AAON, Inc. (NASDAQ: AAON) — A Real Business Wearing a Great-Business Price Tag

Date: June 8, 2026 · Price at analysis: ~$132 · Market cap: ~$11.0–11.5B · EV: ~$11.3B · CIK: 0000824142 Sector: Industrials — Building Products (HVAC / thermal-management equipment) · Fiscal year-end: December

The body of this article (Sections 1–15) takes no buy/sell recommendation and states no price target; valuation is discussed only as embedded expectations and scenarios. The single, deliberate exception is the clearly-labeled Claude’s Take block immediately below.

⚡ Claude’s Take

This block is the author’s own independent, subjective opinion and general information only. It is not investment advice. Everything below this block — the analysis in Sections 1–15 — carries no position and no price target, by design.

Verdict: AVOID at ~$132 / HOLD-don’t-add if owned / accumulate-on-weakness — and explicitly not a short. A genuinely good business that has been re-domiciled into a structurally lower-quality mix, now priced for a near-best-case outcome it has not yet earned. Directional fair-value zone ≈ $85–100 (~24–28× a normalized ~$3.30–3.60 base-case EPS); I’d get constructive in the low-$90s and genuinely interested in the low-$80s — the exact zone where AAON’s own insiders and its corporate buyback actually transacted (purchases $77–82; 2025 buyback ~$81) before selling personally at $135+. At ~$132 there is no margin of safety.

AAON is two businesses stapled together, and the market is paying almost entirely for the second one. The first — premium, semi-custom rooftop HVAC sold through a sticky manufacturer-rep channel — is a real, narrow, demand-side moat that historically threw off ~34% gross margins and ~24% ROIC; but it is now in absolute decline (AAON-branded revenue −2.6% in 2024, −8.3% in 2025). The second — BASX data-center cooling — is a genuine, fast-growing secular winner (backlog +111% to $1.83B, +107% to $2.13B in Q1-26), but it is structurally lower-margin (~24–27% GP), customer-concentrated, cancellable, working-capital-hungry, and a sub-scale follower to Vertiv et al. The whole story is that the high-quality core is shrinking while a lower-quality engine carries the growth — and the stock trades at the 96.5th percentile of its own ten-year valuation history (~93× trailing, ~40× forward, ~45–50× EV/EBITDA) on trough margins, generating no free cash flow (−$204M in 2025; $13K of cash; dividend and buyback funded on the revolver). The framing is quality-compounder-at-the-wrong-price colliding with a late-stage capital cycle: to merely justify today’s price at a sane exit multiple, AAON must roughly quadruple net income to ~$370–440M within five years — a conjunction of sustained 20%-area data-center growth and a margin snap-back toward the legacy peak that the segment mix actively resists (consolidated gross margin was still falling to 25.1% in Q1-2026 despite +54% revenue). The scenario asymmetry is the punchline: bull ≈ $176 (only ~6%/yr even if everything goes right), base ≈ $89, bear ≈ $33. You are being asked to underwrite the bull and handed the bear’s downside.

Conviction: Medium. The demand and backlog are real, which is exactly why this is an avoid, not a short — a doubling order book and an AI-capex tailwind can keep a momentum name expensive far longer than is rational. Flips bullish if consolidated gross margin climbs above ~29% for two consecutive quarters while the data-center mix holds or rises, and free cash flow turns durably positive — that would prove growth and quality can coexist and the 2025 trough was truly transitory. Flips outright bearish (short-worthy) if gross margin stays stuck in the mid-20s through 2026, legacy AAON-branded revenue posts a third straight down year, or a >10%-of-revenue data-center customer cancels/defers and the “firm” backlog visibly shrinks while revolver availability tightens.

One-line tag: “A real business wearing a great-business price tag — the insiders told you the zone: they bought at $80 and sold at $135.”

1. Executive Summary

AAON, Inc. is a Tulsa-based manufacturer of premium, highly configurable commercial and industrial HVAC and thermal-management equipment. Over four years it has been transformed from a single-end-market, semi-custom rooftop-unit (RTU) specialist into a bifurcated enterprise: a mature, high-margin legacy RTU franchise fused — via the 2021 BASX acquisition — to a fast-scaling data-center cooling business. As of FY2025, data-center/BASX-branded product is ~38% of revenue and the entire growth engine; the legacy AAON-branded franchise is in absolute decline. The central analytical question is whether FY2025’s margin and cash-flow collapse is a transient, self-inflicted trough in a quality compounder, or the first clean look at a permanently lower-quality, lower-return business — because the stock is priced almost entirely for the former.

The numbers frame the tension. Revenue grew to $1,442.1M in 2025 (+20.1%), but gross margin compressed from 34.1% (2023) to 26.7% (2025), operating margin halved from 19.5% to 10.1%, and net income fell ~39% from its 2023 peak to $107.6M. Most strikingly, operating cash flow collapsed from $192.5M (2024) to $0.53M (2025) as a ~$236M working-capital build (receivables, contract assets, inventory) consumed essentially all cash earnings; free cash flow was −$204M, cash on hand was $13K, and the company now funds operations — and its dividend and buyback — on a revolver expanded three times in one year to $600M. Beneath the consolidated number, the brand split is decisive: AAON-branded (legacy) revenue fell −8.3% to $894.3M while BASX-branded (data-center) revenue rose +143.5% to $547.8M. The high-margin core is shrinking; the lower-margin engine is carrying everything.

What the analysis concluded. The legacy business has a genuine but narrow and shallow demand-side moat (specification/switching costs, a configure-to-order cost edge, a credible quality reputation) — not a scale moat; AAON is one-fifteenth the size of Trane/Carrier and a single bad year dropped its returns below the commodity giants’. The data-center business is the right growth call but the wrong moat story: lower-margin, project-based, customer-concentrated (three customers each >10% of revenue in 2025), cancellable, and sub-scale against Vertiv, JCI/Silent-Aire, Modine, Munters and STULZ — a textbook late-stage capital cycle into which the whole industry is pouring capacity. The 2025 margin damage is mostly execution/cyclical (an A2L refrigerant transition, a botched Longview ERP go-live, Memphis start-up under-absorption) and Q1-2026 shows the operating line healing — but a structural mix-driven reset is layered underneath, and consolidated gross margin was still falling to 25.1% in Q1-2026. Capital allocation is coherent but, on a per-share-value basis, not yet validated: three years of capex at 2–3× depreciation have coincided with halving margins and returns, an incentive plan that pays for revenue and budgeted profit (no ROIC/EPS/FCF metric), and uniform, discretionary insider selling into the stock’s recovery (≈$51M sold over 24 months vs. ≈$0.6M bought, none above ~$82).

Valuation and verdict-free conclusion. AAON trades at the top of its peer group and the 96.5th percentile of its own decade, on EV/EBITDA (~45–50×) level with the actual data-center-thermal leader, Vertiv — while generating no free cash flow. Embedded-expectations analysis shows that merely to make today’s ~$11.3B EV fair (zero forward return) at a sane 25–30× exit, AAON must roughly quadruple net income to ~$370–440M within five years — requiring ~$2.8–3.7B of revenue at a 15–19% operating margin, i.e. sustained 20%-area data-center growth and a margin recovery toward the legacy peak, two things the segment mix makes difficult simultaneously. A bear/base/bull scenario set spans roughly $33 / $89 / $176, an unfavorable asymmetry in which even a successful base-case partial recovery sits below today’s price. The single observable that will resolve the debate fastest is the trajectory of consolidated gross margin over the next two quarters, read against the direction of the mix. This summary takes no position; the body that follows discusses valuation only as embedded expectations and scenarios. The one explicit opinion in this document is the labeled Claude’s Take block above.

2. Business Overview

AAON, Inc. designs and manufactures premium, highly configurable commercial and industrial HVAC and thermal-management equipment. Founded in 1987 and public since 1992, the company historically built its reputation as a niche specialist in semi-custom rooftop units (RTUs) — equipment engineered to a customer’s exact specification but delivered in 8–12 weeks versus the 16-plus-week lead times of the volume incumbents. Over the past four years the business has been transformed: what was a single-end-market RTU franchise is now a three-segment enterprise in which data-center thermal management is the dominant growth engine and roughly 38% of revenue. Understanding AAON in 2026 requires holding two businesses in view at once — a mature, slow-growth (now shrinking) legacy RTU franchise, and a fast-growing, lower-margin, capex- and working-capital-intensive data-center cooling operation bolted on through the 2021 BASX acquisition.

The three reportable segments. AAON reports along three segments that are organized by manufacturing entity, while products are sold under two brands (AAON-branded and BASX-branded) that cut across the segments — a structure that obscures the underlying mix and must be unpacked (FY2025 10-K, Note 3 / Note 23):

The segment view is deceptive because the data-center business is split across two segments: BASX-branded product is built both at the BASX (Redmond) plant and at AAON Coil Products (Longview). The cleaner lens is the brand disaggregation in Note 3:

This is the single most important fact about the business today: the legacy AAON-branded franchise — the historical moat — has declined in absolute terms for two consecutive years (−2.6% in 2024, −8.3% in 2025), while BASX-branded (overwhelmingly data-center) revenue has compounded ~67% annually since 2022 and now constitutes 38% of the company versus 13% three years ago. Every dollar of net revenue growth in 2024–2025, and then some, came from data center; the core shrank underneath it.

How AAON makes money — the configure-to-order / semi-custom model. For the legacy AAON brand, the economic engine is engineered configurability sold at a premium. AAON manufactures three RTU lines — RQ (2–5 tons), RN (6–140 tons, 26 sizes) and RZ (45–261 tons, 15 sizes) — plus air-handling units, condensing units and geothermal/water-source heat pumps (3–70 tons). Units are built to a customer’s specification before production begins, and the value proposition is “total value proposition rather than initial price” — quality, efficiency, serviceability, and lifecycle cost (FY2025 10-K, Item 1). The premium is real and historically showed up in financial outcomes: AAON Oklahoma carried ~35% segment gross margins as recently as 2024 (35.2% in FY24, 37.3% as a stand-alone in the FY25 comparison) versus a low-single-digit-growth, commoditized RTU market. This is the part of the business that, on the Greenwald taxonomy, plausibly carries a demand-side/customer-captivity advantage (engineered-in specs, rep relationships, faster lead times) — but it is precisely the part now in decline.

For the BASX/data-center business the economics are different and worse on margin: highly customized, job-priced, percentage-of-completion contracts for hyperscale and colocation customers, carrying ~24–27% segment gross margins — structurally below the legacy RTU franchise — and requiring AAON to fund upfront working capital (contract assets, inventory) on multi-phase projects. Pricing is “by the job,” not by published price list. The result is that as BASX scales, consolidated gross margin compresses on mix (34.1% in 2023 → 33.1% in 2024 → 26.7% in 2025), even before the 2025 facility-ramp and ERP-disruption noise.

Recurring vs. non-recurring. AAON is fundamentally a project/equipment business, not a recurring-revenue model. There is no installed-base subscription stream. The closest thing to recurring revenue is aftermarket parts — $80.2M in 2025 (up from $76.9M in 2024 and $67.7M in 2023), or ~5.6% of revenue. This is the highest-quality, most annuity-like slice (replacement parts on a long-lived installed base of AAON equipment), but it is small and growing only mid-single-digits. The BASX data-center revenue is the least recurring: large, lumpy, program-driven orders concentrated in a handful of hyperscaler/colocation customers, with new construction (not replacement) the demand driver. So the mix shift is also a shift away from the more stable replacement/aftermarket-anchored revenue base toward lumpier new-build capex demand.

Distribution — the manufacturer-rep model. AAON-branded products are sold “to property owners and contractors mainly through a network of independent manufacturers’ Representatives,” supplemented by a small internal sales force (FY2025 10-K, MD&A). Management frames this as a deliberate moat: independent reps “attract the most talent,” carry multiple manufacturers’ lines and therefore sell “solutions” rather than captively pushing one brand, and are a more capital-light way to gain share than an owned sales force. The trade-off, which management concedes, is loss of control over the sales process and dependence on a third-party channel (a named risk factor — loss of a major rep could materially hurt results). Critically, BASX does not use the rep model — it sells direct through an internal sales force to “a more concentrated customer base,” which is appropriate for bespoke data-center programs but also explains the rising customer concentration (three customers each >10% of revenue in 2025, up from two in 2024).

The BASX / data-center pivot. BASX, Inc. was co-founded in 2013 by Matt J. Tobolski and acquired by AAON in December 2021 for ~$103M (cash and stock). It is now the strategic center of gravity: in May 2025 Tobolski (the BASX co-founder) became CEO, succeeding Gary Fields, who moved to a board/special-advisor role (8-K, Feb 2025; effective 2025-05-13). The BASX product set has expanded from custom air handlers and cleanrooms into the AI-cooling frontier — in 2025 BASX introduced a proprietary Coolant Distribution Unit (CDU) (the interface between facility water and server cold plates, supporting >100 kW rack densities with N+1 redundant pumps and leak detection) and a water-free Free Cooling Chiller platform for hyperscale heat rejection. The 10-K cites Microsoft, Amazon Web Services, Google Cloud, QTS and Applied Digital as illustrative hyperscale/colocation names in the end-market it serves — not as confirmed ≥10%-of-revenue customers (the filing separately discloses that three unnamed customers each exceeded 10% of revenue in 2025). The data-center channel is reached partly through concentrated rep groups (Texas AirSystems and affiliates Meriton, Ambient, Air Control Concepts). Management has put real capital behind the pivot: a 787,000-sq-ft Memphis, TN plant (purchased December 2024; first profitable quarter Q4 2025; ~$25–30M of quarterly BASX revenue contribution) and a 225,500-sq-ft Longview West Plant expansion completed January 2025, against $204.9M of 2025 capex.

Verdict (Business Overview). AAON is no longer cleanly describable as the high-margin niche RTU compounder of its history. It is a bifurcated business: a mature, premium semi-custom HVAC franchise (≈5% market, low-single-digit growth, ~35% segment margins) that is now in absolute decline, fused to a fast-scaling but lower-margin, working-capital-heavy, customer-concentrated data-center cooling operation that is the entire growth story and is reshaping the consolidated margin structure downward. The rep-based distribution and engineered-configurability model that anchored the legacy moat does not transfer to the direct-sold, job-priced data-center business. The reader should treat AAON as two businesses stapled together — and recognize that the multiple is being paid almost entirely for the second one. The business has been re-domiciled from “niche margin specialist” toward “custom data-center thermal contractor,” and the quality of the franchise should be judged on that basis, not on the legacy AAON story.

3. Industry Dynamics

AAON does not operate in one industry; it straddles two with profoundly different structural physics, and the entire investment debate turns on the collision between them. The first is the mature, replacement-driven North American semi-custom commercial rooftop and applied HVAC market — a structurally attractive but slow-growing niche where AAON has a genuine, if narrow, franchise. The second is the data-center thermal-management market — a violently growing, capital-flooded, AI-fueled arena where AAON, through its 2021 BASX acquisition, is a sub-scale follower chasing a tailwind it does not control. Reading AAON as “an HVAC company” misses the point: as of FY2025 it is a legacy HVAC franchise being reshaped, financially and strategically, by data-center demand. This section maps both, sizes the profit pools, names the competitive set, and applies the Marathon capital-cycle lens to ask the question that matters most for forward returns: is capital flooding into data-center cooling, and what does that imply?

3.1 Market Structure and Size — Three Distinct Profit Pools

Pool 1 — North American commercial/industrial rooftop units (RTUs). This is AAON’s heritage. The North American HVAC rooftop-unit market was roughly $11.1B in 2024, growing at a ~5.2% CAGR through 2033 (Grand View Research). It is a mid-single-digit, GDP-plus market whose growth is anchored not in new construction but in replacement and regulatory upgrade — the installed base of commercial rooftops turns over on a ~15–20 year cycle, and successive efficiency standards force premature replacement (§3.4). AHRI shipment data shows variable-capacity systems up ~27% 2020–2025, with ~5.6M units shipped in the trailing year — a large, fragmented, steady pool. (FACT: Grand View Research, AHRI, 2024–2025 data.)

Pool 2 — Applied/custom air handling and cleanroom systems. A smaller, specification-driven segment of engineered air handlers, makeup-air, energy-recovery, and cleanroom units. This is where AAON’s “configure-to-order, build-before-production” model has historically commanded premium pricing. It overlaps physically with the data-center pool because the same custom air-handling competencies feed hyperscale facilities.

Pool 3 — Data-center cooling — the wave. This is the structural story of the decade and the reason AAON trades at ~95x trailing earnings. The global data-center cooling market was ~$26.3B in 2025, growing at a ~22% CAGR toward ~$128B by 2033; the US market alone was ~$6.58B in 2025 (~16.7% CAGR); and the highest-growth sub-segment, data-center liquid cooling, was ~$5.5B in 2025 and is projected to reach ~$19B by 2030 (~23% CAGR) (FACT: Grand View Research; Virtue Market Research, 2025–2026). The demand driver is unambiguous and quantified in AAON’s own 10-K: between 2022 and 2025, US data-center put-in-place construction spend rose ~240% (FACT: aaon-20251231.htm), against a backdrop of hyperscalers (Microsoft, Amazon, Google, Meta) spending $380B+ on AI infrastructure in 2025 (FACT: industry press). GPU-dense AI racks have pushed thermal densities past the limit of traditional air cooling, accelerating direct-to-chip liquid cooling and rear-door heat exchangers — precisely the BASX product line.

The numbers below frame AAON’s relative scale. It is, decisively, not a scale leader — its entire enterprise is ~7% of Trane’s revenue and smaller than Lennox’s commercial segment alone.

3.2 Competitive Intensity and the Named Competitor Set

AAON’s FY2025 10-K names its competitors explicitly, and the bifurcation is stark. (FACT: aaon-20251231.htm, Item 1, “Competition.”)

• Comfort cooling (legacy RTU): Lennox, Trane, York Light Commercial (Bosch Home Comfort), Johnson Controls, Carrier, and Daikin (Goodman). These are the global majors — each 10–20x AAON’s size — but they compete primarily on standardized, lower-featured, price-led product. AAON’s own language is telling: “We compete on total value proposition rather than initial price… product quality, performance, efficiency, serviceability, reliability, and lifecycle cost of ownership.” It explicitly concedes it competes against “larger manufacturers with greater financial and operational resources.”
• Thermal management (data center): Vertiv (VRT), STULZ, Munters, Silent-Aire (Johnson Controls), Nortek Air Solutions, and Modine (MOD). This is a far more dangerous arena. Vertiv is the #1 thermal-management player, exiting 2025 with a ~$15B backlog (more than double a year earlier), increasingly winning system-level contracts that bundle power and cooling, and rolling up the value chain via >$1B of M&A (PurgeRite, ~$1B Dec-2025; ThermoKey; BMarko). Modine guided its data-center business to +50–70% annual growth with ~5-year demand visibility. (FACT: Vertiv, Modine FY25/FY26 releases.)

The competitive-intensity verdict differs by pool. In legacy RTU, the structure is moderately favorable: a small number of large rational players, high regulatory and certification barriers (AHRI ratings, UL, A2L safety standards), and a sticky specification/replacement channel. In data-center cooling, intensity is rising fast and the field is crowded — the majors (Vertiv, JCI/Silent-Aire), the dedicated thermal players (Modine, Munters, STULZ, SPX), and AAON’s BASX are all expanding into the same hyperscaler demand. AAON sits as a sub-scale, custom/engineered niche supplier against Vertiv’s scale-and-systems leadership — a price-taker, not a price-maker, in this pool.

3.3 Value Chain, Barriers to Entry, and Switching Costs

The North American commercial HVAC value chain runs manufacturer → independent manufacturer’s representative → mechanical/HVAC contractor → building owner. AAON sells through independent rep organizations (e.g., Texas AirSystems, named in the 10-K) plus an internal sales force. This structure is the source of AAON’s narrow moat — and also of a key strategic asymmetry between its two markets:

• Legacy RTU — real, demand-side switching cost. AAON’s edge is concentrated in the replacement and owner-controlled purchase segment, where the building owner (not the contractor) makes the decision and weighs lifecycle/total-cost-of-ownership. The 10-K states AAON has “consistently gained market share by demonstrating superior total cost of ownership.” Once an owner standardizes on AAON’s serviceable, durable, semi-custom units, re-specifying is costly. This is a Greenwald demand-side captivity + intangible (specification) advantage — genuine but local and narrow, and it weakens sharply in new construction where contractors optimize for lowest initial price.
• Data center — no structural switching cost. Hyperscaler and colocation buyers are large, sophisticated, multi-source, and price-disciplined. AAON itself warns that data-center orders “have more risk and we often see shifts in timing, cancellations and re-issuances of orders.” (FACT: aaon-20251231.htm.) There is no rep-channel lock-in, no installed-base captivity — these are project-by-project competitive bids against Vertiv et al.

Barriers to entry are moderate-to-high in both pools (capital intensity, engineering talent, certification, refrigerant-safety compliance), but they protect incumbents collectively far more than they protect AAON individually — and in data-center cooling the incumbents AAON must out-compete are vastly larger.

3.4 The Regulatory/Standards Engine of Replacement Demand

The single most attractive structural feature of the legacy HVAC market is that regulation manufactures recurring, non-discretionary replacement demand and continually obsoletes the installed base. Three overlapping forces are live in 2025–2026:

1. DOE efficiency standards. The 2023 commercial RTU standard (effective 1/1/2023) raised minimum efficiency ~15% over the 2018 level; a further ACUAC/ACUHP step has a recommended 1/1/2029 compliance date. Each step forces the installed base toward newer, higher-margin equipment. (FACT: Federal Register; ACHR News.)
2. AIM Act HFC phasedown + A2L refrigerant transition. The AIM Act mandates an 85% HFC reduction by 2036; effective 1/1/2025, new equipment must use refrigerants below 750 GWP, ending R-410A in favor of mildly-flammable A2L blends (R-454B, R-32). (FACT: EPA AIM Act; ACHR News.) This is a double-edged sword: a multi-year replacement tailwind, but the transition itself drove destocking, pre-buy distortion, and ramp inefficiency in 2024–2025 that is part of AAON’s margin story.
3. ASHRAE 90.1 / IECC building codes continually ratchet baseline efficiency, favoring premium, high-efficiency manufacturers — structurally AAON’s positioning.

Interpretation: the regulatory regime is a durable demand machine for the legacy business — it is why a 5% market is investable. But it confers no special advantage in data-center cooling, where demand is driven by AI capex, not code cycles.

3.5 Marathon Capital-Cycle Lens — Is Capital Flooding In?

This is the decisive analytical question for forward returns, and the evidence is unambiguous: capital is flooding into data-center thermal at or near a cyclical peak — the classic Marathon late-cycle signature. High returns have attracted capital, and on the supply side every credible player is racing to add capacity:

• Vertiv: $15B backlog, >$1B of bolt-on M&A in 2025–2026, a fresh $2.5B revolver and $2.1B notes issuance to fund expansion.
• Modine: record intake, ~5-year visibility, +50–70% annual DC growth, building out capacity.
• AAON itself: capex of $213M (2024) + $205M (2025) — roughly $418M over two years against a ~$1.2–1.4B revenue base — including a new 787,000 sq ft Memphis facility dedicated largely to BASX data-center production, plus Longview and Tulsa expansions. (FACT: aaon-20251231.htm; AAON releases.) This capex is the proximate cause of AAON’s cash collapse: FY2025 operating cash flow fell to $0.53M while total assets jumped +$511M.

The Marathon framework warns that when an entire industry simultaneously expands capacity to chase abnormal returns, future returns mean-revert as the new supply competes margins down. We already see the early footprints in AAON’s own numbers: BASX gross margin of 26.6% and Coil Products at 21.4% versus the legacy Oklahoma franchise’s historical ~37%. The data-center mix is diluting AAON’s franchise economics even as it inflates the top line. The bull case requires demand to outrun capacity for years; the capital-cycle base rate says the window for outsized margins in a capacity-flooded segment is finite. A second-order risk: AAON’s own disclosure that data-center orders are cancellable and frequently re-timed means the eye-popping $1.83B backlog (+111%) is lower-quality than the historically-firm commercial-AC backlog — capacity built for cancellable orders is the textbook setup for a capital-cycle disappointment.

3.6 Verdict

Structurally, this is a tale of two industries, and AAON is well-positioned in the wrong-shaped one. The legacy North American semi-custom commercial RTU market is a structurally GOOD business — moderate concentration, high regulatory/certification barriers, durable replacement demand engineered by DOE/AIM-Act/ASHRAE cycles, a sticky owner/replacement channel, and genuine premium economics (~37% gross margins historically). But it grows only mid-single-digits, and the disconfirming evidence is glaring: AAON’s own legacy Oklahoma segment shrank 6.7% and lost 830bps of gross margin in 2025 — the franchise is being neglected or disrupted, not compounding. The data-center cooling market is structurally explosive but late in its capital cycle, and here AAON is a sub-scale follower with no durable advantage against a far larger, better-capitalized, vertically-integrating Vertiv and a crowded field of dedicated thermal specialists; its growth is real but margin-diluting and order-cancellable. Net: AAON rents a powerful secular tailwind it does not control, at the cost of its franchise margins, just as the whole industry floods the segment with capacity. The industry backdrop is favorable enough to explain the revenue surge but not durable enough to justify pricing data-center hyper-growth as permanent and margin-accretive — which is precisely what the ~45x EV/EBITDA multiple does.

Verdict: Bifurcated — a structurally good but slow (and, for AAON, shrinking) legacy niche fused to a structurally hot but capital-flooded, late-cycle data-center arena where AAON is a price-taking follower. AAON is well-positioned in the narrow business it is partly abandoning and poorly-positioned (sub-scale) in the large one it is chasing.

4. Competitive Position

Verdict in one line: AAON owns a narrow but genuine demand-side moat (switching/specification costs plus a configure-to-order cost-efficiency edge) in its legacy semi-custom rooftop franchise — but that moat is being mix-shifted, not competed, away as the company pivots capital and capacity into data-center cooling (BASX), a structurally lower-margin, more-competitive, project-based business where AAON is a sub-scale niche entrant with no demonstrable advantage. The 2025 margin collapse is mostly execution/cyclical, but it exposed how thin the franchise economics become once the premium-rooftop mix is diluted.

7.3.1 Naming the moat in Greenwald’s taxonomy

Greenwald’s Competition Demystified recognizes only three genuine competitive advantages — (i) supply/cost advantages, (ii) demand-side advantages (customer captivity), and (iii) economies of scale combined with captivity — and demands that any claimed moat survive two tests: market-share stability and persistently above-cost-of-capital ROIC. AAON’s legacy business clears both historically; it does not possess the one advantage the popular narrative assigns it.

What AAON does NOT have: economies of scale. This is the single most common misreading of the name. AAON’s FY2025 revenue was $1.44B (10-K, FY2025, Consolidated Results of Operations). Its named comfort-cooling competitors — Lennox (~$5.2B FY2025), Trane (~$21B), Carrier (~$20–22B), Johnson Controls, Daikin — are 4× to 15× larger (10-K Item 1, “Competition”; peer FY2025 results). AAON management says so explicitly and repeatedly: it competes “against substantially larger manufacturers” with “greater financial and operational resources” (FY2025 10-K, “People and Culture”; “Competition”). A scale moat requires being the largest operator in a market with high fixed costs and customer captivity. AAON is a fractional-share specialist. FACT/INTERPRETATION: AAON has no scale advantage; the giants have it (and AAON’s own filings concede it). Any thesis resting on “scale” is wrong.

What AAON actually has: a demand-side (captivity) advantage rooted in switching/specification costs, reinforced by a niche cost-efficiency edge. Three mechanisms, each tied to a financial outcome:

1. Specification lock-in / engineer-and-rep captivity. AAON sells “highly configurable,” “build-to-order,” “pre-specified before production” equipment through a network of independent manufacturer-representative organizations (FY2025 10-K, “Business and Marketing Strategy,” “Customers First”). Units are engineered into a building’s mechanical design (rooftop footprint, controls, ductwork, electrical) before the order is placed. Once an AAON unit is spec’d by the consulting engineer and installed, the replacement unit is overwhelmingly likely to be AAON again — the curb, footprint, controls protocol (VCCX, BACnet), and rep relationship are already in place. Management states it “consistently gained market share” specifically “in replacement markets and owner-controlled purchases… by demonstrating superior total cost of ownership over equipment lifespans” (FY2025 10-K, “Competition”). This is textbook customer captivity (high search/switching cost), not brand fluff. FACT.

2. A configure-to-order cost-efficiency edge (“mass semi-customization”). AAON’s stated strategy is to combine “the cost efficiency of scaled production with the precision of individual customization” via flexible CAM systems (FY2025 10-K, “Business and Marketing Strategy”). The economically meaningful claim — repeated in both the FY2023 and FY2025 10-Ks — is that “operational efficiency improvements… narrowed the price gap between our semi-custom equipment and competitors’ standardized offerings,” letting AAON win in new construction (where the contractor optimizes initial price) as well as replacement (FY2025 10-K, “Competition”; FY2023 10-K, “Competition”). If true, this is a supply-side/process cost advantage in a defensible niche: AAON delivers near-custom configurations at near-commodity cost and short(er) lead times (AAON-branded lead time 18–26 weeks; 10-K “Backlog”). INTERPRETATION: this is a process edge, not a structural cost moat — it is replicable in principle, and the giants are investing to close it.

3. Reputation/serviceability and a genuinely differentiated R&D/test asset. AAON competes “on total value proposition rather than initial price, emphasizing product quality, performance, efficiency, serviceability, reliability, and lifecycle cost” (FY2025 10-K, “Competition”). The Norman Asbjornson Innovation Center (NAIC) — an AMCA-accredited lab able to test up to a 300-ton AC system / 540-ton chiller under full environmental load, with “the largest sound-testing chamber in the world” for HVAC, capabilities AAON says “exist nowhere else in the world” (FY2025 10-K, “Research and Development”) — is a real intangible asset that lets AAON pre-prove performance on large units that exceed the 63-ton ceiling of standard AHRI/DOE certification. FACT (capability) / INTERPRETATION (durability of the resulting reputation premium). Reputation is the weakest leg: it must be re-earned every cycle and is not, by itself, a Greenwald moat unless it produces measurable customer captivity — which here it does, via the spec/replacement loop above.

Moat type, stated plainly: Demand-side customer captivity (switching/specification costs) in semi-custom commercial rooftop and indoor packaged units, reinforced by a niche process-cost edge and a credible quality/serviceability reputation. Not scale. Not network effects (there are none — pressure-tested below). Patents are explicitly immaterial (“no single patent is material”; FY2025 10-K, “Patents, Trademarks…”), so this is not an IP moat either.

Pressure-test — network effects: none exist. There is no two-sided platform, no installed-base data flywheel that improves the product for the next buyer, no developer ecosystem. The rep channel is a distribution asset (and a risk — see below), not a network effect. The IoT/controls “AI” initiatives are product features, not a network. Any claim of network effects fails. FACT.

7.3.2 The financial fingerprints of the moat — and the 2025 falsification test

A real moat must show up as above-peer returns on capital and a pricing premium. Historically, AAON’s did, emphatically:

In 2023 AAON earned a 34.1% gross margin and ~24% ROIC — well above the ~30% gross margins typical of the volume giants and comfortably above any reasonable ~9–10% WACC. That spread is the moat: customers paid a premium price and AAON converted it to high returns on a relatively asset-light, working-capital-driven model. The market-share-stability test passes in legacy rooftop: management documents consistent share gains in replacement and owner-controlled purchases across cycles (FY2023 and FY2025 10-Ks), and Q3-2025/Q1-2026 commentary cites continued “share gains” (AAON Q3-2025 release; FY2026 10-Q MD&A). The ROIC test passed decisively 2021–2024.

Then 2025 broke it — and the falsification test is the crux of this section. Consolidated gross margin fell 34.1% → 26.7%, operating margin 19.5% → 10.1%, ROIC 24% → 9.5% — i.e., to roughly its own cost of capital. If this were competitive erosion (price being competed away), the moat thesis would be falsified. The evidence says it is mostly execution/cyclical with a real structural mix component layered underneath — a nuanced answer the bulls and bears both oversimplify:

• Execution/cyclical (transitory) drivers, ~the majority of the hit. Management attributes the 2025 compression to: (i) an ERP go-live disaster at the Longview, TX coil plant on 2025-04-01 that “caused some disruptions,” cut coil supply to Tulsa, and “impacted AAON Oklahoma’s ability to ramp up production” (FY2025 10-K, Results highlights) — third-party coverage quantifies a ~50% cut to AAON-branded production and a 37%→28% consolidated GM swing (BeyondSPX/Investing.com, 2025); (ii) the R-454B refrigerant (A2L) transition effective 2025-01-01 that disrupted the supply chain; (iii) Memphis plant startup overhead ($16.1M of cost absorbed in the AAON Oklahoma segment in 2025 for intercompany sales booked at cost in BASX; FY2025 10-K segment MD&A); and (iv) sub-optimal overhead absorption from an 8.3% decline in AAON-branded sales on soft non-residential construction. None of these is competitive. FACT (management attribution) — validated: the recovery signature is already visible. Consolidated GM recovered sequentially to 27.8% in Q3-2025 (from 26.6% in Q2; AAON Q3-2025 release), and AAON Oklahoma segment GM rose to 26.3% in Q1-2026 “due to realization of price increases and increased volume” (FY2026 10-Q MD&A). Pricing power is intact — the franchise is taking price and customers are paying it. That is the single most important disconfirming fact against the “moat is eroding” bear case.

• Structural (durable) driver: mix-shift, not margin erosion. Even fully recovered, AAON’s blended margin will likely sit below the old 34% peak, because the growth — and the backlog — is overwhelmingly BASX data-center cooling, a lower-margin business. Segment gross margins (FY2025 10-K; FY2024 10-K): legacy AAON Oklahoma ~35–37% at peak vs BASX ~25–27% and AAON Coil ~19–21%. Consolidated backlog exploded to $1,828.5M at YE2025 (+110.9% YoY) and $2,129.5M at Q1-2026 (+107% YoY), of which BASX-branded backlog is ~$1.30B (YE2025), up 141%, “most… associated with BASX-branded data center liquid cooling” (FY2025 10-K “Backlog”; FY2026 10-Q). The premium-rooftop franchise (where the moat lives) is becoming a minority of the order book. Management’s own FY2026 GM guide of 29–31% — up from 26.75% in 2025 but well short of the 34% peak — implicitly concedes a structurally lower steady-state mix. INTERPRETATION (well-supported): the moat is being diluted by deliberate mix-shift, not competed away.

7.3.3 AAON vs. Carrier / Trane / Lennox — where it wins and loses

The damning comparison: historically AAON’s gross margin sat above the giants (34% vs ~30%) — the moat premium — even though its operating margin was only roughly in line (19% in 2023 vs Lennox ~20%), because AAON spends more on R&D and SG&A per dollar of revenue given its sub-scale (R&D rose to $58.2M / 4.0% of sales in 2025; SG&A to $239.5M / 16.6%). In 2025 the comparison inverted and got ugly: AAON’s operating margin (10.1%) and ROIC (9.5%) now sit well below Lennox (20.0% / ~29%), Trane (18.6%), and Carrier (~16.5%), all of which held or expanded margins in 2025 (Lennox GM steady at 33.4%, OM +50bps; Trane OM +100bps). The peers did not have AAON’s self-inflicted ERP/refrigerant/startup year — which is consistent with AAON’s problems being idiosyncratic and transitory rather than industry-wide competitive pressure. But it also proves AAON has no margin-of-safety buffer: a single bad operating year drops a “premium niche” compounder below the commodity giants on every return metric. The moat is real but shallow — it generates a price premium, not the structural cost-and-scale fortress the giants enjoy.

7.3.4 Is BASX/data-center a different (weaker) business that dilutes the moat? Yes.

This is the most important forward question, and the Marathon Capital Returns lens sharpens it. Data-center liquid cooling is the textbook late-stage, capital-attracting boom: the market was ~$10–26B in 2025 growing 15–22%+ CAGR, drawing in every well-capitalized player — Vertiv (>11% liquid-cooling share), Schneider Electric, Stulz, Munters, Johnson Controls/Silent-Aire, Rittal, Boyd, Modine, Nortek (AAON 10-K “Competition”; industry data). AAON/BASX is a sub-scale niche entrant here with no demonstrable advantage: it competes against firms with deeper capital, broader thermal-chain portfolios, and incumbent hyperscaler relationships. The business is also structurally lower-quality than legacy rooftop on multiple axes the filings make explicit:

• Lower margin: BASX GM ~25–28% vs legacy AAON Oklahoma 35–37% at peak (segment data).
• Project-based and cancellable: “Orders for the data center market have more risk and we often see shifts in timing, cancellations and re-issuances” (FY2025 10-K “Backlog”); BASX has “lengthy and unpredictable procurement processes” and “large orders with short lead times” that make results “less predictable” (Item 1A risk factor).
• Customer-concentrated: AAON had three customers each ≥10% of revenue in 2025 (up from two in 2024), and three ≥10% of receivables (FY2025 10-K, “Key Customers”/Concentration). The filing names Microsoft, AWS, Google Cloud, QTS and Applied Digital only as illustrative hyperscale/colocation buyers in the end-market — it does not identify them as the ≥10% customers. Hyperscalers are monopsony-grade buyers with their own engineering staffs — they erode supplier captivity, the exact mechanism that gives AAON its moat in fragmented commercial HVAC. A handful of sophisticated hyperscalers spec’ing their own cooling is the antithesis of the thousands-of-small-buyers, rep-locked, replacement-driven dynamic that makes legacy AAON sticky.
• More working-capital and capex hungry: the over-time, milestone-billed contract model drove contract assets up and helped collapse operating cash flow to $0.53M in 2025 despite $107.6M net income — a quality-of-earnings flag the legacy business never produced.

Conclusion on BASX: it is a faster-growing but lower-quality, lower-moat business than legacy rooftop. It is the right growth call (the demand is real and AAON’s engineering/test assets are genuinely transferable to custom air handling) but the wrong moat story. As BASX scales from ~22% to a stated ~$1B revenue target and dominates the backlog, the company’s weighted-average moat strength declines even if every individual unit is well-engineered. INTERPRETATION (high confidence).

7.3.5 The rep channel — moat and fragility

AAON’s go-to-market via independent manufacturer reps is double-edged. Management frames it as an advantage: it “is a more effective way of increasing market share” because the entrepreneurial channel “attracts the most talent” and sells solutions, not just one brand (FY2023 10-K). It deepens captivity (the rep owns the local relationship and service). But it is also a named risk: AAON is “dependent on our third-party representatives,” “certain competitors with greater financial resources… have targeted some of our third-party representatives for exclusive sales channels,” and losing a major rep “could materially and adversely affect” revenue (FY2025 10-K, Item 1A). The disclosed channel-partner names (Texas AirSystems, Meriton, Ambient, Air Control Concepts — some under common ownership) concentrate the risk. A moat that can walk out the door with a rep group is a contingent moat, not an owned one.

7.3.6 Greenwald synthesis and the verdict

Running the full Greenwald checklist: barriers to entry in legacy semi-custom commercial rooftop are moderate — real (spec lock-in, rep network, the NAIC test moat, the configure-to-order process edge) but not insurmountable, since the giants out-resource AAON in R&D and are actively narrowing the price gap. Market-share stability: passes in legacy rooftop (durable replacement gains), unproven and contested in data center. ROIC test: passed 2021–2024 (peak ~24%), failed in 2025 (~9.5%, ≈ WACC) — but the failure is attributable to identifiable, transitory, self-inflicted causes (ERP, A2L, Memphis ramp) plus the structural mix-shift, not to lost pricing power, which Q1-2026 price realization confirms is intact. EPV vs. asset value: at the 2023 earnings level the earnings-power value sat far above asset value (a franchise); at the 2025 trough, EPV compressed toward asset value (a fair business) — the swing itself measures how thin the moat buffer is.

VERDICT: A durable but narrow and shallow moat in the legacy franchise — demand-side captivity plus a niche cost-efficiency edge — that is being mix-shifted (not competed) away as the company pivots into lower-moat data-center cooling. The premium-rooftop business still has real pricing power and replacement stickiness; that is genuine and survives the 2025 stress test (price increases are sticking, share is still being gained). But three sober conclusions follow. (1) This was never a scale moat and never will be — AAON is a fractional-share specialist in a market dominated by 4×–15× larger rivals, and a single bad operating year was enough to drop its returns below those of the commodity giants, proving the buffer is thin. (2) The 2025 collapse is mostly cyclical/execution and is recovering, so the moat is not eroding in the competitive sense — the bear’s “premium competed away” thesis is not supported by the pricing evidence. (3) But the company is deliberately diluting its own moat by routing growth and capital into BASX/data-center, a lower-margin, project-based, customer-concentrated, capital-hungry business where it has no demonstrable advantage against Vertiv/Schneider/Stulz/JCI. The honest characterization is therefore “a premium niche being mix-shifted — not competed — away,” leaving a blended franchise whose steady-state return on capital and margin will likely settle materially below the 2023 peak even after the execution issues clear.

Disconfirming evidence weighed: the strongest counter to this verdict is the FY2026 GM guide (29–31%) plus the $2.13B backlog and Q1-2026 price realization, which a bull reads as “moat fully intact, 2025 was a one-off, re-rate to the old margins.” That is partly right on the execution recovery — but it elides that the backlog is ~70%+ BASX/data-center, so a “recovery” to 29–31% GM is itself the structural-dilution result (a step down from 34%), not a refutation of it. The franchise is good; it is becoming a smaller share of a larger, lower-quality whole.

5. Growth History and Forward Opportunities

Multi-year growth: high magnitude, deteriorating quality. AAON grew revenue from $534.5M (2021) to $1,442.1M (2025) — a 28.2% four-year CAGR — but the composition of that growth changed character completely, and disaggregating it is the whole exercise. The headline trajectory masks three distinct phases:

The 2022–2023 surge was substantially a price event, not a volume/share event. Management implemented a recurring 1% monthly price increase from June 2022 through April 2023 and reinstated it October 2023 through February 2024 (FY2024 10-K, MD&A) — i.e., roughly 17–18 months of compounding ~1%/month list-price increases during the inflation spike. Stripping that out, underlying volume growth in the legacy franchise was far weaker than the reported 66%/32%, and as soon as pricing normalized and the A2L refrigerant transition hit, AAON-branded revenue went negative. The 2024 stall (+2.7%) and 2025 AAON-branded decline (−8.3%) are the tell: the legacy franchise has very little organic volume growth. On a brand basis the AAON-branded CAGR 2022→2025 was only ~5.1% (and turned negative in the last two years), versus a ~67% CAGR for BASX-branded.

Organic vs. acquired. Nearly all durable growth is acquired and then pivoted — BASX (the 2021 acquisition) is the platform on which the data-center business was built, and BASX-branded revenue went from $117.7M (2022) to $547.8M (2025), rising from 13.2% to 38.0% of total revenue. There have been no acquisitions since BASX (none 2022–2024), so post-2021 the data-center scaling is technically “organic” in the GAAP sense — but it is organic growth of an acquired franchise into a new end market, not organic compounding of the historical AAON moat. The legacy AAON-branded business, the actual organic core, has shrunk. This matters for quality: the growth is not the niche RTU specialist taking durable share in its protected market; it is a small custom-cooling business riding the AI data-center capex wave.

Backlog trajectory — genuine visibility, but concentrated and conversion-lagged. AAON’s backlog has more than doubled and provides real forward visibility, which is the strongest pillar of the bull case:

BASX-branded backlog grew 141.3% in 2025 and 160% YoY at Q1 2026, reaching 76% of total backlog — a near-2.4x book-to-bill on the BASX brand in FY2025 (my reconciliation: ($1,302.1M − $539.7M + $547.8M shipped) / $547.8M = 2.39x, matching management’s stated 2.4x). The majority of backlog is “firm” and expected to convert within 12–18 months. But three caveats temper the signal. First, AAON-branded backlog has stopped growing — it ticked down sequentially from $526.4M (YE25) to $509.8M (Q1-26), confirming a soft legacy commercial market (“flattish in 2026,” per management). Second, data-center orders carry more cancellation/deferral risk than the historically near-uncancellable RTU orders (the 10-K explicitly flags this: data-center orders “have more risk”). Third, and most telling, management’s own 2026 guidance implicitly discounts the backlog: against +141% BASX backlog growth, management guides BASX to only ~25% revenue growth in 2026, because the backlog is “longer-duration, multi-phase projects” that ship into 2027+. The $2.1B backlog is not $2.1B of near-term convertible revenue; it is a multi-year pipeline whose timing the company does not fully control.

The data-center demand driver and TAM. The structural tailwind is real and large. US data-center construction put-in-place spending rose ~240% from 2022 to 2025 by management’s count — corroborated externally: data-center construction grew over 100% in two years (and 344% from 2020) to ~$41B in 2025, reaching a ~$50.7B seasonally-adjusted annual rate by April 2026, when it became the largest single segment of US private office construction (US Census via Wolf Street / Marketplace, Feb 2026). The relevant equipment TAM is the data-center cooling / liquid-cooling market, which third-party estimates size at ~$5.1–6.7B in 2025 growing to ~$16–29B by 2030–2033 at ~20–26% CAGRs (Grand View Research, MarketsandMarkets, MarknTel; Dell’Oro pegs broader data-center physical infrastructure at ~$61B by 2029, ~14% CAGR). Rack power densities are migrating from ~15 kW toward 60–120 kW for AI workloads, forcing the air-to-liquid transition that AAON’s CDU and free-cooling-chiller products target. Set against this, AAON’s legacy market — US commercial rooftop units — is ~$2.56B by 2030 growing only ~4.9% (Grand View). The growth math is unambiguous: data center is a ~20%+ CAGR pool an order of magnitude larger and faster than AAON’s mature core. That is why the company has pivoted, and why the market is paying up.

The A2L refrigerant transition — a 2025 distortion, not a durable driver. The EPA AIM Act forced the commercial/light-commercial transition from R-410A to low-GWP A2L refrigerants (R-454B) effective January 1, 2025. AAON “completely transitioned” to R-454B. This created genuine 2025 disruption — supply-chain shortages, A2L cylinder prices up >300%, and the need for new sensors/detectors — and AAON cites the refrigerant change as a direct cause of weak AAON-branded volume and supply constraints in H1 2025 (FY2025 10-K, MD&A). Industry-wide, the transition drove a pre-buy/destocking dynamic (contractors hoarding pre-2025 R-410A inventory), which pulled some demand forward into 2024 and depressed 2025. The investment implication is that part of the 2025 legacy weakness is transitory (a destocking air-pocket that should normalize), but it does not create durable incremental demand — it is a one-time regulatory reset, not a multi-year growth driver, and should not be capitalized into a recovery thesis beyond a modest 2026 normalization.

New-facility capacity — the supply-side response. AAON has committed heavy capital to capacity for the data-center surge: the 787,000-sq-ft Memphis, TN plant (bought Dec 2024, first profitable quarter Q4 2025, ramping toward ~$25–30M/quarter of BASX revenue, though its overhead currently burdens the AAON Oklahoma segment by ~$10–16M as intercompany product is built at cost); the 225,500-sq-ft Longview, TX West Plant expansion (completed Jan 2025); and the Redmond, OR BASX home base. 2025 capex was $204.9M. Through a Marathon capital-cycle lens this is the textbook mid-boom signature: extraordinary returns in data-center thermal are pulling capital and capacity in across the industry — AAON, Vertiv, Modine, Johnson Controls (Silent-Aire), STULZ and Munters are all expanding — which is precisely the condition that historically precedes margin mean-reversion. AAON is currently capacity-constrained (a positive near-term), but the supply response it is part of is the medium-term risk.

Forward drivers and their credibility. Q1 2026 showed the recovery underway: revenue +54.3% to $496.9M (BASX segment +104.5%; AAON Oklahoma +50.7% off a refrigerant-depressed Q1-25 base), with operating margin recovering to 11.5%. Management guides 2026 to 18–20% revenue growth and 29–31% gross margin (vs 26.7% in 2025) — i.e., a margin recovery is underwritten alongside continued growth. The credible drivers: (1) the data-center backlog/book-to-bill is real and gives 12–18-month-plus visibility; (2) Memphis and Longview capacity is now online and ramping; (3) a legacy normalization as A2L destocking passes. The less-credible / unproven elements: the durability of AI data-center capex beyond the current cycle; AAON’s ability to hold or win durable share against far larger thermal competitors as capacity races in; the assumed gross-margin recovery to 29–31% while mix continues shifting toward lower-margin BASX product; and the conversion of a multi-phase $2.1B backlog whose timing the company explicitly cannot pin down (hence the 25% BASX guide against 141% backlog growth).

Verdict (Growth quality): LOW-to-MIXED quality — high magnitude, low durability. The growth is large but fails most quality tests. It is (1) single-end-market and capital-cycle-driven (AI data-center capex), not diversified or replacement-anchored; (2) acquired-and-pivoted (BASX), not organic compounding of the historical moat — which is itself in absolute decline (−8.3% AAON-branded in 2025, ~5% CAGR at best); (3) margin-dilutive, as the mix shift toward ~24–27%-GM BASX product compressed consolidated gross margin ~740 bps from peak; (4) working-capital- and capex-intensive, with the growth funded by a contract-asset/inventory build that collapsed operating cash flow to ~$0.5M in 2025; and (5) customer-concentrated (three >10% customers). The 2022–2023 “growth” was substantially a transitory price event, and the 2024–2025 growth is a derivative of a handful of hyperscaler programs. Backlog visibility is genuine — the single strongest counter-argument — but management’s own guidance signals the backlog overstates near-term convertible revenue. The disconfirming evidence weighed (real backlog, real TAM, real capacity, a credible Q1-26 reacceleration) is meaningful but does not change the character of the growth: this is not the durable, high-ROIC, organic compounding that AAON’s ~95x trailing / ~41x forward multiple implies — it is a cyclical, lower-margin capex-beneficiary whose legacy core is eroding underneath the data-center surge.

6. Financial Quality

All figures reconciled to AAON’s FY2025 Form 10-K (filed 2026-03-02, period 2025-12-31) and Q1-2026 Form 10-Q (filed 2026-05-07, period 2026-03-31). Segment figures are post-intercompany-elimination as presented by the company.

AAON’s reported income statement tells a clean story of a compounder that hit a wall in 2025; the cash-flow statement and balance sheet tell a more complicated one. The right way to read this business is to separate three distinct things that all happened at once in FY2025: (1) a genuine, partly self-inflicted margin trough driven by under-absorption, a refrigerant transition, and an ERP go-live; (2) a structural mix shift toward lower-margin, project-based, working-capital-hungry data-center cooling (BASX); and (3) a cash-flow collapse that is largely — but not entirely — a backlog-funding artifact rather than an earnings-quality fraud signal. The verdict hinges on disentangling the transient from the permanent.

1. Revenue and margin trajectory (five-year)

The shape is unmistakable. Margins peaked in 2023 — a COVID-pricing-and-backlog supercycle that flattered the legacy rooftop business — and have compressed hard since, with the operating margin halving from 19.5% to 10.1% in two years and net income falling ~39% from the 2023 peak despite revenue growing 23% over the same span. (FACT — FY2025 10-K, Consolidated Statements of Income, p.46; D&A per SEC EDGAR XBRL DepreciationDepletionAndAmortization: $46.5M/$62.7M/$79.2M for 2023/24/25.) The FY2024 step (34.1%→33.1%) was a benign give-back of peak pricing; the FY2025 step (33.1%→26.7%) is the one that demands a forensic bridge.

2. Gross-margin bridge: 34.1% → 33.1% → 26.7%

The company gives segment-level gross profit, which lets us build a real bridge rather than rely on the MD&A’s narrative. The decisive year is FY2025, where gross profit fell only $11.4M (-2.9%) in dollars but 640bps in rate. Decomposing by segment (FY2025 10-K, restated comparatives):

The bridge resolves into four forces, in order of magnitude:

• AAON Oklahoma under-absorption (the dominant driver). Legacy rooftop GM collapsed from 37.3% to 29.0% as segment volume fell 6.7%. With largely fixed factory overhead, lost volume in the first half — caused by the R-454B (A2L) refrigerant transition (EPA-mandated for all units built after Jan 1, 2025: new low-GWP refrigerant, added leak sensors/detectors, facility storage retrofits, and supply-chain constraints) and coil shortages from the Longview ERP go-live (April 1, 2025) that starved the Tulsa plant — produced “sub-optimal overhead absorption.” (FACT — FY2025 10-K MD&A p.31-33.) Management partly pre-empted this with a one-time 3.0% AAON-branded price increase effective Jan 1, 2025, but price could not outrun the volume air-pocket. (FACT — 10-K p.717-line equivalent, p.33.)
• A geography/accounting quirk that depresses the rate, not the economics. The new Memphis plant is organizationally inside AAON Oklahoma but builds units at cost that are booked into the BASX segment. So Memphis’s $16.1M of FY2025 overhead lands in AAON OK with no matching revenue, mechanically depressing that segment’s margin and the consolidated rate while it ramps. This is a real cash cost today but a transitory drag, not a permanent structural loss. (FACT — FY2025 10-K MD&A p.33; Q1-2026 10-Q quantifies the Memphis drag at ~$9.8M in the quarter alone — p.34.)
• Mix shift toward structurally lower-margin data-center work. BASX-branded products grew 143.5% (+$322.8M) and now run through both BASX (26.6% GM) and Coil Products (21.4% GM) — both well below the legacy 34-37% rooftop margin. As data-center cooling becomes a larger share of the mix, it dilutes the blended gross margin even when each segment is healthy. This is the permanent component of the reset. (FACT — FY2025 10-K MD&A p.31; segment note p.79-80.)
• Price/raw-material cost: roughly neutral to slightly favorable. The 12-month average input basket was mixed (copper +11.1%, stainless -13.0%, galvanized steel -3.4%, aluminum -0.4%); commodity cost was not the swing factor. (FACT — FY2025 10-K Raw Material Costs table, p.33.)

The critical read on Q1-2026: consolidated GM was 25.1%, below the full-year FY2025 26.7%, even as revenue grew 54.3% and AAON Oklahoma GM recovered to 26.3% on price and Tulsa volume. The reason is mix: BASX surged to 27% of sales at a 23.9% GM (held flat by outsourcing costs), and the Memphis-overhead-in-AAON-OK drag persisted. (FACT — Q1-2026 10-Q MD&A p.33-34.) This is the single most important disconfirming fact for any “margins are about to snap back to 34%” thesis: the consolidated gross margin has not yet inflected upward, and the mix that is depressing it is the part of the business that is growing fastest.

3. Operating-margin walk: opex-led compression

Below the gross line, AAON ran a second, separate margin leak: operating expenses grew far faster than revenue. Over 2023→2025, SG&A rose +39.6% ($171.5M → $239.5M) and R&D +33% ($43.7M → $58.2M) against revenue growth of only +23%. The FY2025 SG&A bridge (10-K p.33) is a mix of investment and deleverage:

The honest characterization is roughly half investment, half deleverage. The salaries build, R&D, and ERP/controls spend are deliberate capacity-building ahead of the data-center ramp; the Memphis incentive fee and a chunk of the consulting are one-time. But a material part of the 2025 margin damage is simply operating deleverage — fixed cost spread over volume that stalled in the first half. The Q1-2026 print is the tell: SG&A fell to 13.7% of sales from 15.9% a year earlier, and operating margin ticked up to 11.5% from 10.9%, as the same cost base spread over 54% more revenue. (FACT — Q1-2026 10-Q MD&A p.34; OM commentary p.32.) That is exactly what you would expect if the opex base was built ahead of a volume recovery — supporting (though not proving) the cyclical-trough interpretation on the operating line.

4. Quality of earnings: the OCF collapse

This is where earnings quality is genuinely tested. Operating cash flow fell from $192.5M (2024) to $0.53M (2025) against net income of $107.6M — a cash-conversion ratio of essentially 0% versus 114% in 2024 and 89% in 2023. (FACT — FY2025 10-K Statement of Cash Flows, p.37.) The decomposition is unambiguous:

A ~$236M working-capital swallow consumed virtually all of $237M of cash earnings. The question is whether that build is backlog-driven and recoverable or structural and impairing. The evidence leans recoverable, with two caveats:

Evidence it is recoverable / backlog-funded:

• Backlog more than doubled. Total backlog rose 110.9% to $1,828.5M at year-end 2025 (BASX backlog +141.3% to $1,302.1M), and again to $2,129.5M at Q1-2026 (+107.4% YoY; BASX +160%). (FACT — FY2025 10-K p.78; Q1-2026 10-Q p.32.) The inventory and contract-asset builds are management explicitly pre-funding this backlog — “significant purchases of inventory related to data center orders… allocated to customer jobs and show as increases to our contract assets,” and BASX jobs “require the Company to fund the upfront working capital.” (FACT — 10-K p.37.) Contract assets are over-time revenue recognized but not yet billed on customized BASX units — a function of the over-time accounting model on a doubling order book, not slow collection.
• Q1-2026 already shows partial reversal. OCF swung to +$34.0M (from -$9.2M in Q1-2025), and crucially accounts receivable reversed +$24.3M (cash collected) versus a -$17.6M build a year earlier. (FACT — Q1-2026 10-Q Statement of Cash Flows, p.4.) The single scariest FY2025 line — the AR balloon — began unwinding within one quarter, which is strong (if early) evidence that it was year-end timing, not deteriorating collectibility.

Caveats / what still needs watching:

• The AR build was disproportionate to revenue. AR more than doubled ($147.4M → $314.4M) while revenue grew only 20%, pushing DSO from 45 to 80 days. Contract-asset days rose from 41 to 63 and inventory days from 85 to 90. The DSO doubling is larger than a pure volume story justifies and likely embeds a shift toward large data-center customers with bigger, lumpier, slower-paying terms. (FACT — computed from 10-K balance sheet p.46.) The Q1-2026 collection is reassuring but is one quarter; the durability of working-capital intensity at the new BASX-heavy mix is an OPEN QUESTION.
• No accounting red flags in the non-cash adjustments. AR provisions and inventory obsolescence reserves are small and stable; there is no aggressive capitalization masking opex. The OCF collapse is a working-capital phenomenon, not an accruals-quality fraud signal. (INTERPRETATION — 10-K/10-Q cash-flow detail.)

Net read: the cash collapse is largely a recoverable, growth-funding build on a doubling backlog of project-based data-center work, not an earnings-quality alarm — but it permanently re-rates AAON’s working-capital intensity higher. This is no longer a self-funding rooftop manufacturer; it is a project-finance-intensive cooling business that must carry the cash gap on its balance sheet.

5. Free cash flow: deeply negative

With OCF near zero and capex running ~$190M, FCF is sharply negative:

(FACT — FY2025 10-K Statement of Cash Flows, p.37.) AAON was FCF-negative even in 2024 once capex is counted; in 2025 it burned ~$204M. Capex intensity is ~13% of sales, and the 2026 capex guide is ~$190M. (FACT — 10-K p.38.) This is a structurally capital-hungry business in an investment phase (Memphis, Longview, Redmond, Parkville capacity), and even on a recovering OCF base in 2026, FCF is likely to remain modest-to-negative until the working-capital build normalizes. The valuation implication is stark: a name trading at ~50x trailing EBITDA / ~36x on Q1-2026 annualized EBITDA is generating no free cash flow today.

6. Returns on capital: above-to-around cost of capital

(Effective tax rates 18.4%/16.4%; equity $824.6M → $895.0M per 10-K p.46; invested capital ≈ total debt + equity − cash.) The 2025 collapse in ROIC to ~9.3% is the financial expression of the whole story: margins halved on the operating line while invested capital ballooned +$314M on capex and working capital. Crucially, 2025 ROIC sits at or just below a reasonable WACC (~9-10%) — meaning at the trough, AAON is barely creating economic value. In 2024 it comfortably exceeded its cost of capital. The Greenwald test is decisive here: a business with a durable moat should sustain ROIC well above WACC through a cyclical air-pocket; AAON’s slide to break-even economics in a single bad year suggests the “moat” is real but cyclically fragile, and the returns are now being earned on a much larger, lower-returning asset base. Whether ROIC re-rates back toward the high teens depends entirely on margin recovery and working-capital normalization — neither yet proven.

7. Balance sheet and liquidity

The balance sheet is adequate but visibly stretched by the growth investment, and the firm is now running on its revolver:

• Cash: $13K. Unrestricted cash is effectively zero; AAON funds operations day-to-day off the revolving credit facility. (FACT — 10-K balance sheet p.46.)
• Debt ~$414.6M (revolver $398.3M + NMTC obligations $16.3M), up from ~$171M a year earlier; the revolver balance rose further to $425.2M at Q1-2026. Leverage is 1.77x (covenant ceiling 3.0x), and EBIT/interest coverage is ~8.3x. (FACT — 10-K p.34-35; Q1-2026 10-Q p.35.)
• Revolver capacity expanded three times in 2025 ($200M → $230M → $500M → $600M via accordion), expiring May 2030 — management is deliberately building borrowing headroom to fund the data-center working-capital cycle. Availability was $201M at year-end, $173.5M at Q1-2026 — adequate but shrinking, and worth monitoring. (FACT — 10-K p.34; Q1-2026 p.35.)
• Interest cost surge: interest expense jumped from $2.9M (2024) to $17.7M (2025) on the revolver draw at ~5.7% — a real, growing drag on the pre-tax line as the balance climbs. (FACT — 10-K p.46.)
• Off-balance-sheet: New Markets Tax Credit structures (2019/2023/2024) carry modest low-rate debt with potential future forgiveness; not a hidden liability. Accounting (Grant Thornton, auditor since 2004) flags over-time BASX revenue recognition as the sole critical audit matter — appropriate given $547.8M of BASX revenue recognized on estimated-cost-to-complete. (FACT — 10-K p.34-35, 44.)

Dilution and shareholder returns are benign but, tellingly, debt-funded in 2025. Shares rose only 0.3% (81.4M → 81.7M); SBC was $18.0M (1.2% of revenue). The company repurchased $39.7M and paid $32.6M in dividends (~30% payout) — but with FCF at -$204M, both the dividend and buyback were funded off the revolver, not free cash flow. (FACT — 10-K p.36-37.) That is sustainable at 1.77x leverage but is a flag: returning capital you are simultaneously borrowing is a capital-allocation choice that only works if the trough is genuinely temporary.

8. Verdict — do economics improve with scale; trough or structural reset?

The answer is split, and saying so plainly is the honest call: 2025 is a cyclical/transitional trough in margin and a partial structural reset in margin level. The transient drags — R-454B refrigerant disruption, the Longview ERP go-live, Memphis under-absorption, and headcount built ahead of volume — are real and fading, and Q1-2026 confirms the operating line is already healing (OM 11.5%, SG&A leverage returning to 13.7% of sales, AAON Oklahoma GM back to 26.3%, AR collecting). On that evidence, the operating margin should recover toward the mid-teens as volume absorbs the cost base and the one-timers roll off.

But the proposition that “economics improve with scale” — the core test of a quality compounder — does not hold cleanly here. AAON is scaling into a structurally lower-margin, more capital-intensive, more working-capital-hungry mix. The fastest-growing segment (BASX/data-center) runs 24-27% gross margins versus the legacy 34-37% rooftop business; the consolidated gross margin was still falling in Q1-2026 (25.1%) precisely because the favorable-mix segment is shrinking as a share and the unfavorable-mix segment is exploding. Returns on capital halved to ~WACC in a single year, FCF is deeply negative (-$204M), capex runs ~13% of sales indefinitely, and the dividend and buyback are now financed with borrowed money. The 34% gross-margin, self-funding, high-ROIC AAON of 2023 is unlikely to return; the steady-state AAON is a faster-growing but lower-margin, lower-return, more leverage-dependent business. Bottom line: margins and cash flow are at a cyclical trough that should improve from here, but the level they recover to is structurally below the 2023 peak — investors paying ~50x trough EBITDA are underwriting both a sharp cyclical snap-back and a margin recovery that the most recent quarter does not yet show.

7. Capital Allocation

Framing. Capital allocation is where AAON’s premium narrative collides most violently with the arithmetic. Over FY2023–FY2025 the company has executed a deliberate, large-scale capacity build to convert itself from a niche semi-custom rooftop maker into a data-center-cooling supplier — and in doing so has consumed essentially all of its operating cash flow, levered a previously debt-light balance sheet, and watched returns on a rapidly-growing capital base fall by roughly half. Management is making a coherent bet; the question is whether the bet has so far created per-share value or merely grown the asset base. On the evidence to date, the latter dominates. The single most important capital-allocation fact is that in FY2025 AAON generated $0.53M of operating cash flow against $190.6M of capital expenditure — a roughly −$190M free-cash-flow year — and still raised the dividend and bought back stock, funding the entire shortfall with debt (FACT, FY2025 10-K cash-flow statement).

The BASX acquisition (December 2021) — the pivot that reset the business and the margin profile

AAON acquired BasX Solutions (Redmond, OR) on December 10, 2021. Consideration: $100M cash upfront + up to $80M of contingent earn-out payable in AAON stock (2021–2023 milestones) + a separate $22M purchase of the BasX real estate — up to ~$202M all-in (FACT: AAON 8-K/press release, Dec 13, 2021; cleanroomtechnology.com, Dec 2021). The cash-flow statements record $103.4M cash paid for the business combination in 2021 and the $22M building payment in 2022 (FACT, FY2022 10-K, Investing Activities). Goodwill stepped up from $3.2M (2020) to $85.7M (2021) — i.e. ~$82.5M of goodwill, implying AAON paid a meaningful premium over BASX’s net tangible assets, consistent with a high-single-digit-to-low-double-digit revenue/EBITDA multiple for a then-~$100M-revenue custom-cooling business (FACT: EDGAR XBRL Goodwill concept; INTERPRETATION on the multiple, as FY2021 PPA detail is not in the trailing-36-month corpus).

Has BASX created value? The honest answer is strategically yes, financially not-yet-proven, and margin-dilutive so far:

• Strategic value (real). BASX is the vehicle for AAON’s entire data-center-cooling exposure — the growth story underwriting the ~95x trailing / ~41x forward multiple. Without BASX, AAON is a low-double-digit-grower rooftop business; with it, it has a credible AI/hyperscale-capex narrative. The 2022 revenue surge (+66% to $888.8M) and 2025 re-acceleration (+20% to $1,442.1M) lean heavily on this segment (FACT: EDGAR revenue series; INTERPRETATION on attribution).
• Financial value (unproven and arguably negative to date). The pivot coincides exactly with the margin and cash collapse. Consolidated gross margin fell from 34.1% (2023) to 26.7% (2025) and operating margin from 19.5% to 10.1% as the lower-margin, more capital- and working-capital-intensive data-center mix scaled (FACT: EDGAR). Data-center custom air-handling is a project business — large, lumpy, working-capital-heavy, and structurally lower-margin than AAON’s legacy configure-to-order rooftop franchise. The acquisition did not buy a higher-return business; it bought a faster-growing, lower-return one, and the consolidated economics now reflect that.

Verdict on BASX: a defensible strategic acquisition at a full price that has delivered growth but has diluted the very return and margin characteristics that justified AAON’s premium rating. It is a growth purchase, not a quality purchase. (INTERPRETATION, weighing disconfirming evidence: if data-center backlog converts at the margins management implies and the new capacity reaches scale, BASX could yet prove accretive on a per-share basis — but that is a forward hope, not a realized result.)

The capex super-cycle — disciplined growth investment, or capital intensity that erodes FCF and ROIC?

This is the heart of the capital-allocation read. AAON has run a multi-year capacity build across Longview TX (coils/components and data-center), Memphis TN (new greenfield, partly grant-funded), and Redmond OR / BASX. Quantified against depreciation and operating cash flow:

(FACT: all figures from FY2022–FY2025 10-K cash-flow statements and EDGAR XBRL; capex excludes separately-disclosed intangible/ERP additions of $5.2M/$17.5M/$14.3M in 2023/2024/2025; 2022 also included the $22M BASX building.)

Three observations not to miss:

1. Capex has run at 2.2x–3.1x depreciation for three straight years — this is a genuine growth capex program, not maintenance. Cumulative 2023–2025 capex of ~$490M against ~$188M of D&A means ~$300M of net new asset investment. PP&E (net) more than doubled from $304.7M (2022) to $631.3M (2025), and total assets ballooned from $814M to $1,686M (FACT: EDGAR). The asset base is growing far faster than profits.

2. The return on that base is falling, not rising. Operating income fell from $227.5M (2023) to $146.2M (2025) — a 36% decline — while invested capital roughly doubled. On a rough NOPAT basis, pre-tax operating ROIC compressed from the high-20s%/low-30s% in 2023 to roughly the low-teens in 2025 (INTERPRETATION: op income ÷ approximate invested capital of debt+equity; equity rose to $895M and revolver debt to $398M by FY2025). ROE on reported net income fell from ~24% (2023, $177.6M on ~$745M equity) to ~12% (2025, $107.6M on ~$895M equity) (INTERPRETATION from EDGAR net income + equity). This is the textbook signature of Marathon’s “capital pouring in as returns fall” warning — the asset-growth anomaly in real time.

3. The cash conversion has broken. FY2025 OCF of $0.5M against $107.6M net income is a stark quality-of-earnings flag: inventory rose $73.8M (to $261.2M) as the company pre-bought for data-center orders and navigated the A2L refrigerant transition (FACT: FY2025 10-K). Even granting that the working-capital build is largely backlog-driven and recoverable, the simultaneous capex peak means AAON funded its dividend, its $30M buyback, and ~$190M of net capex almost entirely with debt: the revolver rose from $138.9M (2024) to $398.3M (2025), +$259M (FACT: EDGAR LineOfCredit), and cash sits at $13K (FACT: EDGAR/FY2025 10-K).

Where in the capital cycle? AAON sits at the peak-investment phase of a Marathon capital cycle: high prior returns (2023 ~30% ROIC) attracted a wave of internal reinvestment; the new capacity is ahead of the earnings it is meant to generate; and the consolidated return is mean-reverting downward. The bull case is that this is the self-funded version of the cycle — AAON is adding capacity to capture a genuine secular data-center demand wave, and returns will re-rate as the plants fill. The bear case is that AAON has converted a high-return niche into a mid-return capital-intensive contract manufacturer and the market has not yet repriced it. Verdict on the capex cycle: this is growth investment, made with intent and into real demand, but it is not yet disciplined on a returns basis — three years of 2–3x-depreciation capex have coincided with halving operating margins, a ~190M negative-FCF year, and a doubling of debt. Discipline is proven only by the returns the new assets ultimately earn, and that proof does not yet exist.

Buybacks, dividends, and issuance — a modest, balanced, mostly-self-disciplined return program

AAON’s direct shareholder returns are small relative to the capex program but reasonably structured:

(FACT: FY2022–FY2025 10-K financing activities; DPS from EDGAR CommonStockDividendsPerShareDeclared, split-adjusted, moved to quarterly $0.10 in 2025.)

The 2024 $100M buyback was the standout — opportunistically large and well-timed against a stock that had de-rated on the 2024 earnings stall, and it shrank the share count meaningfully in a year when the stock was cheap. In 2025, with the stock recovering and the balance sheet stretched, repurchases sensibly fell back to $30M. The dividend is small (yield ~0.30%, payout ~28%) and has been raised steadily, but at $32.6M it is not covered by free cash flow in 2024 or 2025 and is being sustained on the revolver — a yellow flag if the negative-FCF period extends.

On dilution and share count. The headline “shares rose from ~52M to ~82M” is largely an artifact of the 3-for-2 stock split (effective Aug 16, 2023), not dilution (FACT: FY2023 10-K). Split-adjusting 2021 basic shares of 52.4M (×1.5 = ~78.6M) against FY2025 basic of 81.5M implies roughly +3.7% net dilution over four years — modest, and notable given ~$18.0M of annual stock-based compensation (FY2025) (FACT: EDGAR SBC series 2022–2025: $13.7M → $16.4M → $16.7M → $18.0M). The recurring buybacks have therefore largely neutralized SBC creep rather than letting the count drift up — a genuine point in management’s favor. The BASX earn-out added a small slug of stock on top of the split. Net read: issuance discipline is good; SBC is contained and offset; the buyback is opportunistic rather than chronic. This is the cleanest part of AAON’s capital allocation.

The proxy — what is management actually paid to maximize?

Reading the 2026 DEF 14A (filed April 1, 2026), the incentive design is the central capital-allocation concern, because it pays for growth and budget-hitting, not for returns or per-share value:

• Annual cash bonus (the short-term incentive) is weighted “Operating Profit (67%)” + “Net Sales (33%)” against the “Opportunity Budget” (FACT, verbatim, 2026 DEF 14A CD&A, p.35). “Operating Profit” is defined as “the Company’s budgeted net income before profit sharing and income taxes, but after bonus accrual.” There is no ROIC, ROE, EPS, free-cash-flow, or margin metric in the annual bonus. A management team paid 33% on top-line and 67% on a budgeted (not absolute or per-share) profit number is structurally incentivized to grow revenue and hit an internally-set budget — precisely the behavior that produces a capex super-cycle into a lower-return mix. This is a material mis-alignment with per-share value creation. (To management’s partial credit, the formula did respond to the 2025 operating miss: the weighted bonus factor came in at 0.31 of target, so NEOs were paid roughly 31% of target bonus — the plan is not cosmetic. FACT, 2026 DEF 14A, p.35–36.)

• Long-term equity is better aligned: the 2025 mix is PSUs 50% / stock options 25% / restricted stock 25%, and the PSUs vest on three-year relative total shareholder return vs. the S&P 400/600 Building Products Industry Group, with an absolute-TSR cap (payout capped at 100% if absolute TSR is negative) (FACT, verbatim, 2026 DEF 14A, p.36–37). Relative TSR is a legitimate per-share-value-linked metric and the negative-absolute cap is a sensible guardrail. The most recent 2023–2025 PSU cycle paid out at 83.2% of target (TSR 56.6%, 46.6th percentile) — i.e. the relative-TSR plan delivered a below-target payout, which is appropriate given the operating deterioration. So the long-term incentive is reasonably aligned; the annual incentive is not.

• Ownership requirements and guardrails are robust: CEO must hold 6x base salary in stock, EVP/CFO/COO 3x, directors 6x cash retainer; 75% net-share retention until met; hedging and pledging are prohibited; both a discretionary and a Dodd-Frank mandatory clawback are in place (FACT, 2026 DEF 14A, p.38–39). Pay is benchmarked to a 17-company building-products/manufacturing peer set with Meridian as consultant; 81.5% of the CEO’s target pay is “at risk.”

• Founder/insider ownership. Founder-chairman Norman Asbjornson owns 13.73M shares / 16.8% (incl. trusts and foundation) — a large, aligned, long-tenured holder whose stake is the strongest single alignment fact in the file (FACT, 2026 DEF 14A, p.28). But note the operating leadership owns relatively little: new CEO Matthew Tobolski holds 335,291 shares + 30,174 options (<1%) and former CEO Gary Fields holds just 49,622 shares + 133,160 options (<1%) (FACT, 2026 DEF 14A, p.29). There was a CEO transition in May 2025 — Tobolski (formerly President/COO) replaced Fields, who moved to “Special Advisor” — a material governance change executed without disclosed turmoil but worth flagging (FACT, 2026 DEF 14A, p.34).

Insider selling — a conviction signal that cuts against the premium multiple

The May–June 2026 Form 4 activity is uniform, discretionary, and bearish-leaning at a ~95x trailing P/E (all FACT, EDGAR Form 4 XML, parsed 2026-06):

• Former CEO Gary Fields: open-market sale of 19,000 sh @ $140.20 (~$2.66M) on 2026-06-01, leaving only 15,252 shares directly held; plus a 2026-05-27 exercise-and-sell of 21,173 sh @ $140.34 (~$2.97M). No 10b5-1 plan is referenced in the filings.
• CAO Rebecca Thompson: open-market sale of 4,230 sh @ $143.42 on 2026-06-05, reducing her directly-held position to zero.
• EVP Gordon Wichman: exercise-and-sell of 3,000 sh @ $140.39 on 2026-05-27.

None of these are routine 10b5-1 dispositions on the face of the XML — they read as discretionary sales clustered after the stock’s recovery toward its 52-week high. A CAO selling to zero direct shares and a recent-CEO cutting to a stub are not, on their own, a thesis-breaker (executives diversify; Fields is departing the operating role), but the pattern — uniform selling across CEO/CAO/EVP, discretionary, into strength, at ~95x trailing earnings — is a negative conviction signal that sits awkwardly against consensus’s underwriting of a sharp 2026–2027 earnings recovery. The aligned long-term holder is the founder; the current operators are net sellers with small stakes.

Verdict — has management allocated capital intelligently on a per-share basis?

Mixed, tilting negative on the evidence to date — call it “ambitious growth investment not yet validated by returns.” The constructive elements are real and should not be dismissed: the BASX pivot positioned AAON in front of a genuine secular demand wave; the buyback is opportunistic and has neutralized SBC dilution; the share count (split-adjusted) is barely up; the dividend is conservative; ownership/clawback governance is strong; the founder owns 16.8%; and the relative-TSR PSU plus the 0.31x 2025 bonus show the pay machine does respond to underperformance. But the dominant facts are unfavorable on a per-share-value basis: three years of 2–3x-depreciation capex have coincided with operating margins halving, ROIC/ROE roughly halving, a ~$190M negative-FCF year, cash at $13K, and revolver debt tripling to $398M — the classic peak of a Marathon capital cycle, with capital growing far faster than the returns it earns. The annual incentive pays for revenue and budgeted profit, not for returns or per-share value, which structurally encourages exactly this empire-building reinvestment. And the people running the business are selling. Management has allocated capital aggressively and coherently, but the proposition that it has done so intelligently on a per-share basis depends entirely on a returns recovery that has not yet appeared in the financials. Until the new asset base demonstrably earns its cost of capital, the prudent read is that AAON has grown its balance sheet faster than it has grown per-share value.

Verdict: Capital allocation is ambitious and self-aware at the margins (good buyback discipline, strong governance, founder alignment) but, on the central question, has not yet proven intelligent on a per-share basis — a returns-dilutive, debt-funded capex super-cycle into a lower-margin mix, rewarded by an incentive plan that pays for growth rather than returns, with current operators selling into the multiple.

8. Changes and Headwinds — Last Two Years

SEC Filings Sweep & Insider read. Built from the trailing 36-month corpus mirrored in output/AAON/sources/ — four 10-Ks, the 10-Q set through Q1 FY2026, ~49 8-Ks, the 2026 DEF 14A, and all 129 Form 4s filed since June 2024 (parsed to 298 transaction rows). Numbers reconcile to the filings; management framing is treated as hypothesis and checked against the financials.

The last two years rewrote AAON’s story twice over. The legacy semi-custom rooftop business — the franchise — stalled and then shrank, while a regulatory refrigerant transition and a botched ERP cutover gutted margins. Simultaneously, a secular data-center cooling boom (the BASX/Longview franchise acquired in 2021) detonated the order book to record levels. The result is a business whose backlog more than doubled while its operating margin halved — the central tension of the entire thesis, and the reason the run-rate is so hard to read.

Material-event timeline (8-K / 10-K / 10-Q corpus)

Two corrections to the working brief, both load-bearing. First, Gary Fields is no longer CEO — he stepped down to “special advisor to the Board” on May 13, 2025, and Matt Tobolski (age ~42, BASX PhD co-founder) is the sitting CEO (8-K 2025-02-20; 8-K 2025-05-16). Second, Norman Asbjornson is not chairman — he retired as Executive Chairman in May 2022 and is now an ordinary non-employee director (age 90, 38-year tenure); the board is independent-chaired (A.H. McElroy II is Independent Chair) (DEF 14A 2026-04-01). These matter because the AZI feed’s “CEO Gary Fields” selling label is mis-titled (Fields sells as a departed director, not the running CEO), and the governance is less founder-controlled than the brief implies.

The 2024 stall and the 2025 margin collapse — what actually happened

The 2024 stall was the A2L refrigerant transition (EPA mandated lower-GWP R-454B for equipment manufactured on/after Jan 1, 2025) colliding with a soft nonresidential construction market: AAON-branded products fell 2.6%, and the AAON Oklahoma segment dropped 4.4%, with BASX data-center growth (+35.1%) papering over the consolidated number to a barely-positive +2.7% (FY24 10-K MD&A).

2025 is where the damage shows. Revenue grew a healthy +20.1% to $1,442.1M, but gross margin compressed from 33.1% to 26.7% and operating margin halved from 17.4% to 10.1%, dragging net income down ~36% to $107.6M. Management’s own attribution (FY25 10-K MD&A, lines 916-1071) is unusually specific, and it is mostly self-inflicted execution, not pricing or competition:

• ERP cutover at Longview (Apr 1, 2025) disrupted AAON Coil Products and, by starving Tulsa of coils, also throttled AAON Oklahoma volume in H1;
• A2L refrigerant supply-chain constraints early in the year;
• Sub-optimal overhead absorption on the H1 AAON-brand volume shortfall;
• Memphis plant startup — because Memphis is housed in the AAON Oklahoma segment but builds intercompany product for BASX at cost, it loaded $16.1M of overhead into AAON Oklahoma with the revenue/margin booked in BASX.

AAON-branded revenue fell −8.3% in 2025 even as BASX-branded revenue rose +143.5%. This is no longer the same company that compounded on premium rooftop units: the growth engine, the margin profile, and the customer base have all shifted toward custom data-center cooling.

The bull-side offset: the backlog re-rate

The reason the stock is up despite halved margins is the order book. Consolidated backlog reached $1,828.5M at Dec 31, 2025 — up 110.9% YoY — with BASX-branded backlog of $1,302.1M (+141.3%), “most of these orders associated with BASX-branded data center liquid cooling solutions” (FY25 10-K). Q1 FY2026 extended it to $2,129.5M (+107.4% YoY), and Q1 revenue grew +54.3% with AAON Oklahoma recovering (+$82.1M as Tulsa came back online against the easy refrigerant comp). The recovery the consensus is underwriting (EPS TTM $1.42 → CY est $2.26 → NY est $3.29) is visibly beginning — but Q1 gross margin was still only 25.1% (vs 26.8% a year earlier), with $9.8M of Memphis overhead still a drag. The order book is real; the margin recovery is, so far, a hypothesis with one soft data point.

One-time items distorting the run-rate

• 2023 SG&A carries the $7.5M Havtech settlement — normalize it out before any 2023→2024→2025 SG&A trend read.
• 2024-2025 D&A step-up from ERP capitalization inflates SG&A optically (FY24 D&A +49.3%; FY25 +$7.2M) — partly non-cash investment, not core cost creep.
• 2025 SG&A includes a $6.1M real-estate broker incentive fee tied to the Memphis acquisition and $17.4M of technology/consulting — transitional, not permanent run-rate.
• Memphis overhead in the wrong segment ($16.1M FY25, $9.8M Q1’26) artificially depresses AAON Oklahoma’s reported margin and flatters BASX’s — segment margins are not clean comparables until Memphis matures.

Insider transactions — the mandatory read (129 Form 4s, trailing 24 months)

The signal is one-directional distribution, but it is benign in character, not a red-flag founder dump.

Buying is negligible. Just five open-market purchases totaling ~$0.6M in two years — Wakefield (EVP, ~$490K at $77-82 in Dec 2025), Shaub (EVP, $35K, Sep 2025), and director Stewart ($81K, Mar 2025). All were near the 2025 lows (low-$80s); no insider has bought a single share above ~$82, against a current price of ~$132. That is a meaningful absence of conviction at the re-rated level.

Selling is large but mostly mechanical. Of $50.9M sold, $31.0M (61%) was Gary Fields — the departed CEO — exercising deep-in-the-money options (strikes $31-49) and immediately selling into a $134-140 market. That is rational monetization by an exiting executive, not an active-management signal. The founder, Asbjornson (16.8% / 13.7M shares), made only one open-market sale in 24 months (11,022 shares, ~$0.96M — under 0.1% of his stake); his activity is dominated by 854K shares of gifts (to trusts/foundation — estate and charitable planning) and low-strike option exercises, which is what you’d expect from a 90-year-old founder, not a loss of faith. The sitting CEO, Tobolski, made one $1.08M sale (8,000 shares, May 2026) while gifting ~125K shares into trusts — his economic stake is largely intact.

Two caveats sharpen the negative tilt. First, the AZI-flagged May-June 2026 cluster is confirmed and uniform — Fields ($31M cumulative; ~$5.6M across May-June 2026 alone), CAO Thompson ($6.6M cumulative; sales June 5 at $143), EVP Wichman, and CAO Kidwell all sold — and it coincides with the stock’s run from a $62 52-week low to $140+. 2026 H1 was the single heaviest selling half-year (~$20.0M across 18 sales). Selling into a doubling is not damning, but no one is leaning in. Second, 10b5-1 protection is almost absent: only 6 of 226 disposition events (89,838 shares) were under a Rule 10b5-1 plan — ~96% of insider sales/exercises were discretionary. That undercuts the usual “it’s just automatic diversification” defense; these were active decisions to sell.

A final juxtaposition worth attention: in FY2025 the company repurchased $39.7M of stock at an ~$81 average ($30.0M open-market at $80.81; $9.7M for option/tax) — funded off a revolver while operating cash flow was ~$0 — while insiders were exercising and selling at $134-140. Buying low with borrowed money on the corporate account while insiders distribute high on personal accounts is an uncomfortable, if not unusual, pattern.

Risk inputs for the Lead’s matrix (ranked)

Verdict — these two years WEAKEN the thesis on quality while STRENGTHENING it on demand, and the insider signal tilts mildly negative. The franchise that earned AAON its premium — high-30s-gross-margin semi-custom rooftop units — shrank and de-rated on margin, displaced as the growth driver by lower-margin, more cancellable, more concentrated custom data-center cooling. The backlog re-rate (+111% to $1.83B) is genuine and is the bull case in one number, but it has been delivered alongside a halving of operating margin, a collapse of operating cash flow to ~zero, a cash balance of $13K, and a revolver drawn to two-thirds of a freshly-tripled $600M facility — the company is funding a debt-financed growth sprint whose returns are not yet visible in the P&L. The insider tape reinforces the caution: zero buying above $82, ~$51M of overwhelmingly discretionary (non-10b5-1) selling concentrated into the stock’s doubling, and the corporate buyback executed ~40% below where insiders chose to sell. None of it is a fraud flag or a founder bailout — Asbjornson and Tobolski retain their economics — but on balance the changes raise the execution bar and the insider behavior offers no support for the re-rating, leaving the burden of proof squarely on the unproven 2026 margin recovery.

9. Risk Analysis (Risk Matrix)

The matrix below grades each material risk by likelihood and impact (Low/Med/High), each tied to an evidence basis from the filings, proxy, Form 4 corpus, and industry analysis. It is the consolidated risk view feeding the Variant Perception (§11) and What Must Be True (§14) sections.

Likelihood and Impact graded Low / Medium / High. “Impact” measures the consequence to intrinsic value and/or the multiple if the risk materializes. Evidence is drawn from the FY2025 10-K, Q1-2026 10-Q, the proxy, the Form 4 corpus, and the industry analysis.

10. Valuation Discussion (Embedded Expectations)

All multiples as of 2026-06-08 (price ~$132.05; market cap ~$11.0B; EV ~$11.3B; 81.9M shares; net debt ~$443M). Peer multiples via Yahoo Finance (yfinance, accessed 2026-06-08) and cross-checked to the financial analysis; treat as third-party and reconcile to filings. No price target and no BUY/SELL — this section reverse-engineers the expectations embedded in the price and stress-tests them against scenarios.

AAON trades like a hyper-scaler-adjacent secular-growth compounder, but it earns like a cyclical HVAC manufacturer at a margin trough. That gap — between the multiple and the trailing economics — is the entire valuation question. The stock changes hands at ~93x trailing GAAP earnings, ~40x forward, ~50x trailing EV/EBITDA (45x on the lower-quality yfinance EBITDA), ~7.8x EV/revenue and ~7.6x P/sales on FY2025 actuals. On any of these, AAON is priced at or near the top of its peer group and at the 96th–99th percentile of its own ten-year valuation history (below). The market is not paying for what AAON is; it is paying for what AAON might become if the data-center mix scales, margins snap back toward the 2023 peak, and the AI-cooling tailwind persists for years. The body’s job is to make that embedded bet explicit and to price the alternatives.

10.1 Where the stock trades — sector-appropriate multiples and the peer comp

The first observation is that AAON’s trailing multiple is distorted by the FY2025 margin trough (10.1% operating margin, $107.6M net income, $1.42 TTM EPS). At 93x trailing P/E the stock looks absurd, but that denominator is a trough; the forward P/E of ~40x is the more honest starting point because it already assumes a partial recovery (consensus CY EPS $2.26, NY EPS $3.29 — a near-doubling off TTM in two years). The peer table below uses consistent yfinance multiples so the relative premium is apples-to-apples, then contrasts AAON’s economics against the set.

(FACT — Yahoo Finance, 2026-06-08. AAON’s EV/EBITDA shows ~45x on yfinance’s figure and ~50x on a FY2025 EBITDA of $225.4M; both are correct under different EBITDA conventions.)

Three conclusions fall out of the table. First, AAON carries the highest forward P/E of the entire group except a tie with Comfort Systems and Vertiv (~40x vs. FIX 35x, VRT 34x), and the highest trailing P/E except Modine — whose 121x is itself a trough artifact, not a richer valuation. Against the diversified HVAC majors (CARR, TT, JCI, LII at 19–27x forward), AAON trades at a 50–110% forward-P/E premium; against distribution-quality Watsco (27x) a ~50% premium; against the closest quality analog Trane (27x forward, 24.7x EV/EBITDA) a ~50% premium on P/E and roughly double on EV/EBITDA. Second, on EV/EBITDA (~45–50x) AAON sits at the very top, level with or above Vertiv (48.8x) — the actual data-center-thermal leader with a $15B backlog and 30% revenue growth — despite being one-fifteenth Vertiv’s size and a sub-scale follower in that arena (Industry §3.5). Third, and most damning, AAON pairs a top-of-group multiple with bottom-of-group current profitability. Its FY2025 operating margin of 10.1% is half Trane’s mid-teens-to-high-teens, below Lennox’s high-teens, and roughly in line with low-margin distribution/contracting models (Watsco, Comfort Systems) — but AAON is priced like neither a high-margin compounder nor a capital-light distributor.

The peer that matters most for the thesis is the pairing of Trane (the quality benchmark) and Vertiv (the growth benchmark): AAON is priced as if it deserves Trane’s quality premium and Vertiv’s growth premium simultaneously, while currently demonstrating neither — Trane-beating growth (Q1-26 +54%) but Trane-half margins, and Vertiv-like end-market exposure but as a price-taking sub-scale supplier (§3.2). The multiple is underwriting a future AAON that has resolved this tension. The valuation question is whether that resolution is probable.

10.2 Own-history percentile — AAON is near its most expensive ever

AAON’s valuation against its own ten-year history removes the cross-sectional argument that “HVAC just re-rated.” An own-history valuation index (as of 2026-06-05) is decisive:

(FACT — own-history valuation index, n_components=3, dated 2026-06-05; percentile = 0 cheapest, 100 most expensive vs. AAON’s own trailing history.) The only legitimate use of these percentiles is comparison against the stock’s own past, never cross-sectionally. The signal is unambiguous: even setting peers aside, AAON has rarely if ever been this expensive on its own multiples. The P/E percentile (99.3) is partly a trough-earnings artifact, but P/B at the 95.6th and P/S at the 94.4th percentile are denominator-clean — they do not depend on the depressed earnings line — and they independently confirm the stock is priced near the top of its decade. Buying here is buying AAON at a valuation it has essentially never sustained, on trough margins, on a thesis that the data-center pivot permanently re-rates the franchise. (INTERPRETATION.)

10.3 Embedded-expectations — reverse-engineering the ~$11.3B EV

The disciplined way to value a stock priced on hope is to invert it: hold the multiple constant at a sane exit level and solve for the earnings the current price requires. AAON’s ~$11.0B market cap on 81.9M shares is ~$132/share. To answer “what must be true,” I solve for the net income/EPS required to (a) merely justify today’s price at a defensible mature multiple, and (b) deliver a market return from here.

(a) Justify-today math (zero forward return — i.e., today’s price is “fair”):

• At a 25x exit P/E (already generous for a mid-teens-margin HVAC manufacturer — Trane trades 27x forward as the best-in-class), today’s $11.0B cap requires ~$440M of net income, or ~$5.37 EPS.
• At a 30x exit P/E, it requires ~$367M net income, ~$4.48 EPS.
• On EV/EBITDA: $11.3B EV at a mature 20x EBITDA requires ~$565M EBITDA; at 25x, ~$452M. At a normalized ~20% EBITDA margin, $565M EBITDA implies ~$2.8B of revenue.

Anchor those against reality: FY2025 net income was $107.6M ($1.42 EPS) and EBITDA $225.4M. So merely to make today’s price fair (zero return) at a 25–30x exit, AAON must roughly quadruple net income to $367–440M and more than double EBITDA to $450–565M. Consensus already embeds part of this — CY EPS $2.26, NY EPS $3.29 — but even the next-year $3.29 consensus, capitalized at 30x, supports only ~$8.1B of market cap (~$99/share), i.e., the consensus two-years-out number, fully credited at a rich multiple, does not reach today’s price. The price is discounting numbers beyond the visible consensus window.

(b) Return math (to earn ~10%/yr from here): to compound at 10% annually, the market cap must grow to ~$13.3B in three years or ~$17.7B in five. Holding a 25x exit P/E:

• 3-year, 25x: requires ~$7.03 EPS (~$576M net income) — a ~5.4x increase off TTM in three years.
• 5-year, 25x: requires ~$8.50 EPS (~$696M net income) — a ~6x increase.
• At a richer 30x exit, the bar eases to ~$5.86 EPS (3yr) / ~$7.09 EPS (5yr) — still a 4–5x net-income increase.

(All figures: Yahoo Finance anchors + financial analysis; ASSUMPTION on tax ~18%, ~$25M interest, flat share count.)

What does $440M+ of net income require operationally? At an ~18% effective tax and ~$25M interest, ~$440M net income needs ~$560M of operating income. To get there:

• At a 15% operating margin (the base-case partial recovery), revenue must reach ~$3.7B — a ~2.6x increase off FY2025’s $1.44B, or roughly a 20%+ revenue CAGR for five years.
• At a 19% operating margin (the 2023 peak, fully restored), revenue must reach ~$2.95B — a ~12–13% CAGR — but this requires the margin to climb 900bps while the lower-margin BASX mix grows from 38% toward a majority of revenue (§7.5 shows consolidated GM was still falling in Q1-26 to 25.1%).

This is the crux: the embedded expectation requires AAON to do two contradictory things at once — sustain ~20%+ data-center-led revenue growth (which structurally lowers gross margin, per §3 and §7.5) and simultaneously expand the operating margin back toward its high-teens legacy peak. The math only closes if you assume both the data-center volume of the bull case and the legacy-rooftop margin of 2023 — a combination the segment economics actively resist. The market is underwriting a margin recovery and a growth surge that are, to a meaningful degree, mutually exclusive. (INTERPRETATION — the central valuation tension.)

10.4 Scenario analysis (bear / base / bull) — explicit assumptions and implied value

I model three five-year paths to 2030 with explicit revenue, margin and exit-multiple assumptions, then translate to implied market cap/share. These are expectation scenarios, not a target. Tax ~18%, interest ~$25M, ~81.9M shares (light dilution ignored), ~$443M net debt.

(Scenario construction: financial, business, and industry analysis; ASSUMPTION on every margin/multiple input. Current price ~$132; current mkt cap ~$11.0B.)

• Bear (margins stay ~10–12% / data-center commoditizes / multiple de-rates). The capital-cycle risk (§3.5) plays out: Vertiv, Modine, JCI/Silent-Aire, STULZ and Munters flood capacity, BASX’s price-taking position compresses data-center margins toward low-20s gross / low-double-digit operating, the legacy core keeps eroding (~flat-to-down), and the market re-rates AAON to a normal cyclical HVAC multiple (18x P/E, 14x EV/EBITDA — still above Lennox). Result: ~$33/share, ~75% below today. This is not a tail — it is the Marathon base rate for a capacity-flooded segment, and Q1-26’s still-falling 25.1% gross margin is consistent with it.
• Base (partial recovery to ~14–16% op margin). Revenue compounds low-teens as data-center backlog converts but legacy stays soft; operating margin recovers to ~15% as Memphis/Longview absorb and one-timers roll off, but mix caps it well below the 19% peak; the multiple normalizes to a premium-but-sane 25x P/E / 20x EV/EBITDA. Result: ~$7.3–10.2B (~$89/share), ~30% below today. Notably, even a successful partial recovery — the most defensible outcome on current evidence — does not support today’s price.
• Bull (snap-back toward ~18–19% peak margins + sustained DC growth). Data-center demand outruns capacity for five more years, AAON holds share against Vertiv, the margin recovers to the 2023 peak despite the mix headwind, and the market keeps paying a 30x growth multiple. Result: ~$14.4B (~$176/share), ~33% upside. This is the outcome the price is leaning on — and even fully realized, it offers ~33% upside over five years (~6%/yr) for accepting bear-case ~75% downside. The asymmetry is unfavorable: the bull case is roughly priced in, while the base and bear cases are not.

The probability-weighted read is stark. Splitting the scenarios even-handedly (say 30% bear / 45% base / 25% bull) yields an expected value materially below today’s price — the upside in the bull case (~$176) is swamped by the downside in bear (~$33) and the still-negative base (~$89). The market is implicitly assigning a high probability to the bull path and a low one to the capital-cycle bear path that the industry structure (§3.5) and the most recent quarter’s margins argue is far from a tail. (INTERPRETATION.)

10.5 What the market is underwriting — correctly vs. incorrectly

Underwritten correctly: (1) The data-center cooling TAM is real, large, and growing ~20–23% (§3.1) — AAON’s revenue growth is not in doubt; Q1-26 +54% and a $2.1B backlog (§7.4) make the top-line case. (2) The legacy refrigerant/ERP/under-absorption drags are partly transient — Q1-26 operating margin ticked to 11.5% with SG&A leverage returning (§7.5), so some margin recovery is genuine and the trailing 93x P/E overstates the steady-state multiple. (3) Backlog gives unusual 12–18-month forward visibility for an equipment manufacturer.

Underwritten incorrectly (or too optimistically): (1) That growth and margin recovery are simultaneous. They are structurally in tension — the fastest-growing segment (BASX, 24–27% GM) is dilutive, and consolidated gross margin was still falling in Q1-26 to 25.1% (§7.5). The market is paying for a 19% peak-margin recovery on a mix that mechanically resists it. (2) That AAON deserves a Vertiv-level EV/EBITDA (~48x) as a one-fifteenth-the-size, price-taking follower with no structural switching cost in data center (§3.3) and a shrinking legacy moat (−8.3% AAON-branded in 2025). (3) That the capital cycle stays favorable — the entire industry is adding capacity into the same demand (§3.5; AAON’s own ~$205M/yr capex), the textbook setup for margin mean-reversion that the bull multiple ignores. (4) That the cash economics support the multiple — AAON generated ~$0.5M of operating cash flow and ~−$204M of FCF in 2025, funding its dividend and buyback off the revolver (§7.5); a name at ~7.8x EV/revenue producing no free cash flow is being valued on a future cash profile it has not yet demonstrated. (5) The forward consensus itself, fully credited at a rich 30x, lands below today’s price — the market is discounting beyond consensus.

10.6 Verdict — what must be true, and how probable

For today’s ~$132 / ~$11.3B EV to be merely fair (zero forward return), AAON must roughly quadruple net income to ~$370–440M and more than double EBITDA to ~$450–565M within ~5 years — which requires revenue of ~$2.8–3.7B (a 12–21% CAGR) AND an operating margin recovered to 15–19% AND a sustained premium exit multiple of 25–30x. To earn even a market return from here, the bar rises to ~$580–700M of net income (a 5–6x increase off the FY2025 trough). The price is underwriting the bull scenario — sustained 20%-area data-center growth and a snap-back to near-peak margins and a durable growth multiple — three things that the segment mix, the capital cycle (§3.5), and the still-compressing Q1-26 gross margin make difficult to achieve together.

Probability assessment: the growth leg is probable (high confidence); the margin-recovery-to-peak leg is possible but unproven and structurally headwinded (low-to-medium confidence — most recent quarter disconfirms); the durable-premium-multiple leg depends on the capital cycle staying favorable, which the Marathon base rate argues against over a five-year horizon (low-to-medium confidence). The conjunction of all three at once — which today’s price requires — is low-probability. The scenario set is sharply asymmetric: a fully-realized bull case offers ~33% upside over five years (~6%/yr) against a credible bear case of ~75% downside, with the most defensible base case (~15% op margin, low-teens growth) still implying a price ~30% below today. The honest valuation conclusion is that AAON is priced for a near-best-case resolution of a genuine tension it has not yet resolved, at the 96th percentile of its own ten-year history, with no free cash flow to anchor the multiple — the embedded expectations are demanding and the asymmetry is unfavorable.

Verdict: The market is correctly pricing AAON’s data-center growth but incorrectly pricing a simultaneous margin snap-back and durable premium multiple as near-certain. What must be true — ~$2.8B+ revenue at ~15–19% operating margins capitalized at 25–30x within five years — is a low-probability conjunction; today’s price discounts the bull path and offers little compensation for the base and bear paths that the segment economics and capital cycle make more likely.

11. Variant Perception

This section synthesizes the preceding analysis into the consensus view, the strongest bull and bear cases, the handful of assumptions that actually decide the outcome, and the specific evidence that would falsify each side. No recommendation or price target appears here; those are reserved for the opening Claude’s Take block.

11.1 The Consensus Belief — what the ~95x trailing multiple is pricing

The market is treating AAON as a premium quality-compounder that has earned a temporary, self-inflicted demotion into a once-in-a-decade secular tailwind. The consensus narrative, distilled, runs as follows: AAON is a structurally advantaged, engineering-led niche HVAC franchise whose 2025 was a transient trough — the unhappy collision of an EPA-mandated A2L refrigerant transition (R-454B, effective 1/1/2025), a botched Longview ERP go-live (4/1/2025) that starved Tulsa of coils, and start-up under-absorption at the new Memphis plant ($16.1M of overhead booked into the AAON Oklahoma segment with no matching revenue). Strip those one-timers out, the argument goes, and the real AAON is the one whose order book just more than doubled to $1,828.5M (+110.9%) at year-end 2025 and $2,129.5M (+107.4%) at Q1-2026, powered by a BASX data-center backlog up 141.3% to $1.30B. The market is underwriting a sharp earnings snap-back — the consensus EPS path runs $1.42 TTM → $2.26 CY → $3.29 NY (a ~2.3x rebuild over two years) — and a margin recovery to management’s guided 29–31% gross margin (from 26.7% in 2025), validated by a Q1-2026 print that showed revenue +54.3%, AAON Oklahoma segment GM recovering to 26.3% “due to realization of price increases,” and operating margin ticking up to 11.5%. (FACT — consensus EPS; FY2025 10-K; Q1-2026 10-Q; Q4-2025 earnings call.) On this reading, ~41x forward / ~45x EV/EBITDA is the price of admission to a >20% CAGR liquid-cooling pool ($5–6.7B in 2025 → $16–29B by 2030–33 at ~20–26% CAGR) where AAON is an early, credible, engineering-differentiated supplier with proprietary CDU and free-cooling-chiller products and named hyperscaler customers (Microsoft, AWS, Google Cloud, QTS, Applied Digital). The third-party Wall Street target of ~$143 (context only; not adopted) embeds essentially this view.

The consensus is not naive — it is anchored to two genuinely strong facts (the backlog and the Q1-2026 reacceleration). Its vulnerability is what it elides: that the backlog is overwhelmingly a lower-margin business, that the legacy premium franchise is shrinking in absolute terms, and that the 34%-gross-margin AAON it implicitly extrapolates from is gone on mix, not just on cyclical noise.

11.2 The Strongest Bull Case

The bull case is real and must be stated at full strength, because it is the reason the stock has tripled off its $62 low. AAON has bolted a genuine secular growth engine onto a defensible niche franchise, and the order book proves the demand is here now, not hoped-for.

1. The backlog is the single best forward indicator in the file, and it is exploding. Total backlog +110.9% to $1.83B (YE25) and +107.4% to $2.13B (Q1-26); BASX book-to-bill of ~2.4x in FY2025. This is not a soft pipeline — the majority is “firm,” and it gives 12–18-month-plus revenue visibility into a pool growing 20%+ annually. (FACT — FY2025 10-K Backlog; Q1-2026 10-Q.)
2. The 2025 margin collapse is mostly self-inflicted and identifiable — therefore recoverable. The competitive-position falsification test is decisive on this point: pricing power is intact. AAON took a 3.0% price increase 1/1/2025, a 6.0% tariff surcharge 4/1/2025, and Q1-2026 AAON Oklahoma GM recovered to 26.3% “due to realization of price increases and increased volume.” The moat is being mix-shifted, not competed, away — the bear’s “premium is eroding” claim is unsupported by the pricing evidence. (FACT — FY2025 10-K; Q1-2026 10-Q MD&A.)
3. The OCF collapse is a backlog-funding artifact, already reversing. OCF fell to $0.53M in 2025, but the ~$236M working-capital swallow was contract-asset and inventory build pre-funding a doubling order book, not an accruals-quality fraud signal. Q1-2026 OCF swung positive to +$34.0M and accounts receivable reversed +$24.3M — the scariest FY2025 line began unwinding in one quarter. (FACT — FY2025/Q1-2026 cash-flow statements.)
4. Capacity is now in place and ramping into the demand. Memphis (787k sq ft) hit its first profitable quarter in Q4-2025; Longview West Plant expansion completed Jan 2025. The ~$418M of 2024–2025 capex is largely behind the company, setting up FCF inflection as utilization climbs and the working-capital build normalizes.
5. Alignment and governance are sound. Founder Asbjornson still owns 16.8% (13.73M shares); long-term incentives use a relative-TSR PSU (which paid below target at 83.2% for 2023–25); buybacks have neutralized SBC dilution; split-adjusted net dilution is only ~3.7% over four years.

The bull’s distilled claim: 2025 was the trough of a quality compounder mid-transformation; consensus EPS ($1.42→$2.26→$3.29) and a 29–31% GM recovery are conservative against a $2.1B order book, and the multiple compresses as earnings rebuild.

11.3 The Strongest Bear Case

The bear case is, in my assessment, the better-evidenced of the two — and it does not require the demand to disappear. It requires only that the quality of the business be honestly re-rated to what the numbers now show.

1. The premium franchise — the entire reason AAON ever deserved a premium multiple — is structurally shrinking. AAON-branded (legacy) revenue fell −2.6% in 2024 and −8.3% in 2025 to $894.3M, while BASX-branded grew +143.5% to $547.8M (now 38% of revenue vs 13% in 2022). This is the decisive fact: the high-margin core is in absolute decline, and every dollar of growth (and then some) is lower-margin data-center work. The company is not compounding its moat; it is mix-shifting out of it. (FACT — FY2025 10-K Note 3 brand disaggregation.)
2. The gross-margin reset is partly structural, not merely cyclical — and Q1-2026 proves it. Consolidated GM fell 34.1% → 26.7%, and the single most damning disconfirming fact for the snap-back thesis is that Q1-2026 consolidated GM was 25.1% — below the FY2025 full year — despite revenue +54.3%. The reason is mix: the fastest-growing segment (BASX, ~24–27% GM; Coil, ~21% GM) is structurally below the legacy ~35–37% franchise. Management’s own 29–31% guide concedes a permanent step-down from the 34% peak. A “recovery” to 29–31% is the dilution, not a refutation of it. (FACT — Q1-2026 10-Q MD&A; FY2025 10-K segment tables.)
3. Quality of earnings has deteriorated and the balance sheet is stretched to near-zero slack. OCF of $0.53M against $107.6M net income (0% cash conversion); FCF of −$204M in 2025; cash of $13K; the company now runs day-to-day on a revolver expanded three times in one year to $600M and drawn to $425.2M at Q1-2026 with only $173.5M of availability left. The dividend and buyback in 2025 were funded with borrowed money while FCF was deeply negative. ROIC collapsed to ~9.3% — roughly its own cost of capital. A genuine moat sustains returns through an air-pocket; AAON’s slid to break-even economics in a single bad year. (FACT — FY2025 10-K; Q1-2026 10-Q.)
4. The growth is single-end-market, customer-concentrated, and cancellable. Three customers were each >10% of revenue in 2025 (up from two in 2024); the order book is ~76% BASX/data-center; and the 10-K explicitly warns that data-center orders “have more risk and we often see shifts in timing, cancellations and re-issuances” — unlike historically near-uncancellable RTU orders. Management’s own 2026 guide of only ~25% BASX revenue growth against +141% backlog growth is a tacit admission that the headline backlog overstates near-term convertible revenue. The growth is a derivative of a handful of hyperscaler programs in a Marathon mid-boom where capacity (Vertiv’s $15B backlog and >$1B M&A, Modine +50–70%, JCI/Silent-Aire, Munters, STULZ) is racing in across the entire industry — the textbook configuration that precedes margin mean-reversion. (FACT — FY2025 10-K; competitor releases.)
5. The insider tape and management transition offer zero support for the re-rating. ~$51M of overwhelmingly discretionary (only 6 of 226 dispositions were 10b5-1) insider selling over 24 months versus only ~$0.6M of buying — and no insider has bought a single share above ~$82 against a current ~$132. The May–June 2026 cluster (departed-CEO Fields ~$5.6M, CAO Thompson to zero direct shares, EVP Wichman) is uniform and into strength. Layered on top: a CEO transition (May 2025), three senior-finance changes in under 12 months, and two AAON-unit GM reshuffles in 12 months. None of it is a fraud flag, but at ~95x trailing earnings it is a conspicuous absence of conviction from the people who know the business best. (FACT — Form 4 corpus; 8-K timeline.)

The bear’s distilled claim: AAON has converted a high-ROIC, self-funding, ~35%-margin niche franchise into a lower-margin (~27%), capital- and working-capital-hungry, customer-concentrated, revolver-dependent data-center contract manufacturer — and is being priced at ~95x trailing / ~45x EV/EBITDA as if the old quality were intact. There is no margin of safety against any slip in the unproven 2026 margin recovery.

11.4 The 3–5 Assumptions That Matter Most

Everything reduces to five swing variables. The first three are where the market is most likely wrong.

11.5 What Evidence Would Falsify Each Side

Falsify the bear (confirm the bull): Consolidated gross margin inflects up through 2026 toward the guided 29–31% while the BASX/data-center mix continues to rise — i.e., margin recovery survives the mix headwind, not despite a pause in it. Concretely: two consecutive quarters of consolidated GM ≥29% with BASX share of revenue flat-to-up, and OCF/FCF turning durably positive (multi-quarter, not one-print) as the working-capital build reverses without re-accumulating. That would prove the 2025 trough was genuinely transitory and that the data-center business can be grown at near-franchise economics — vindicating the multiple.

Falsify the bull (confirm the bear): Consolidated gross margin stalls in the mid-20s through 2026 (the Q1-2026 25.1% persists rather than inflects), legacy AAON-branded revenue posts a third year of decline or flat results, OCF/FCF stays weak as working capital fails to release, or a single >10% data-center customer cancels/defers a major program and the “firm” backlog visibly shrinks. Any one of these would expose the re-rate as having capitalized a cyclical, lower-quality, capacity-cycle peak as if it were durable franchise compounding.

11.6 Verdict — Where Is the Market Most Likely Wrong, and in Which Direction?

The market is most likely wrong on quality and margin level, in the direction of having over-extrapolated the old AAON onto the new one — i.e., the risk is skewed to the downside on the multiple, not on the demand. The bull case correctly identifies a real backlog and a real secular tailwind; the consensus error is not “the data center isn’t growing” — it manifestly is — but rather conflating a lower-margin, customer-concentrated, working-capital-heavy, cancellable, sub-scale data-center contract business with the high-ROIC, self-funding, ~35%-margin niche franchise that historically justified a premium multiple. Three facts the market is under-weighting carry the verdict: (1) the legacy premium core is in absolute decline (−8.3%), so the moat is shrinking as a share of the whole; (2) Q1-2026 consolidated GM of 25.1% came in below the FY2025 full year despite +54% revenue, directly contradicting the snap-back thesis and confirming a structural (not merely cyclical) margin reset; and (3) the company funds its growth, dividend, and buyback on a revolver against $13K of cash while ROIC sits at its cost of capital — there is no balance-sheet or earnings-quality buffer beneath a ~95x trailing / ~45x EV/EBITDA valuation that demands the snap-back arrives essentially on schedule. The defensible variant-perception conclusion is that AAON is a genuinely good business that has been re-domiciled into a structurally lower-quality mix and is priced as though that re-domiciling never happened — the asymmetry is unfavorable because the multiple already capitalizes a margin and returns recovery that the most recent quarter does not yet show, leaving the entire downside protection resting on an unproven 2026 inflection. The single most likely way the market is wrong is that steady-state margins and returns settle structurally below the 2023 peak even after the one-timers clear, and the multiple — not the demand — is what corrects.

12. Fact vs. Interpretation

The table separates what the filings establish (Fact — every figure below ties to EDGAR XBRL and/or the FY2025 10-K / Q1-2026 10-Q, independently verified) from the judgment this article draws (Interpretation). The investment debate lives almost entirely in the right-hand column.

13. Open Questions

These are the genuinely unresolved questions that would most change the analysis. They are deduplicated and ordered by how decisively each would move the thesis; several are answerable from the next one or two quarterly filings.

1. Steady-state gross margin — structural or transitional? Once Memphis/Longview capacity is absorbed and the ERP stabilizes, does consolidated gross margin recover toward management’s guided 29–31% (and hold there while the dilutive BASX mix keeps rising), or is the mid-to-high 20s the new structural ceiling? Q1-2026’s 25.1% — below the trough year — has not yet answered this, and it is the single most decisive open question. (Resolvable from the next two 10-Qs.)

2. Single-customer concentration. The 10-K discloses three customers each ≥10% of revenue in 2025 but does not quantify whether any single data-center customer is >15–20%. Given the explicit disclosure that data-center orders are cancellable and re-timed, the “firmness” of the $2.1B backlog — and the tail risk to revenue — hinges on this undisclosed number.

3. Does free cash flow ever inflect positive, durably? The Q1-2026 working-capital reversal (+$34M OCF, +$24.3M AR collected) is one quarter. At a permanently BASX-heavy, project-billed mix with ~13%-of-sales capex, is the business structurally dependent on the revolver, or can it self-fund its capex, dividend, and buyback within a year or two?

4. Is the legacy −8.3% decline transitory destocking or structural share loss? A2L pre-buy/destocking would normalize in 2026; genuine share loss or neglect (as capital and attention shift to data center) would not. AAON-branded backlog ticking down sequentially (YE25 → Q1-26) and management’s “flattish 2026” commentary argue this is more than a one-year air-pocket — but two more quarters will tell.

5. Can AAON win durable data-center share, or is it an overflow taker? Against Vertiv ($15B backlog, >$1B M&A), JCI/Silent-Aire, Modine, Munters and STULZ — all adding capacity — is AAON/BASX a differentiated supplier or a sub-scale price-taker into a capital cycle that historically mean-reverts margins? There is no rep-channel lock-in or switching cost in this pool.

6. Will the new capex base earn its cost of capital? Capex ran 2–3× depreciation for three years into a lower-margin mix while ROIC halved to ~9%. Does the new capacity ultimately earn AAON’s historical ~25% ROIC once utilized, or has the mix shift permanently reset consolidated returns to the low-teens? And does 2026 capex step down from the ~$190M peak, signaling the build is complete?

7. Is consensus EPS itself too high? The path $1.42 → $2.26 → $3.29 underwrites a sharp margin recovery that Q1-2026’s still-falling gross margin does not yet support. If consensus is optimistic, the “justify-today” valuation bar is even further above the current price than the headline forward multiple implies.

8. Accounting/estimate risk during the scale-up. With a new CFO (Andy Cheung, April 2026 — the third senior-finance change in under 12 months) and $547.8M of BASX revenue recognized over-time on estimated cost-to-complete (the auditor’s sole critical audit matter), is there risk of future cost-to-complete revisions reversing recognized margin, or changes to percentage-of-completion judgments and ERP/software capitalization policy?

9. Will the Compensation Committee fix the incentive design? Given the visible returns deterioration, will the annual bonus (currently revenue- and budgeted-profit-weighted, with no returns or per-share metric) be revised to align pay with per-share value — or will it continue to reward the capex super-cycle?

10. BASX deal economics (data gap). The FY2021 purchase-price allocation and implied acquisition multiple are not in the trailing-36-month corpus, so the “full price” characterization of the BASX deal is an interpretation rather than a sourced figure; the FY2021 10-K would resolve it.

14. What Must Be True (Bull and Bear)

Each side reduced to its load-bearing claim plus a single concrete, near-term falsification test to monitor.

The Bull Case — What Must Be True

The premium multiple is justified only if AAON can grow the data-center business at near-franchise economics — i.e., the 2025 margin/cash trough was genuinely transitory and the BASX mix scales without permanently re-rating consolidated returns lower. Specifically: (a) consolidated gross margin inflects up through 2026 toward 29–31% even as BASX/data-center rises as a share of revenue; (b) the $2.1B backlog converts on schedule without a major hyperscaler cancellation; © working capital releases and FCF turns durably positive so the business stops funding itself, its dividend, and its buyback on the revolver; and (d) consensus EPS rebuilds along the $1.42 → $2.26 → $3.29 path. In short, the data-center pivot must prove to be growth and quality, not growth at the expense of quality.

Bull falsification test (single, concrete): If, over the next two reported quarters (Q2 and Q3 FY2026), consolidated gross margin fails to climb above ~28% — i.e., it remains stuck near the Q1-2026 25.1% level — while BASX/data-center revenue share holds or rises, the “transitory trough” thesis is falsified: the margin reset is structural mix, not a recoverable air-pocket, and the ~95x/~45x multiple is being paid for compounding that no longer exists at the implied return.

The Bear Case — What Must Be True

The skeptical case is correct if AAON has structurally re-domiciled into a lower-margin, lower-return, capital-hungry, customer-concentrated data-center contract business while being priced as though the old ~35%-margin, high-ROIC franchise were intact. Specifically: (a) steady-state gross margin settles materially below the 34% peak (high-20s to ~30%) on permanent mix; (b) the legacy premium franchise continues to shrink or stagnate, so the moat is a declining share of the whole; © ROIC stays near or below cost of capital on the enlarged, lower-returning asset base; and (d) the multiple — not the demand — is what ultimately corrects as the market re-rates quality. The demand can be entirely real and the bear case still holds: it is a quality and price argument, not a demand argument.

Bear falsification test (single, concrete): If, within the next two reported quarters, consolidated gross margin durably re-rates back to ≥31% (toward/through the top of guidance) and legacy AAON-branded revenue returns to year-over-year growth, the structural-dilution thesis is falsified: the 2025 trough was cyclical, the franchise re-accelerated as the one-timers cleared, and AAON is once again the high-margin compounder the multiple assumes — in which case the bear is paying too much attention to a single bad year.

Synthesis: The two falsification tests share one observable — the trajectory of consolidated gross margin over the next two quarters, read against the direction of the mix. That single number, watched in context, will resolve the central variant-perception question faster than any other data point: whether AAON’s 2025 was a transitory trough in a quality compounder (bull) or the first clean look at a permanently lower-quality, lower-return business priced as if it were the old one (bear). On the weight of the evidence assembled across the analysis — a shrinking premium core, a Q1-2026 margin that came in below the trough year, a balance sheet at $13K of cash, and insiders selling into the re-rate — the burden of proof sits squarely with the bull, and the market’s most probable error is having capitalized a cyclical, lower-quality, capacity-cycle peak as durable franchise compounding.

APPENDIX A — Standard Diligence Questionnaire

Supplemental to the research memo. Answers are grounded in the FY2025 10-K (filed 2026-03-02), Q1-2026 10-Q (2026-05-07), the 2026 DEF 14A, the trailing-36-month SEC corpus, and SEC EDGAR XBRL. Fact / Interpretation / Assumption are labeled where it matters. Greenwald (Competition Demystified) and Marathon (Capital Returns) frameworks applied where they add insight.

General

What thoughtful questions have other investors asked about this company? The sophisticated debate centers on five questions: (1) Is the 2025 margin collapse (gross margin 34%→27%) cyclical/self-inflicted or a structural mix-driven reset? (2) Is the $2.1B backlog “firm,” near-term-convertible revenue, or a multi-year, cancellable, customer-concentrated pipeline? (3) Can AAON earn franchise-like returns in data-center cooling as a sub-scale follower to Vertiv, or is it a price-taker? (4) Does the working-capital-heavy, capex-intensive new mix ever generate free cash flow, or is it structurally revolver-dependent? (5) Does a ~93× trailing / ~45–50× EV/EBITDA multiple survive if the margin recovery disappoints? The bulls focus on backlog and TAM; the bears focus on the shrinking legacy core, the still-falling Q1-2026 gross margin, and the no-FCF balance sheet.

Cyclicality & Earnings Nature

Are earnings at a cyclical high or low? A cyclical/transitional low. FY2025 net income ($107.6M) is down ~39% from the 2023 peak ($177.6M); operating margin (10.1%) is half the 2023 level (19.5%). (Fact.) Consensus underwrites a recovery (EPS $1.42 TTM → $2.26 → $3.29). Interpretation: earnings are depressed, but the normalized level is below the 2023 peak because of permanent mix shift — so this is a trough, but not against a recoverable 2023-level baseline.

Driven by the external environment or internal actions? Both, roughly half each. External/regulatory: the EPA-mandated A2L (R-454B) refrigerant transition (effective 1/1/2025) caused supply-chain disruption and a destocking air-pocket; soft nonresidential construction hurt legacy volume. Internal/self-inflicted: a botched Longview ERP go-live (4/1/2025) starved Tulsa of coils; Memphis start-up loaded $16.1M of unabsorbed overhead into the AAON Oklahoma segment; opex was built ahead of volume. (Fact — FY2025 10-K MD&A.)

How stable are revenues? Decreasingly stable. The legacy RTU business was replacement-anchored and relatively stable; the new BASX/data-center mix (38% of revenue) is project-based, lumpy, customer-concentrated (three customers ≥10% in 2025), and explicitly cancellable/re-timeable per the 10-K. Interpretation: revenue visibility improved (backlog doubled) but revenue stability/quality declined.

Outlook for products/services? Bifurcated. Data-center cooling demand is strong (backlog +107% YoY at Q1-2026; TAM growing ~20%+ per third-party estimates); legacy commercial RTU is “flattish 2026” per management with AAON-branded backlog ticking down sequentially. (Fact + third-party estimate.)

How big will this market be? The data-center cooling/liquid-cooling pool is sized by third-party vendors (Grand View, MarketsandMarkets, Dell’Oro et al.) at ~$5–6.7B (2025) growing to ~$16–29B by 2030–33 (~20–26% CAGR) — third-party estimates with wide dispersion, not facts. The legacy US commercial RTU market is ~$2.56B by 2030 (~4.9% CAGR). Domestic-led with some Canada; AAON sells in the US and Canada.

Business Quality & Competitive Moat

Is the industry getting more or less competitive? More, in the segment that matters for growth. Data-center thermal is a late-stage capital cycle with capacity flooding in from Vertiv ($15B backlog, >$1B M&A), JCI/Silent-Aire, Modine, Munters and STULZ. Legacy RTU is moderately concentrated and more stable, but AAON’s own legacy franchise is shrinking within it.

How profitable is the business (ROIC, ROE)? Historically excellent, now compressed. ROIC ~24% (2023) → ~9.3% (2025); ROE ~27% → ~12%. (Fact — computed from EDGAR.) At the 2025 trough, returns sit at roughly the cost of capital — a genuine moat should hold returns above WACC through an air-pocket, so the slide signals the buffer is thin.

How profitable is the industry — competitors, barriers to entry? Legacy RTU is a structurally good industry (moderate concentration, regulatory/certification barriers, durable replacement demand). Data-center cooling is structurally hot but margin-pressured by capacity inflows. Peer operating margins (per peer releases): Lennox ~20%, Trane ~18.6%, Carrier ~16.5–17%; AAON at 10.1% (2025) fell below all of them.

Can the business be easily understood? Mostly. It is an equipment manufacturer with a clear (if recently complicated) segment/brand structure. The complications: the data-center business straddles two reportable segments, and Memphis overhead is booked in the “wrong” segment — both obscure the true mix. Interpretation: understandable, but the reported segments require unpacking (use the brand disaggregation).

Can it be undermined by foreign low-cost labor? Low risk. AAON manufactures domestically (Tulsa, Longview, Memphis, Redmond, Parkville), and the value proposition is engineered configurability, lead time, and serviceability, not lowest-cost commodity production. Tariffs (a 6% surcharge added April 2025) are a cost input, not an existential threat.

Do brands matter? Moderately, via the specification/reputation loop, not consumer branding. The AAON name carries a real quality/serviceability reputation among engineers and owners that drives replacement stickiness; BASX is building a reputation in data-center thermal. Patents are explicitly immaterial. Interpretation: brand matters as part of the demand-side moat in legacy, not in data center.

What is the nature of competition? In legacy: total-value-proposition (quality, efficiency, lifecycle cost), not initial price, against far larger majors. In data center: project-by-project competitive bidding against larger, vertically-integrating thermal players for sophisticated hyperscaler buyers.

Customers’ switching costs? Real and narrow in legacy (engineered-in specs, curb/footprint/controls lock-in, rep relationships drive replacement re-orders); effectively absent in data center (multi-source, price-disciplined monopsony buyers). This asymmetry is central: the growth segment lacks the very mechanism that gives AAON its moat.

Financial Condition & Balance Sheet

Assets not fully recognized on the balance sheet? The Norman Asbjornson Innovation Center (NAIC) test lab is a genuinely differentiated R&D/intangible asset carried at cost. The legacy brand/reputation and rep relationships are unrecognized intangibles. None is a hidden liability.

Off-balance-sheet liabilities? Minimal. New Markets Tax Credit structures (2019/2023/2024) carry modest low-rate debt with potential future forgiveness — not a hidden liability. Operating leases are standard. No material off-balance-sheet exposure identified.

How conservative is the accounting? Generally conservative, with one watch item: BASX over-time (percentage-of-completion) revenue recognition — $547.8M recognized on estimated cost-to-complete — is the auditor’s (Grant Thornton, since 2004) sole critical audit matter. Non-cash adjustments and reserves are stable; there is no aggressive capitalization masking opex. Interpretation: conservative overall, but the over-time revenue estimate is a legitimate area to monitor, especially amid finance-team turnover.

How CapEx-hungry is the business? Very, currently. Capex ran 2.2–3.1× depreciation for three years (~$490M cumulative 2023–25; ~13% of sales), driving PP&E from $305M (2022) to $631M (2025). 2026 capex is guided ~$190M. Interpretation: a structurally capital-hungry investment phase; whether it steps down post-2026 is an open question.

Capital Allocation & Management

How much FCF does the business generate; how is it used; what is the philosophy? Currently negative: FCF was −$204M in 2025 (OCF $0.53M − capex $190.6M) and −$20.6M in 2024. (Fact.) The philosophy is growth-first: fund a data-center capacity build, supplemented by a small dividend (~30% payout) and opportunistic buybacks — but in 2025 the dividend and buyback were funded on the revolver, not free cash flow.

Significant acquisitions recently? Only BASX (December 2021, ~$100M cash + up to $80M contingent stock + $22M real estate). None since. The BASX deal is the strategic pivot; it has delivered growth but diluted margins and returns — a growth purchase, not a quality purchase.

Buying back shares? Yes, modestly and opportunistically: $100M in 2024 (well-timed against a de-rated stock at ~$81 average), $30M open-market in 2025. Buybacks have roughly neutralized SBC dilution; split-adjusted net dilution is only ~3.7% over four years. (Fact.)

Issuing large amounts of new shares to insiders? No. SBC is contained (~$18.0M / 1.2% of revenue in 2025) and offset by buybacks. The 2021→2022 share-count jump was the BASX stock consideration plus a 3-for-2 split, not chronic dilution.

Compensation policy of directors/management? The concern. The annual cash bonus is weighted Operating Profit 67% / Net Sales 33% against budget — no ROIC/ROE/EPS/FCF metric — structurally rewarding growth and budget-hitting over per-share returns. The long-term plan is better aligned (3-yr relative-TSR PSUs, 50% of the mix, with a negative-absolute-TSR cap; paid below target at 83.2% for 2023–25). Ownership guidelines (CEO 6× salary), retention, anti-hedging/pledging, and clawbacks are robust. (Fact — 2026 DEF 14A.)

Motivations of management? Founder Asbjornson (age 90, 16.8% / 13.7M shares) is strongly aligned but is now an ordinary director, not chairman or operator. The operating leadership owns little (CEO Tobolski <1%, ex-CEO Fields <1%) and has been a net seller. Interpretation: founder alignment is real; current-operator conviction, judged by the insider tape (~$51M sold vs ~$0.6M bought over 24 months, none above ~$82), is conspicuously absent at the re-rated price.

Valuation & Market Data

Is the stock an ADR, MLP, or K-1 issuer? No — a standard US C-corporation common stock (NASDAQ: AAON), Form 1099 dividends. No K-1.

Dividend policy? Small and growing: $0.40/share in 2025 (moved to $0.10/quarter), ~0.28–0.30% yield, ~28–30% payout. Raised steadily, but not covered by free cash flow in 2024–2025.

How profitable is the business? At a trough: 26.7% gross / 10.1% operating / 7.5% net margin in 2025, down sharply from the 2023 peak (34.1% / 19.5% / 15.2%). Returns (ROIC ~9.3%, ROE ~12%) are at roughly the cost of capital.

Is net income diverging from cash from operations? Dramatically — the central quality-of-earnings flag. FY2025 net income $107.6M vs OCF $0.53M (0% cash conversion vs 114% in 2024), driven by a ~$236M working-capital build (receivables, contract assets, inventory) pre-funding the backlog. Q1-2026 began reversing (+$34M OCF, +$24.3M AR collected) — reassuring but one quarter. Interpretation: largely a recoverable backlog-funding artifact, not fraud, but it re-rates working-capital intensity permanently higher.

Risks & Downside

What factors would cause the stock to decline? A gross-margin recovery that stalls in the mid-20s (the Q1-2026 25.1% persists); a third straight down year in legacy AAON-branded revenue; a >10% data-center customer cancellation shrinking the “firm” backlog; FCF staying negative and revolver availability tightening; or simply multiple compression from a 96.5th-percentile-of-history valuation on any disappointment. The valuation offers no margin of safety.

Risk of a catastrophic loss? Low in the near term. Leverage is moderate (1.7–1.8× vs a 3.0× covenant), interest coverage ~8×, and demand is real. The balance sheet is stretched (cash $13K, revolver-dependent) but not distressed. Interpretation: the dominant risk is valuation/multiple de-rating, not solvency.

Chance of a total loss? Very low. AAON is a profitable, ~$1.4B-revenue manufacturer with real assets, a doubling backlog, and a founder with a 16.8% stake; a permanent total loss would require a severe, prolonged demand collapse plus a refinancing failure — not the base case. The realistic downside is a large price decline (bear scenario ~$33, ~75% below today), not a zero.

Recent News & Events

Has the business environment changed recently? Yes, materially over two years: the A2L refrigerant transition reset the legacy demand pattern; the AI data-center capex wave detonated the BASX order book (+107% YoY); and consolidated margins halved on execution + mix. The environment shifted AAON from a niche margin specialist toward a data-center thermal contractor.

Significant acquisitions? None since BASX (2021).

Change in accounting policies? No major policy change, but a 3-for-2 stock split (Aug 2023), a segment-reporting recast (moving intercompany coil sales to cost, which re-stated segment margins), and the ERP go-live (April 2025) are relevant. The over-time BASX revenue remains the critical audit matter.

Recent changes — new markets, facilities, management? All three. Markets: deepened into data-center liquid cooling (proprietary CDU and free-cooling-chiller products). Facilities: the 787,000-sq-ft Memphis plant (bought Dec 2024, first profitable quarter Q4 2025) and the Longview West Plant expansion (completed Jan 2025). Management: CEO transition (Matt Tobolski, BASX co-founder, effective May 2025, succeeding Gary Fields); three senior-finance changes in under 12 months (including new CFO Andy Cheung, April 2026); two AAON-unit GM reshuffles in 12 months. (Fact — 8-K timeline / DEF 14A.)

APPENDIX B — Source Appendix

Public primary sources underpinning the verified claims. All SEC filings accessed via EDGAR; XBRL facts via the SEC companyfacts/companyconcept API. Access date for all web/third-party sources: 2026-06-08.

A. SEC filings (primary — EDGAR, CIK 0000824142)

B. SEC EDGAR XBRL (primary, structured data — companyfacts/companyconcept, CIK 0000824142)

Used to independently confirm the multi-year series (in $000 unless noted): RevenueFromContractWithCustomerIncludingAssessedTax; GrossProfit; OperatingIncomeLoss; NetIncomeLoss; NetCashProvidedByUsedInOperatingActivities (534 for FY2025); InventoryNet; AccountsReceivableNetCurrent; ContractWithCustomerAssetNetCurrent; CashAndCashEquivalentsAtCarryingValue (13); Assets; StockholdersEquityIncludingPortionAttributableToNoncontrollingInterest; LineOfCredit; PaymentsToAcquireMachineryAndEquipment / PaymentsToAcquireProductiveAssets (capex); PaymentsToAcquireBusinessesNetOfCashAcquired (103,430, 2021); PaymentsToAcquireBuildings (22,000, 2022); Goodwill; PaymentsOfDividendsCommonStock; PaymentsForRepurchaseOfEquity; ShareBasedCompensation; ResearchAndDevelopmentExpense; DepreciationDepletionAndAmortization; CommonStockSharesOutstanding; WeightedAverageNumberOfDilutedSharesOutstanding. (Source: https://data.sec.gov/api/xbrl/companyfacts/CIK0000824142.json and companyconcept endpoints; accessed 2026-06-08.)

C. Public industry, market-size, and peer sources (third-party — treat as estimates/ranges, not primary facts)

• Data-center / liquid-cooling TAM: Grand View Research; MarketsandMarkets; Virtue Market Research; MarknTel Advisors; Dell’Oro Group (DCPI). Figures span a wide cross-vendor range ($5.1–6.7B 2025 → $16–29B 2030–2033; ~20–26% CAGR) — present with vendor attribution and as a range, not a point fact.
• US commercial HVAC / rooftop-unit market: Grand View Research, “U.S./North America HVAC Rooftop Units Market” (~$2.56B US by 2030 / ~$11.1B N.A. 2024; ~4.9–5.2% CAGR).
• Data-center construction spending: U.S. Census Bureau construction put-in-place (primary public series), reported via Wolf Street (2026-02-28) and Marketplace (2026-02-27): +>100% in two years / +344% from 2020 to ~$41B (2025); ~$50.7B SAAR by April 2026.
• Refrigerant transition: EPA AIM Act HFC phasedown / A2L (R-454B, <750 GWP, effective 2025-01-01); DOE 2023 commercial RTU efficiency standard (Federal Register); ACHR News coverage.
• Peer FY2025 results (company releases / SEC filings): Lennox International (8-K/IR), Trane Technologies (Q4/FY2025 release), Carrier Global (8-K earnings exhibits), Vertiv Holdings (FY2025 results, ~$15B backlog). Used for relative size and margin comparison; exact peer percentages taken per each company’s release (not re-derived in this sweep).
• Public earnings transcript: AAON Q4-2025 earnings call, The Motley Fool, 2026-03-02 (2026 guidance 18–20% revenue / 29–31% GM; BASX book-to-bill ~2.4x) — public secondary; management commentary, treat as hypothesis.