Advanced Energy Industries, Inc. (NASDAQ: AEIS) — The Picks-and-Shovels Power Play, Re-Rated to Perfection on an AI Build-Out It Doesn’t Lead

Independent equity research, as of 2026-06-27. The body (Sections 1–15) carries no recommendation and no price target; the single exception is the labeled Claude’s Take block immediately below.

⚡ Claude’s Take

This block is the author’s own subjective opinion. It is not investment advice and is general information only. Everything below it (Sections 1–15) takes no position and sets no price target.

Verdict: HOLD / AVOID-here / accumulate only on a cyclical reset. Not a short. Conviction: medium.

Advanced Energy is a genuinely better company than it was three years ago — gross margin has climbed from ~35.7% to 40.1%, the data-center business it bought in 2019 (Artesyn) has quietly become the swing factor, more than doubling in a year, and its plasma-power franchise still holds co-leadership in RF generators for leading-edge etch and deposition. The problem is not the business; it is the price the market now pays for it. At ~$348, AEIS trades at roughly 39x FY26 non-GAAP EPS, ~25–27x EV/EBITDA, ~6x forward sales, and 10.6x book — its richest valuation ever (AZI composite 96th percentile; P/B 98.7th, P/S 98.5th of its own decade) — for a business whose through-cycle return on invested capital is ~8% on a total-capital basis and only reaches the low-teens net of cash in good years. That is a quality-compounder multiple stapled to a mid-cycle-return cyclical. The market is capitalizing the most cyclical revenue surge of the company’s history (data center +102% YoY, semiconductor riding a fresh WFE up-cycle) and the promise of 800-volt HVDC dollars that don’t meaningfully arrive until 2027–2028 — for a part where, tellingly, AEIS is absent from NVIDIA’s published 800V partner list, just like Vicor.

The framing is momentum/quality-at-a-rich-price, not value: the stock is +310% off its April-2025 low, beta ~2.0, with the entire move coming from multiple expansion on the AI narrative rather than a step-change in returns on capital. I’d want to own this franchise — net cash, no value-destructive M&A, a shrewd 0%-coupon 2031 convert that monetized the peak equity — but at roughly $180–230 (~22–27x non-GAAP EPS, P/B ~6–7x), a zone that credits the data-center growth and 43% GM goal without paying for perfection and a not-yet-earned re-rating to compounder status. The single fact that flips me bullish: durable evidence that the 800V/HVDC and eVoS/eVerest/NavX wins lift structural ROIC into the high-teens-plus (margin expansion that survives the next semi downturn). The single fact that flips me bearish: a data-center air-pocket (hyperscaler in-sourcing or demand-mix reset) plus a semi-cycle stall, which would expose a ~10% GAAP-operating-margin cyclical trading at 40x. Tag: “Forty-percent margins, mid-cycle returns, a top-percentile price.”

📈 Stock Price Action — Five-Year Event Map

Over the trailing five years AEIS went from a forgotten, range-bound cyclical to one of the loudest AI-power re-ratings in semicap. The arc: a 5-year low of ~$68 (June 2022), a long $80–115 range through 2022–early 2025 (cyclical trough plus a two-year industrial/medical inventory correction), then a near-vertical +310% move from ~$95 (April 2025) to an all-time high of ~$389 (May 1, 2026) as data-center revenue doubled and the market recast AEIS as an AI-infrastructure name. It currently trades at ~$348, ~10% below the ATH, with a 52-week range of roughly $93–$397. The price moves below are FACT (AZI price history); the attributed drivers are INTERPRETATION cross-referenced to earnings prints, 8-Ks, and the news feed.

Cycle narrative. (1–2) AEIS spent 2022–2024 as a left-for-dead cyclical: the 2022 growth-stock purge took it to ~$68, and a WFE downturn plus a punishing two-year industrial/medical inventory correction kept revenue and EPS flat in an $80–115 box. (3) Early 2025 added a tariff scare and a fresh semi air-pocket, marking the ~$89–95 bottom. (4–6) From April 2025 the story changed: the Artesyn-legacy data-center business went from a minor line to $587M (+107%) in FY2025 and kept doubling (Q1-26 +102%), gross margin pushed through 40% for the first time since 2019, and management raised the FY26 growth target twice — and the market re-rated the multiple from ~15x to ~40x non-GAAP, quadrupling the stock to a $389 ATH. (7) The recent ~10% pullback is consolidation after a parabolic move; the $1.0B zero-coupon 2031 convertible (May 2026) opportunistically termed out debt at the equity peak, and the sell-side piled on with $400–$450 targets. This is a re-rating, not just an earnings move — the central tension the rest of this memo examines.

1. Executive Summary

Advanced Energy Industries is a Denver-based maker of precision power-conversion, measurement, and control products. It sells into four markets: semiconductor process/plasma power (~47% of FY25 revenue — RF and pulsed-DC generators and matching networks injected into etch and deposition chambers, sold to wafer-fab-equipment OEMs and fabs), data-center computing (~33% — merchant/embedded server and rack power supplies, the legacy of the 2019 Artesyn acquisition), industrial & medical (~16%), and telecom & networking (~5%). FY2025 revenue was $1.80B (+21%); the stock has roughly quadrupled in fourteen months and trades near an all-time high.

The investment tension is valuation versus returns on capital, not growth. Three facts are simultaneously true. First, the business is genuinely improving: gross margin rose from ~35.7% to 40.1% (Q1-26), with a credible path to management’s 43% goal, and the data-center segment more than doubled to $587M in FY2025 and again in Q1-26 (+102% YoY). Second, the franchise is real but mediocre in returns: through-cycle ROIC is ~8% on a total-invested-capital basis (it cleared a ~9–10% WACC only in up-cycle years), gross margin is the lowest in its peer group (MKS ~47%, Vicor ~52%, Monolithic Power ~55%), and AEIS is a co-leader/#2 in plasma power and a merchant box-maker in data center — it does not lead the value chain. Third, the price is the richest in company history — ~39x FY26 non-GAAP EPS, ~25–27x EV/EBITDA, ~6x forward sales, 10.6x book (98.7th percentile of its own ten-year range).

Quality assessment: a decent-but-mediocre cyclical with a narrow, product-line-level moat — not a high-quality compounder. AEIS has genuine switching costs in plasma power (generators are qualified/“copy-exact” into OEM tools and fab recipes), but it is contested (#1 in conductor etch/deposition, #2/#3 in dielectric etch), faces powerful buyers on both sides (top three customers = 54% of FY25 revenue), and earns returns that only marginally exceed its cost of capital. The data-center business has no durable moat — it wins today on power density, efficiency, and speed-to-market against Delta, Lite-On, and Flex, in a fast-moving architecture transition where the silicon IP belongs to others (Infineon, Monolithic Power, Navitas, TI) and AEIS is absent from NVIDIA’s named 800V partner list.

Capital allocation is above-average but not elite. No value-destructive M&A and zero impairments; the May-2026 $1.0B zero-coupon convertible was a shrewd monetization of peak equity; the balance sheet is net-cash. But buybacks are pro-cyclical (heaviest at $80–120, ~$167M authorization sitting idle at $300+), the dividend is a flat, never-raised $0.40, compensation contains no ROIC/ROE hurdle (revenue + non-GAAP operating income + cash flow + relative TSR), say-on-pay fell 99%→86%, insider ownership is just 1.3%, and insiders have made zero open-market purchases across 121 Form 4s while the CEO sold $15.8M within ~$15 of the all-time high.

Embedded expectations. At ~$348 the market is underwriting (a) data-center growth that stays well above the broader cycle and survives the 800V transition where AEIS isn’t NVIDIA-named; (b) gross margin reaching and holding 43%; and © a structural lift in ROIC into the high-teens that the historical record has not yet shown. If those hold, the multiple is defensible; if data center air-pockets or the semi cycle stalls, a ~10% GAAP-operating-margin cyclical at 40x has substantial downside. The body that follows takes no position; it lays out the evidence on both sides.

2. Business Overview

Advanced Energy was incorporated in 1981 and went public in 1995. It designs and manufactures precision power-conversion, measurement, and control solutions — equipment that takes raw electrical power and shapes it into the highly controlled forms that demanding industrial processes require. The company organizes its revenue into four end markets.

Semiconductor (FY2025: $839.9M, ~46.7% of revenue). This is the historical core and the highest-quality piece. AEIS makes RF (radio-frequency) and pulsed-DC plasma power generators, matching networks, and remote plasma sources that are integrated into etch and deposition chambers inside wafer-fab equipment. Plasma — an ionized gas — is the medium through which chips are etched and thin films are deposited; the power generator that strikes and controls that plasma is a mission-critical subsystem, because power stability and waveform precision directly determine etch uniformity, throughput, and yield. AEIS’s flagship leading-edge product lines are eVoS, eVerest, and NavX. Customers are the major wafer-fab-equipment OEMs (Applied Materials, Lam Research, Tokyo Electron, KLA) and, increasingly, the fabs themselves. AEIS also sells “system power” — the power that runs the machine itself (between the wall and the tool) — into semiconductor test and wafer-fab equipment, a newer adjacency.

Data Center Computing (FY2025: $587.3M, +106.7%, ~32.7% of revenue). This is the legacy of the 2019 Artesyn Embedded Power acquisition and the engine of the recent re-rating. AEIS makes merchant and embedded server and rack power supplies and OCP-compliant power shelves (CRPS units up to ~2.4kW, ORv3 power shelves up to 100kW) that convert facility power into the forms AI servers consume. As AI racks have grown more power-hungry, AEIS’s content per rack has risen, and the segment grew from a minor line to the swing factor for the whole company. In June 2026 it launched the ADH-series 800VDC→50VDC converters (98.2% peak efficiency) and bus converters aimed at the next-generation 800V/HVDC “megawatt rack” architecture.

Industrial & Medical (FY2025: ~$282M, −10.7%, ~16% of revenue). High- and low-voltage power for medical devices (therapeutic, diagnostic, life-science), test and measurement, factory automation, robotics, and battery-backup. This market is the most fragmented and is where management targets bolt-on M&A. It spent 2023–2025 in a deep post-COVID inventory correction and was inflecting upward in Q1-26 (bookings +14% sequential, highest since 2023).

Telecom & Networking (FY2025: ~$90M, ~5% of revenue). Power for networking and telecom equipment; recently boosted by AI-related networking programs (Q1-26 +16% YoY, highest since 2023).

Business model and revenue character. AEIS sells hardware through a direct salesforce, sales reps, channel partners, and distributors, plus a smaller after-market services stream (calibration, conversions, upgrades, refurbishment, repair). Revenue is largely transactional product revenue — there is no large recurring/subscription base — though plasma-power positions, once designed into an OEM tool and qualified into a fab recipe, can recur for years across the life of that platform. The economics are cyclical (tied to WFE and now AI data-center capex) and concentrated: the top three customers were 23%, 19%, and 12% of FY2025 revenue (54% combined), up from 37% the prior year.

Verdict. AEIS is a diversified, mission-critical power-conversion supplier with a high-quality plasma-power core and a fast-growing but lower-quality data-center box business. The revenue is transactional and cyclical with sticky design-in positions in the semiconductor segment. It is a real franchise — but a supplier’s franchise, not a platform’s.

3. Industry Dynamics

AEIS sits in two very different industries, and conflating them is the most common analytical error on this name.

Semiconductor process/plasma power — a good-but-cyclical, fragmented industry. The RF-plasma-generator market is roughly $2.0–2.7B in 2025, growing ~6–9% to ~$2.7B by the early-to-mid 2030s (third-party estimates). AEIS’s broader served power TAM (adding high-voltage, system power, and sensing/controls) is larger. The defining features:

• WFE cyclicality. Demand tracks wafer-fab-equipment spending, which is cyclical. AEIS semiconductor revenue ran $743.8M (2023) → $792.5M (2024) → $839.9M (2025) through a WFE downturn and recovery, and was “flattish YoY” in Q1-26.
• A genuine secular tailwind: rising plasma intensity per chamber. Advanced nodes — gate-all-around, high-NA EUV patterning, 3D-NAND layer stacking — require more plasma process steps and more complex, multi-level pulsed RF power per chamber. This is the structural reason plasma-power TAM can grow faster than WFE units over time. (That intensity rises is FACT; that AEIS captures a disproportionate share of it is the thesis to be proven.)
• Fragmented competition, powerful buyers. The 10-K names COMET, Daihen, MKS Instruments, and TRUMPF Hüttinger (Kyosan is another industry player) and states plainly that “no single company dominates any of our markets.” The customers are a handful of enormous WFE OEMs who dual-source and dictate terms. This is a structurally good-not-great industry: high barriers (qualification, IP, copy-exact) but fragmented supply and concentrated, powerful buyers — Marathon’s capital-cycle lens would note that high returns here are competed away by capable rivals (MKS, COMET, Hüttinger).

Data-center/AI rack power — a fast-growing but structurally commoditized industry. The AI build-out is expanding power content per rack dramatically and driving an architecture transition (54V → 48V → 800V/HVDC megawatt racks). That is a real, large TAM. But the merchant power-supply layer is a crowded, low-barrier assembly business: the 10-K names Delta Electronics, Flex, Lite-On, and Cosel — Delta and Lite-On are vast, low-cost Asian power-supply manufacturers — and at the component level the differentiated IP and margin concentrate in the silicon (Infineon, Monolithic Power, Navitas, Texas Instruments, Power Integrations, onsemi), not in the box integrator. The clearest structural tell: the named silicon and infrastructure partners for NVIDIA’s 800V Vera-Rubin/Kyber architecture are ADI, AOS, EPC, Infineon, Innoscience, MPS, Navitas, onsemi, Power Integrations, Renesas, Richtek, ROHM, STMicro, TI (silicon) and ABB, Eaton, Vertiv (infrastructure) — AEIS, like Vicor, is not on the list. AEIS competes at the rack/shelf integration layer on density, efficiency, and speed — advantages that are real today but transient in a fast-moving transition.

Verdict: a good-but-cyclical-and-fragmented semiconductor power industry bolted to a hyper-growth-but-commoditized data-center power industry. The blended industry is more attractive than a year ago because of AI capex, but neither half is a structurally protected profit pool. The plasma-power industry has barriers but capable competitors and powerful buyers; the rack-power industry has scale and growth but low barriers and value capture migrating to silicon. Structurally mixed — good enough to earn a return on capital, not good enough to compound one durably above peers.

4. Competitive Position

Plasma power: a real but narrow, product-line-level switching-cost moat — and #2-grade, not company-wide. Plasma generators are designed-in to the OEM’s tool and qualified into the fab’s process recipe (“copy-exact”); re-qualifying an alternative is slow, costly, and risky, so an incumbent socket is sticky and can recur across a platform’s life. This is a genuine Greenwald demand/captivity (switching-cost) advantage — the same mechanism that protects MKS Instruments. But it must be pressure-tested:

• It is contested, not owned. AEIS itself states it has “leading market share” in conductor etch and deposition but “lower market share” in dielectric etch — a precise admission that it is #1 in some niches and #2/#3 in others. Third-party data puts AEIS ~13.2% and MKS ~11.8% of the narrow RF-generator market (AEIS nominally ahead), but the broader process-power franchise leadership belongs to MKS, which bundles generators + matching networks + reactive-gas/remote-plasma + Baratron/MFC metrology (“surround-the-wafer”). The honest read: AEIS and MKS are co-leaders/a duopoly in RF generators specifically; MKS is the broader process-power leader. State both; don’t crown a single winner.
• Switching costs cut both ways and cap pricing power. Copy-exact locks AEIS in once it wins — but re-opening qualification is the customer’s lever, the OEMs dual-source, and AEIS is contesting (not harvesting) many sockets. The financial proof is decisive: gross margin has been stuck at 35.7–40.1% across the cycle, versus MKS ~47%, Vicor ~52%, MPWR ~55%. A real, company-wide pricing-power moat shows up as a higher, rising gross margin. AEIS’s does not (yet).

Data-center rack power: no durable moat. This is a merchant box-assembly business winning today on power density, 98%+ efficiency, and time-to-market against Delta/Lite-On/Flex — a temporary execution edge in a fast-moving architecture transition, which Greenwald would classify as the most transient kind of advantage. The vulnerabilities are structural: (i) absence from NVIDIA’s named 800V partner list; (ii) hyperscalers increasingly design custom racks and could in-source or multi-source power shelves; (iii) the real IP and margin sit in the silicon AEIS integrates, not in AEIS; (iv) Delta Electronics is a vastly larger, lower-cost competitor. If rack power were a durable franchise, the data-center segment would carry a high, rising gross margin; the fact that consolidated GM only reached 40% with data center at record mix implies the segment is dilutive-to-neutral to corporate margin — consistent with a commodity-box read.

Moat-to-financial-outcome test. The discipline is: if a claimed moat can’t be tied to a financial outcome that would deteriorate without it, it isn’t a moat. AEIS’s plasma switching costs do tie to an outcome — recurring socket revenue and a gross margin that, while not elite, is durable through the cycle and rising. The data-center “moat” does not — its margin is undifferentiated and its position depends on staying ahead in a transition it doesn’t architect.

Verdict: a genuine but narrow, product-line-level switching-cost moat in plasma power (co-leader in RF generators, #2/#3 in adjacencies) and essentially no durable moat in data-center rack power. AEIS sits in the “squeezed middle” of the semicap value chain — essential, fragmented, powerful buyers on both sides — with returns (~8% ROIC on total capital) that confirm participation rights rather than pricing power.

5. Growth History and Forward Opportunities

History: flat-with-cyclicality, until the 2025–26 data-center surge. Revenue ran $1,416M (2020) → $1,456M (2021) → $1,845M (2022 peak) → $1,656M (2023) → $1,482M (2024 trough) → $1,799M (2025). For four years the top line essentially round-tripped — the company was a flat cyclical. GAAP diluted EPS over the same span: $3.49 / $3.51 / $5.29 / $3.39 / $1.43 (2024 trough) / $3.84. The inflection came in 2025–26, driven almost entirely by data-center computing: $249.9M (2023) → $284.2M (2024) → $587.3M (2025, +107%), with Q1-26 at $194M (+102% YoY). Semiconductor recovered modestly with the WFE cycle; industrial/medical shrank through a two-year inventory correction ($474M in 2023 → $282M in 2025) and is only now inflecting.

Organic vs. acquired. The recent surge is organic (the data-center growth is volume/content on the existing Artesyn product base), but the platform that enabled it was acquired in 2019. Recent M&A has been tiny (SL Power $145.6M in 2022; Airity $19.6M in 2024) and immaterial to growth.

Forward drivers, ranked by durability:

1. AI WFE intensity (medium-high durability). More plasma steps and RF content per advanced-node chamber is a structural tailwind; AEIS guides 2H-26 semi revenue up >30% YoY and expects “continued growth in 2027.” Timing is cyclical, but the direction is real. The eVoS/eVerest/NavX “outgrow-WFE via share gain” thesis is plausible but unproven this cycle — Q1-26 semi was only “flattish YoY,” i.e., tracking WFE, not yet dramatically beating it. New-product revenue becomes “meaningful” only late 2026 and into 2027–28.
2. AI rack power, incl. 800V/HVDC (medium-LOW durability). A large TAM, but the headline +102% rate is unsustainable (management itself guides FY26 data center to “mid-30s,” a sharp deceleration) and concentrated in a few hyperscalers. The 800V prize — where AEIS isn’t NVIDIA-named — produces only small revenue in 2026 and ramps in 2027–28. This is real but fragile and contested.
3. Industrial & Medical recovery + bolt-on M&A (medium). Off a depressed base, with a fragmented market that supports tuck-ins; this is a genuine second leg if the recovery sustains.
4. Gross-margin self-help to 43% (medium). Real, but partly cyclical operating leverage on the volume surge rather than purely structural mix.

Capacity is being built ahead of the demand. Management is scaling revenue-generating capacity to >$2.5B exiting 2026 (Malaysia, Philippines, Mexico) and >$3.5B once the new 500,000-sq-ft Thailand factory is fully built, pulling forward Thailand spending into late 2026. FY26 capex is guided to $170–180M (up from $107M in 2025 and ~$57M in 2024).

Verdict: medium-quality growth. The end-market tailwinds are real, but the headline growth is cyclical (WFE) plus a concentrated, contested AI-rack spike, with the durable share-gain thesis plausible-but-unproven. Growth that earns ~8% ROIC is, by this framework, only marginally value-creating; the open question is whether the new products and 800V transition durably lift returns or simply add lower-return revenue and a large new fixed-cost capacity base.

6. Financial Quality

Revenue and margins. FY2025 revenue $1.80B (+21%); Q1-26 $511M (+26% YoY). The margin trajectory is the genuinely encouraging part of the story:

The defining fact: returns barely clear the cost of capital through the cycle. On a total-invested-capital basis (equity + gross debt), ROIC was 3.8% in the 2024 trough and ~7.8% in good-year 2025 — below a ~9–10% WACC even in a recovery. Net of the company’s large cash balance, ROIC reaches the low-teens in up-years (~12.8% in 2025) but collapsed to ~6.2% in 2024. ROIC.ai reports return_on_inv_capital of 8.16% for 2025, consistent with the total-capital computation. This is the hard financial signature of a narrow moat: a business whose economics improve with scale and cycle but do not durably compound above its cost of capital.

Margins are improving — but partly cyclically. The +560bps YoY jump in Q1-26 operating margin is substantially operating leverage on a +26% revenue surge, not purely structural mix. A sobering historical comparison: pre-Artesyn FY2018 AEIS earned ~50.9% gross margin, ~24.5% operating margin, and ~26% ROIC — the data-center business diluted AEIS’s margin structure even as it added growth and scale. The 43% gross-margin goal, if reached and held through a downturn, would be the proof that the improvement is structural; that proof does not yet exist.

Quality-of-earnings flags.

• The GAAP-to-non-GAAP gap is large and dominated by permanent items. FY2025 GAAP operating income $168.0M (9.3%) → non-GAAP $284.0M (15.8%) — a $116M / 6.5-point gap. The bridge: SBC $55.7M (the largest single add-back), amortization of acquired intangibles $22.1M, restructuring/impairment $12.5M, facility/transition costs $19.9M, acquisition costs $5.8M. The headline “earnings power” partly rests on adding back stock comp.
• SBC has tripled and is rising: $15.7M (2021) → $19.8M → $31.0M → $45.9M → $55.7M (2025, ~3.1% of sales) — real, ongoing dilution, not a one-time item. The 19.1% Q1-26 non-GAAP operating margin adds it back.
• “Restructuring” is quasi-recurring, not one-time: $25.1M (2023), $28.1M (2024), $6.3M (2025) — three consecutive plans for footprint consolidation, the 2025 plan running into 2027.
• A tailwind to flatter the GAAP trend: acquired-intangible amortization (~$22M/yr) declines as Artesyn/SL Power intangibles age out, mechanically lifting GAAP EPS in coming years.
• Cash conversion is decent but cyclical. FCF (OCF − capex): $111M (2021), $125M (2022), $148M (2023), $74M (2024 trough), $126M (2025). Q1-26 operating cash flow was actually a small outflow (−$6M) as inventory built $48M to support growth and buffer supply — a working-capital use that should reverse but bears watching. Capex is stepping up sharply ($107M in 2025 → $170–180M guided 2026), so near-term FCF is constrained by the capacity build; management guides FY26 FCF “at or above” 2025.

Balance sheet: a genuine quality positive. Net cash (~$224M at YE2025; ~$131M in Q1-26 pre the May raise), $791M cash, goodwill only $300.8M — this is not a levered roll-up. In May 2026 AEIS issued $1.0B of 0% convertible notes due 2031 and is redeeming the remaining $136.7M of its 2.5% 2028 converts, leaving it with substantial cash, zero cash interest on the new paper, and a $600M undrawn revolver. Liquidity and solvency are not risks.

Verdict. Economics do improve with scale and cycle — gross margin is genuinely climbing and the balance sheet is pristine — but the through-cycle return on capital is mediocre (~8% on total capital), the lowest-gross-margin profile in its peer group, and the non-GAAP “earnings power” leans on growing, permanent SBC. This is a financially decent cyclical, not a high-quality compounder.

7. Capital Allocation

M&A: a clean record built on one transformational deal. The 2019 Artesyn Embedded Power acquisition (~$400M) created the data-center, telecom, and embedded-power businesses now carrying the stock — a clear winner. Since then, capital deployment into M&A has been small and disciplined: SL Power Electronics ($145.6M, 2022, medical/industrial power; no impairment) and Airity Technologies ($19.6M, 2024, high-voltage; immaterial, though $11.1M of the consideration is retention SBC). NavX, eVoS, and eVerest are internally developed, not acquired. There have been no impairments anywhere in the five-year corpus — a genuinely clean record. The caveat: the data-center upside is a 2019-vintage legacy, not recent capital deployment, and management is now signaling renewed M&A appetite in fragmented industrial/medical.

Shareholder returns: pro-cyclical and thin. Buybacks by year: $11.6M (2020), $78.1M (2021, heaviest, stock ~$80–110), $26.6M (2022), $40.0M (2023), $1.8M (2024, halted as the stock began to run), $30.2M (2025, most before the parabolic move; Q4-25 only ~$7M at $205+). ~$167M of the $200M authorization sits unused at the richest-ever price — the right instinct on a buyback (don’t chase), but it means today’s shareholder gets neither buybacks nor a meaningful dividend. The dividend has been a flat $0.10/quarter ($0.40/yr), never raised across the entire period — a token ~10% payout, not a capital-return story.

The May-2026 convert is the single shrewdest move in the file. Swapping 2.5%-coupon paper (struck at ~$137) for $1.0B of 0% paper due 2031 at a much higher conversion price monetizes the inflated equity through cheap financing rather than buying stock at the top, terms out the maturity, and locks in zero cash interest. This is opportunistic and smart — though it does create some future dilution on the in-the-money spread of the converting 2028 notes.

Compensation — the structural weakness. The proxy contains no ROIC or ROE metric anywhere. The annual bonus pays on Revenue (40%) + non-GAAP Operating Income (40%) + Adjusted Cash Flow (20%) — achievement was 144.4% in FY2025 as the data-center surge blew past targets. Long-term PSUs vest 70% on relative TSR vs. the S&P 1000 + 30% strategic goals. So management is paid for growth and the stock price — the very multiple expansion that carried the shares — not for capital efficiency. For a business that lives or dies on ROIC, the absence of a return-on-capital hurdle is a real misalignment. Supporting signals: say-on-pay approval fell 99% → 86% (a meaningful investor rebuke that still passed), and directors-and-officers own just 1.3% of the company — low skin in the game.

Insider behavior: one-way selling. Across 121 Form 4s in 2024–2026 there were zero open-market purchases (code P). Transaction codes: 58 sells, 60 option/RSU exercises, 32 tax-withholdings, 0 buys. The CEO sold 50,000 shares for $15.8M in a single 10b5-1 block in March 2026 — within ~$15 of the all-time high; the CFO’s sales scaled up with the price ($125 → $316). All material sales were 10b5-1-planned, which softens the read, but the complete absence of any conviction buying while the stock tripled is the opposite of the board-buying-the-bottom signal one looks for.

Verdict: above-average, not elite. Clean, disciplined M&A and a genuinely shrewd convert refinancing, on a net-cash balance sheet — but pro-cyclical/idle buybacks, a token never-raised dividend, no ROIC hurdle in pay, falling say-on-pay, low ownership, and pure insider selling. Management allocates the business’s capital reasonably; it does not return capital opportunistically to today’s shareholder, and its incentives reward the momentum the stock has already enjoyed.

8. Changes and Headwinds — Last Two Years

Thesis-strengthening changes (validated by numbers, not just narrative):

• The data-center pivot is real and paying off. DCC went from a minor line to $587M (+107%) in FY2025 and kept doubling (Q1-26 +102%) — the swing factor in both revenue and the re-rating.
• Gross margin is structurally climbing: 35.7% → 37.7% → 40.1% (Q1-26), the highest since the 2019 Artesyn deal, with a stated path to ~41% by YE26 and a 43% long-term goal.
• Guidance raised twice: FY26 revenue growth lifted from high-teens to low-to-mid-20s, data center from “>30%” to mid-30s.
• Deleveraging/clean-up: the $1.0B 0% convert and redemption of the 2.5% notes term out debt at zero cash cost.
• Management stability: CEO Steve Kelley (since March 2021) and CFO Paul Oldham (since May 2018) intact; no succession overhang.

Live headwinds (all grounded in the Q1-26 call and filings):

• Data-center demand-mix volatility. Management flagged “frequent customer changes in demand mix” and guided Q2-26 data center to moderate sequentially — the growth is lumpy and order-timing-sensitive.
• Customer-side (downstream) constraints. Near-term data-center revenue is partly capped by bottlenecks at customers, so the top line is hostage to their build pace.
• Customer concentration. Top three customers were 54% of FY25 revenue (up from 37%); the diversifying “second-wave” data-center customers don’t contribute meaningfully until ~2027.
• The biggest HVDC prize is still a promise. 800V is late-2026 first production, 2027–28 ramp — the $400–450 street targets capitalize revenue not in the run-rate, on a part where AEIS isn’t a named NVIDIA partner.
• Tariffs / input-cost inflation. Management conceded “supply and cost challenges have begun to surface”; the mitigation (a +$48M inventory build, dual-sourcing) carries working-capital and margin risk.
• Capacity-build / over-build risk. Simultaneous Thailand + Malaysia/Philippines/Mexico ramps to >$3.5B capacity against a still-cyclical demand base — classic over-build exposure if data center or semi stalls.
• Cyclical semi exposure remains and the 2026 recovery is back-half-weighted (“2H likely up >30% YoY”) — not yet delivered.

Verdict. The last two years strengthen the business. The thesis tension is that the improvement is now a momentum re-rating that prices forward (2027–28) HVDC dollars and a not-yet-delivered semi recovery, against a demand stream management itself describes as volatile and customer-constrained. The changes are net-positive on fundamentals and net-cautionary on risk/reward at the current price.

9. Risk Analysis

The dominant risk is valuation, compounded by the cyclicality of both engines — a re-rated cyclical is doubly exposed if either the multiple or the cycle turns. There is no balance-sheet or going-concern risk; a “catastrophic loss” scenario is remote.

10. Valuation Discussion (Embedded Expectations)

No price target and no recommendation here — only what the current price implies and where the market may be right or wrong.

Where the multiple sits. At ~$348 (≈37.7M basic / ~40.6M+ diluted shares including converts), market cap is ~$13.1B (basic) to ~$14–15B (fully diluted). Net cash is modestly positive after the May-2026 $1.0B raise, so EV ≈ $13B. Against that:

• EV/sales: ~7.2x trailing (FY25 $1.8B), ~5.9x forward (FY26E ~$2.2B on low-to-mid-20s growth).
• EV/EBITDA: ~25–27x forward non-GAAP (Q1-26 adj. EBITDA $108M annualizes >$450M).
• P/E: ~39x FY26E non-GAAP EPS (~$9); ~73x GAAP TTM EPS ($4.75) — the GAAP figure is distorted by amortization and is not the right lens, but even the non-GAAP multiple is rich.
• P/B 10.6x; P/S 7.25x — 98.7th and 98.5th percentiles of AEIS’s own ten-year range; composite 96th. This is, on its own history, the most expensive AEIS has ever been. (The P/E percentile, 91st, is understated because GAAP EPS is depressed by amortization — P/B and P/S are the cleaner own-history tells here, and both scream richest-ever.)

What the price embeds. To justify ~39x non-GAAP / ~6x forward sales for a business with ~8% total-capital ROIC, the market must be underwriting a permanent step-change in both growth and returns:

1. Data-center growth stays well above the cycle and survives the 800V transition — i.e., AEIS holds or gains share into 2027–28 megawatt racks despite not being a named NVIDIA partner, and the segment compounds rather than air-pockets after the +100% base effect.
2. Gross margin reaches and holds ~43%, lifting operating margin durably toward the high-teens/low-20s — and holds through the next semi downturn, proving it structural rather than cyclical leverage.
3. ROIC steps up into the high-teens as the new capacity fills and high-value products mix up — a return profile AEIS has never sustained post-Artesyn.

Scenario sketch (illustrative, non-GAAP, not a target):

• Bear (~$140–180): semi recovery disappoints and/or data center air-pockets; revenue plateaus ~$1.9–2.0B, margins give back cyclical gains (op margin to low-teens), the multiple compresses toward a still-premium ~18–22x mid-cycle non-GAAP EPS (~$7–8). This is roughly a return to early-2026 levels — not a collapse, because the balance sheet and franchise are real.
• Base (~$210–260): FY26 plays out near guidance (~$2.2B revenue, GM ~41%, non-GAAP EPS ~$9), data center decelerates to a healthy clip, 800V ramps modestly in 2027; the market pays ~24–29x — a premium that credits the improvement without capitalizing perfection.
• Bull (~$400–470, where the sell-side sits): 800V wins land, semi outgrows WFE on eVoS/eVerest/NavX, GM hits 43%, ROIC structurally re-rates; the market sustains ~40x+ on >$11–12 of out-year non-GAAP EPS. This is the priced-in case — it requires nearly everything to go right.

Embedded-expectations read. The market is paying a quality-compounder multiple for a mid-cycle-return cyclical, capitalizing the most cyclical revenue surge in the company’s history and forward HVDC dollars that aren’t yet in the run-rate. AEIS may well grow into part of this — the franchise is real and improving — but the current price leaves little margin of safety and embeds a structural ROIC re-rating the company has not yet demonstrated. The valuation is the thesis risk.

11. Variant Perception

Consensus belief. AEIS is being recast from a sleepy cyclical semiconductor-power supplier into an AI-data-center power-infrastructure compounder — the 800V/HVDC transition layered on a 2026 semiconductor up-cycle and a march to 43% gross margin. The sell-side is uniformly constructive (Cantor Overweight $400, BofA Buy $450; ~11 analysts, average “Buy”). The tape agrees: +310% in fourteen months, relative strength rs_12m 161.65, near an all-time high.

Strongest bull case. A genuinely improving business at the intersection of two secular tailwinds (AI WFE intensity + AI rack power), with gross margin breaking 40% for the first time since 2019, a pristine net-cash balance sheet, a clean M&A record, a shrewd 0%-convert that monetized peak equity, and design-win momentum (eVoS/eVerest/NavX, 800V) that should drive “more meaningful” revenue in 2027–28. If the margin and share-gain stories prove structural, ROIC re-rates and the multiple is defensible.

Strongest bear case. A mediocre-return cyclical (~8% total-capital ROIC, lowest gross margin in its peer set) re-rated to its richest-ever multiple on a non-repeatable +100% data-center base effect and forward HVDC promises — for a part where it isn’t a named NVIDIA partner and the IP belongs to the silicon vendors. Customer concentration is 54%, the data-center stream is “volatile” and customer-constrained by management’s own words, capacity is being built ahead of a cyclical demand base, compensation has no ROIC hurdle, and insiders are pure sellers (CEO selling within $15 of the high). Beta ~2.0 means a multiple de-rate and a cycle turn would compound brutally.

The 3–5 assumptions that matter most:

1. Is the 40%→43% gross-margin gain structural or cyclical? (Structural = bull; cyclical leverage = bear.)
2. Does data center compound past the base effect, and does AEIS hold share into 800V despite not being NVIDIA-named?
3. Does the eVoS/eVerest/NavX share-gain thesis convert to AEIS outgrowing WFE — or is semi just tracking the cycle?
4. Does ROIC structurally re-rate into the high-teens, or revert toward ~8% as new capacity dilutes returns?
5. Does the multiple hold, or mean-revert from the 96th percentile of its own history?

Factor-positioning read (FactorsToday + AZI). AEIS is a high-beta (market beta ~1.45 in the sector model, raw ~2.0), modest-momentum, strongly anti-low-volatility name (LowVolatility loading −0.72) with positive Quality (+0.27) and negligible Value (−0.02). Risk-adjusted track record is extraordinary but unsustainable-looking: y1 Sharpe 3.40 (return +182% annualized), m6 Sharpe 3.23 — the signature of a name in a one-way momentum run, against a lifetime max drawdown of −78.6% (and −39.9% even over the last five years). Factor-similar peers (UCTT, TER, DIOD, FORM, ONTO) are all cyclical semicap names, confirming the market still treats AEIS as a high-beta semicap cyclical even as the narrative sells it as an AI compounder. This is evidence that consensus is positioned in a crowded momentum trade, not a defensible low-vol quality holding — the positioning most exposed to a growth-or-multiple disappointment. Treat as positioning input, not a price call.

12. Fact vs. Interpretation

13. Open Questions

1. Segment gross margins. AEIS doesn’t disclose data-center vs. semiconductor gross margin. Is data center dilutive (the commodity-box read) or accretive? This is the single most important undisclosed number for the quality verdict.
2. 800V share. What is AEIS’s actual design-win position in 2027–28 megawatt racks, given its absence from NVIDIA’s partner list? Are the “second-wave” wins with reference designs or bespoke programs?
3. Structural vs. cyclical margin. How much of the 43% goal survives a WFE downturn? What is the decremental margin on the new capacity if demand stalls?
4. eVoS/eVerest/NavX revenue. Management won’t quantify new-product revenue; when does it become “meaningful,” and does it actually let semi outgrow WFE?
5. Customer identity and concentration. Which hyperscaler is the 19% (or 12%) customer, and how exposed is AEIS to that single roadmap?
6. Capacity utilization. With >$3.5B capacity targeted against ~$1.9B revenue, what utilization underpins the 43% margin goal, and what happens to returns if the cycle disappoints?

14. What Must Be True

Bull case — what must be true:

• Data center compounds past the +100% base effect (mid-30s in 2026, then durable growth), and AEIS holds or gains share into the 800V transition despite not being NVIDIA-named.
• Gross margin reaches ~43% and holds through the next semiconductor downturn — proving it structural.
• eVoS/eVerest/NavX convert to AEIS outgrowing WFE, and ROIC steps up into the high-teens as new capacity fills.
• Falsification test: a single data-center quarter that air-pockets (segment down YoY) or a semi 2H-26 that fails to deliver the guided >30% growth or gross margin that stalls below 41% — any of these breaks the “structural step-change” thesis the multiple requires.

Bear case — what must be true:

• The +100% data-center surge proves a cyclical/concentration spike, not a durable compound; share leaks at 800V to NVIDIA-named silicon vendors and lower-cost box makers.
• Gross-margin gains reverse with volume; ROIC reverts toward ~8% as the large new capacity base dilutes returns.
• The multiple mean-reverts from the 96th percentile as growth normalizes; beta ~2.0 amplifies the de-rate.
• Falsification test: two-plus consecutive quarters of data-center growth re-accelerating with rising segment margin, plus a gross margin that holds ≥42% into a flat-to-down WFE quarter, would prove the franchise has structurally re-rated and the bear is wrong.

15. Source Appendix

See the separate Source Appendix (Appendix B in the combined report) for the full list of primary and secondary sources, with URLs and access dates.

Primary sources: AEIS FY2021–FY2025 Forms 10-K, 10-Q, 8-K, and DEF 14A (SEC EDGAR, CIK 0000927003); AEIS Q1-2026 earnings call transcript (2026-05-04); AEIS Q4/FY2025 results (2026-02-10). Quantitative data: ROIC.ai (financial statements, ratios, enterprise value); AZI/azitrading.com (valuation percentiles, price history); FactorsToday (factor loadings, leaderboard). Secondary: third-party RF-plasma market sizing; NVIDIA 800V partner disclosures (Data Center Frontier); AEIS product launches (Power Electronics News); sell-side notes (Cantor, BofA) reported as sentiment signal only.

APPENDIX A — Standard Diligence Questionnaire

Standard Diligence Questionnaire — Advanced Energy Industries, Inc. (NASDAQ: AEIS)

Supplemental appendix, as of 2026-06-27. Fact / Interpretation / Assumption labeled where it matters.

General

What thoughtful questions have other investors asked about this company? The recurring institutional questions: (1) Is the data-center surge durable AI capex or a concentration/cyclical spike? (2) Does AEIS hold share into the 800V/HVDC transition given it isn’t a named NVIDIA partner? (3) Is the 40%→43% gross-margin path structural or cyclical operating leverage? (4) How much of “non-GAAP earnings power” is just adding back growing SBC? (5) Why no ROIC hurdle in compensation, and why are insiders only selling? (6) Can a ~8%-ROIC cyclical sustain a near-40x non-GAAP multiple?

Cyclicality & Earnings Nature

Are earnings at a cyclical high or low? Interpretation: closer to a high than a low. Semiconductor is recovering off a WFE downturn; data center is at a record run-rate (+102% YoY) that management itself expects to decelerate to “mid-30s”; gross margin is at a post-2019 peak (40.1%). Industrial & medical is the one segment off a trough. Net: earnings are being driven by a cyclical/secular up-leg, not depressed.

Driven by external environment or internal actions? Both. External: AI capex (data center), the WFE up-cycle (semi), the I&M inventory-correction recovery. Internal: genuine gross-margin self-help (footprint consolidation, new-product mix) and design wins. The recent surge is more external (cycle/AI) than internal.

How stable are revenues? Fact: not very. Revenue round-tripped $1.46B→$1.85B→$1.48B→$1.80B over 2021–2025; the data-center line is “volatile” and customer-constrained by management’s own description. This is a cyclical, transactional revenue base with sticky design-in positions, not a recurring/subscription model.

Outlook for products/services; how big will this market be? Fact: AEIS guides FY26 revenue growth to low-to-mid-20s, data center to mid-30s, semi 2H up >30% YoY. The plasma-power TAM (~$2.0–2.7B, mid-single-digit growth) expands faster than WFE on rising per-chamber intensity; the AI rack-power TAM is large and growing but commoditized. Markets are growing, international (manufacturing in Malaysia/Philippines/Mexico/Thailand; customers global).

Business Quality & Competitive Moat

Is the industry getting more or less competitive? Interpretation: stable-to-more in plasma (capable rivals MKS, COMET, Hüttinger, Daihen; powerful WFE buyers); structurally crowded in data-center boxes (Delta, Lite-On, Flex), with value capture migrating to silicon.

How profitable is the business (ROIC, ROE)? Fact: ROE 13.8% (FY25 good year), 5.4% (FY24 trough); ROIC ~8% on total capital, ~12.8% net of cash in 2025. Mediocre — at/near WACC through the cycle.

How profitable is the industry; how many competitors; barriers to entry? Fact: the 10-K states “no single company dominates any of our markets.” Plasma power has real barriers (qualification, IP, copy-exact) but several capable competitors; rack power has low barriers. Industry profitability is moderate and competed.

Can the business be easily understood? Yes — it sells power-conversion hardware into identifiable end markets; the main complexity is the GAAP-to-non-GAAP bridge and segment-margin opacity.

Can it be undermined by foreign low-cost labor? Interpretation: in data-center boxes, partly — it competes directly with low-cost Asian makers (Delta, Lite-On). In plasma power, less so — the moat is IP/qualification, not labor cost. AEIS already manufactures in low-cost Asian locations.

Do brands matter? Nature of competition? Switching costs? Brands matter modestly (reputation/reliability in mission-critical power). Competition is on technology, qualification, density/efficiency, and price. Switching costs are real in plasma (re-qualification risk) but a customer lever; low in rack power.

Financial Condition & Balance Sheet

Assets not fully recognized on the balance sheet? Internally developed product IP (eVoS/eVerest/NavX) and qualified design-in positions are not capitalized — genuine unrecognized value, partly offsetting the lowish ROIC.

Off-balance-sheet liabilities? Operating leases (modest); no material pension/litigation overhang flagged. The $1.0B 0% convert and converting 2028 notes create future dilution (on-balance-sheet but equity-dilutive).

How conservative is the accounting? Interpretation: reasonably conservative on the balance sheet (net cash, modest goodwill $300.8M, no impairments), but the non-GAAP presentation is aggressive in adding back growing, permanent SBC ($55.7M, ~3.1% of sales) and quasi-recurring restructuring. Read GAAP and non-GAAP together.

How CapEx-hungry is the business? Historically modest (~$57M), now stepping up sharply ($107M in 2025 → $170–180M guided 2026) for the Thailand/Asia capacity build — a growth-capex bet that compresses near-term FCF.

Capital Allocation & Management

How much FCF does it generate; how is it used; philosophy? FCF ~$74–148M/yr (cyclical), constrained near-term by capacity capex. Uses: token dividend ($15M/yr), pro-cyclical buybacks (idle now), bolt-on M&A, and growth capex. Philosophy: invest in capacity/products first, return capital opportunistically (well, intermittently) second.

Significant acquisitions recently? Artesyn (2019, transformational, ~$400M, the data-center maker); SL Power ($145.6M, 2022); Airity ($19.6M, 2024). No impairments. Renewed appetite signaled for industrial/medical.

Buying back shares? Yes but pro-cyclically — heaviest at $80–120 (2021–23), ~$167M authorization idle at $300+. Fact.

Issuing large amounts of new shares to insiders? SBC is rising ($55.7M) and dilutive but not egregious (~3% of sales); share count is stable ~37.7M. Fact.

Compensation policy of directors/management? Red flag: no ROIC/ROE metric; bonus on revenue/non-GAAP op-income/cash-flow, LTI 70% relative TSR. Pays for growth and stock price, not capital efficiency. Say-on-pay fell 99%→86%.

Motivations of management? Interpretation: aligned with growth and the share price (rTSR), and monetizing via 10b5-1 sales (zero buys; CEO sold near the high). Insider ownership only 1.3% — modest skin in the game.

Valuation & Market Data

ADR, MLP, or K-1 issuer? No — ordinary U.S. common stock, NASDAQ-listed, 1099 reporting.

Dividend policy? Flat $0.10/quarter ($0.40/yr), never raised; ~10% payout; ~0.1% yield. Token.

How profitable is the business? Covered above — mediocre returns (ROIC ~8%), improving margins (GM 40.1%).

Is net income diverging from cash from operations? Fact: Q1-26 OCF was a small outflow (−$6M) on a $48M inventory build despite strong net income — a working-capital divergence that should reverse but bears watching. Over full years, OCF tracks earnings reasonably (cash-flow-to-net-income ~1.3–1.6x), aided by D&A and SBC add-backs.

Risks & Downside

What factors would cause the stock to decline? A valuation de-rate from the 96th percentile; a data-center air-pocket or mix reset; a semi-cycle stall; gross-margin disappointment; customer concentration shock; over-build of capacity into a soft cycle. With beta ~2.0, a multiple de-rate and a cycle turn compound.

Risk of a catastrophic loss? Low — net cash, no covenant stress, diversified end markets, no going-concern risk.

Chance of a total loss? Negligible. The risk here is valuation/drawdown (the stock has a −78.6% lifetime max drawdown and −39.9% even over five years), not insolvency.

Recent News & Events

Has the business environment changed recently? Yes, materially and favorably on fundamentals: data-center inflection (+107% FY25), gross margin breaking 40%, guidance raised twice, the AI-power re-rating. The risk environment also changed — tariffs/input costs surfaced, and the valuation became the dominant risk.

Significant acquisitions / accounting changes / new markets/facilities/management? Airity (2024, tiny); the $1.0B 0% 2031 convert and redemption of 2.5% 2028 notes (May–Jun 2026); new Thailand 500k-sq-ft factory (broke ground 2023, ramping); Malaysia/Philippines/Mexico expansions; 800V/HVDC product launches (2025–26). Management (Kelley/Oldham) unchanged. No accounting-policy red flags.

APPENDIX B — Source Appendix

Source Appendix — Advanced Energy Industries, Inc. (NASDAQ: AEIS)

Compiled 2026-06-27. Primary sources before secondary.

Primary — SEC filings (EDGAR, CIK 0000927003)

• AEIS Form 10-K for FY2025 (filed 2026-02-13) — https://www.sec.gov/Archives/edgar/data/927003/000110465926014731/aeis-20251231x10k.htm — segment revenue, customer concentration (top 3 = 23/19/12%), competition (“no single company dominates”), GAAP→non-GAAP reconciliation, SBC, restructuring, goodwill/intangibles, buyback authorization, convert terms.
• AEIS Form 10-K for FY2024 (filed 2025-02-18); FY2023 (2024-02-20); FY2022 (2023-02-17); FY2021 (2022-03-16) — multi-year revenue, margins, M&A notes (SL Power $145.6M 2022; Airity $19.6M 2024), capex, dividend history.
• AEIS DEF 14A proxy (filed 2026-03-26) — incentive metrics (revenue/non-GAAP op-income/cash-flow STI; 70% relative-TSR LTI; no ROIC), say-on-pay (~86%), insider ownership (1.3%).
• AEIS Forms 8-K (2024–2026) — quarterly earnings; FY2025 results (2026-02-10); Q1-26 results (2026-05-04); $600M revolver/term loan (2025-05-08); $1.0B 0% convertible notes due 2031 (2026-05-13/14); redemption notice for remaining $136.7M of 2.5% 2028 notes (2026-06-12); say-on-pay result (2025-04-25).
• AEIS Forms 3/4/5 (2024–2026, 121 filings) — insider transactions: zero open-market buys; 58 sells (CEO Kelley 50,000 sh / $15.8M 10b5-1 near ATH; CFO Oldham scaling sales).

Primary — Transcripts & releases

• AEIS Q1-2026 earnings call transcript (2026-05-04), via ROIC.ai — segment detail (Semi $219M, DCC $194M +102%, I&M $72M, T&N $25M), GM 40.1%, non-GAAP op margin 19.1%, EPS $2.09, FY26 guide raised to low-to-mid-20s, 43% GM goal, 800V solutions, capacity (>$2.5B exiting '26, >$3.5B with Thailand), capex $170–180M.
• AEIS Q4/FY2025 results (2026-02-10) — revenue $1.80B (+21%), DCC $587.3M (+107%), non-GAAP EPS $6.41 (+73%), record OCF $235M.

Quantitative data sources

• ROIC.ai MCP (accessed 2026-06-27) — income statement, balance sheet, cash flow, profitability/credit ratios, enterprise value, per-share data, valuation multiples; earnings-call transcripts. Third-party aggregated; reconciled to filings.
• AZI / azitrading.com (accessed 2026-06-27) — valuation_index own-history percentiles (composite 96.2; P/E 91.4th, P/B 98.7th, P/S 98.5th); 5-year daily price/OHLCV CSV; news feed (Cantor OW id 401551; BofA Buy $450 id 407157; convert redemption id 401988).
• FactorsToday (accessed 2026-06-27) — /stock-loadings (market beta ~1.45 sector model; LowVol −0.72; Momentum +0.36; Quality +0.27; Value −0.02), /leaderboard (y1 return +182% ann., Sharpe 3.40; lifetime max drawdown −78.6%), /stock-info (beta 2.01, rs_12m 161.65), /related-stocks (UCTT, TER, DIOD, FORM, ONTO).

Secondary — industry, competitive, market

• RF plasma generator market sizing (~$2.0–2.7B 2025; ~6–9% CAGR) — Mordor Intelligence; OpenPR/market research (Jun 2026), incl. narrow-RF share estimates (AEIS ~13.2%, MKS ~11.8%).
• NVIDIA 800V/HVDC partner list (AEIS absent) — Data Center Frontier / DCD (2025–26).
• AEIS ADH-series 800V product launch — Power Electronics News; StockTitan (Jun 2026).
• AEIS Q1-2026 deep dive — Alphastreet (Apr 2026).
• Thailand factory groundbreaking — BusinessWire / AEIS IR (2023-10-03).
• Sell-side notes (reported as sentiment signal only, not adopted): Cantor Fitzgerald initiation Overweight $400 (2026-06-11); BofA Securities Buy, PT raised to $450 (2026-06-23) — via Benzinga/AZI.