Jabil Inc. (NYSE: JBL) — A 5%-Margin Assembler Wearing an AI Architect’s Multiple

Target: Jabil Inc. (NYSE: JBL) · Electronics Manufacturing Services (EMS) · FY ends Aug 31 Report date: 2026-06-19 · Price (2026-06-18 close): $371.88 · Market cap: ~$28.4B · EV: ~$31B Independent equity research

⚡ Claude’s Take

This is the author’s own independent opinion and general information, not investment advice. The analysis that follows takes no position and carries no price target — it discusses valuation only as embedded expectations and scenarios.

Verdict: HOLD / AVOID-here for new capital · not-a-short · accumulate-on-weakness toward ~$200–240 (~16–18x FY27 core EPS / ~0.6–0.7x EV-sales — i.e., back toward where its closest factor twin Flex trades). Medium conviction.

Jabil is the best-run, most-diversified, most-shareholder-friendly assembler in the EMS comp set — and that is precisely the problem with owning it here. The market has re-rated a structurally thin-margin contract manufacturer to the single richest valuation in its own decade-plus history (composite valuation 97th percentile; price-to-sales 99.6th percentile) on a clean, real, but ultimately commoditizable AI-infrastructure story. The tell that the body builds out below: Jabil’s Intelligent Infrastructure (AI/cloud/datacenter) segment earns the same ~5.4% segment margin as its automotive and consumer segments. Jabil is capturing the AI build-out’s volume, not its rent — it earns the assembly spread, while the actual economics accrue to Nvidia (60%+ operating margin), the network-silicon owners, and the hyperscalers who design their own racks and can in-source or multi-source at will. Pay 0.95x EV/sales and ~24x forward core EPS for that, at the 99.6th percentile of the stock’s own price-to-sales range, and you are underwriting a triple-compound bet — peak-cycle hyperscaler capex keeps growing 30–50%/yr, core margin breaks above 6%, and the buyback keeps shrinking the float — all priced as if it is already certain. The June 17 Q3 print was a genuine beat-and-raise (FY26 core EPS guided to ~$12.70, AI revenue to ~$13.6B), and the stock spiked to $428.93 intraday… then closed down at $374.98. After a near-doubling in seven months, even good news now sells. That is what the late innings of a momentum trade look like.

The framing, grounded in the factor read, is a crowded, high-beta (1.54), peak-cycle momentum trade that has already run (relative strength 89th percentile, −3.6% off the all-time high, idiosyncratic vol ~31%, lifetime max-drawdown −90%) — not a falling knife and not a value compounder. I’d genuinely respect a long entry near Flex-like multiples (~$200–240) where you get the real AI-volume franchise, ~20% ROIC, and the best customer diversification in the group without paying for the AI option. Here, you are paying full architect prices for the assembler. Conviction: medium. The single fact that flips me bullish: durable core operating margin sustained above ~6.5% through a capex deceleration, proving Jabil is capturing real design content (rent), not just volume (spread). The single fact that flips me bearish: a hyperscaler-capex air-pocket or customer in-sourcing event that resets AI growth below ~20% on a fixed-cost base — the EMS down-cycle that the −90% lifetime drawdown remembers and this multiple has forgotten. Tag: paid for the AI rack, earning the assembly spread.

📈 Stock Price Action — Five-Year Event Map

Factual price history (price moves = Fact; attributed drivers = Interpretation). No recommendation, no price target.

The arc. Jabil has been one of the great quiet melt-ups of the cycle. From a post-COVID base of roughly $41–50 (2021–mid-2022) the stock climbed steadily, dipped to a 52-week low of $192.49 (Nov 20, 2025), then nearly doubled in seven months to an all-time-high close of $385.63 (June 15, 2026) — with an intraday spike to $428.93 on June 17 — before settling at $371.88 (June 18). That is roughly 7–8x off the 2022 base, ~2x off the November-2025 low, and just −3.6% off the all-time high. The entire late-stage move is an AI-infrastructure re-rating layered on a 27% reduction in the share count.

Cycle narrative. (1–2) The 2022–2023 leg was a fundamental margin/FCF re-rating off depressed levels as Jabil’s diversification and buyback machine got recognized. (3) The Mobility/BYD divestiture (Sep 2023 announcement, Dec 2023 close) is the strategic pivot of the period — Jabil shed ~$6B of concentrated, low-margin, Apple-adjacent mobility revenue near a peak. (4) That sale mechanically reset the revenue base ($34.7B FY23 → $28.9B FY24) and, combined with soft 5G/EV/renewables end-markets, drove a ~25% drawdown in 1H-2024. (5) The October-2024 CEO transition (Mark Mondello → Mike Dastoor) and the FY2025 three-segment reorganization re-told the story around “Intelligent Infrastructure,” coinciding with the broad AI-capex narrative; a December-2024 Amazon stock-purchase warrant signaled deepening hyperscaler alignment. (6) A 2025 consolidation reflected recurring market debates over whether hyperscaler capex was durable. (7) From November 2025, the AI-revenue inflection became undeniable — management raised AI-related revenue from ~$9B (FY25) toward ~$13.6B (FY26) and flagged a similar growth rate for FY27 — and the stock doubled. (8) The June 17, 2026 Q3 print confirmed the story (revenue +12% YoY, core EPS +24%, guide raised) yet the stock reversed hard intraday — the classic signature of a richly-priced momentum name where good news is already in the price.

1. Executive Summary

Jabil is one of the three or four largest Electronics Manufacturing Services (EMS) providers in the world: a global, asset-light contract manufacturer that builds electronic and mechanical products to its customers’ specifications across roughly 100 sites in ~30 countries with ~135,000 employees. FY2025 revenue was $29.8B at a ~8.9% gross margin and ~5.4% segment operating margin — the thin, uniform conversion economics that define the industry. Effective September 1, 2024, Jabil recast into three segments: Regulated Industries (auto & transport, healthcare & packaging, renewables; $11.9B, 40%), Intelligent Infrastructure (AI/cloud/datacenter, capital equipment incl. semicap test, networking/optics; $12.3B, 41% — now the largest and fastest-growing segment), and Connected Living & Digital Commerce ($5.6B, 19%).

The investment debate is entirely about valuation versus business quality. On business quality, the honest verdict is mixed-positive: Jabil has no durable competitive moat in the Greenwald sense — EMS is a structurally bad, fragmented, customer-power industry where Foxconn out-scales Jabil ~7x — but Jabil is an unusually well-run operator with the cleanest customer diversification in the comp set (top-5 customers ~36%, a single 16% customer), a genuinely high ~20–21% ROIC (an asset-light artifact of tiny fixed-asset and working-capital denominators), and a shrewd capital-allocation record anchored by the well-timed Mobility/BYD sale and a buyback that cut diluted shares 27% in four years.

On valuation, the picture is stark. At ~$372, Jabil trades at the richest multiple in its own history — composite valuation 97th percentile, price-to-sales 99.6th percentile, GAAP P/E 50x (93rd percentile). On a forward core (adjusted) basis the multiple looks more digestible — ~29x FY26 core EPS (~$12.70) and ~24x a plausible FY27 — but that is still a wholesale re-rating of a 5–6%-margin assembler that historically traded at 8–12x core earnings. The re-rating rests on the AI/Intelligent Infrastructure ramp: AI-related revenue grew from ~$9B (FY25) to a guided ~$13.6B (FY26, +50%) and is expected to grow at a “similar percentage rate” in FY27 (implying ~$20B). The skeptic’s central observation, developed throughout this memo, is that Jabil captures AI volume at the same ~5.4% margin it earns everywhere else — it is collecting the assembly spread, not the AI rent — yet it is priced as if it owns the rent. The recent near-doubling and the June-17 “sell-the-news” reversal suggest the easy money has been made. No recommendation or price target appears below; the body discusses valuation strictly as embedded expectations and scenarios.

2. Business Overview

What Jabil does. Jabil describes itself as “one of the leading providers of worldwide manufacturing services and solutions” (FY2025 10-K, Item 1). In plain terms it is a turnkey EMS / ODM-adjacent contract manufacturer: customers (OEMs, hyperscalers, brand owners) design a product or system; Jabil procures the components, builds, tests, integrates, and often manages the supply chain, logistics, and aftermarket service. Substantially all revenue comes from “manufacturing services” — tangible products built to customer specification — with materials largely a pass-through cost (FY2025 10-K MD&A). This is the structural reason gross margin sits at ~8.9%: a large fraction of “revenue” is pass-through componentry on which Jabil earns a thin conversion fee.

Segment structure (recast FY2025). Effective 9/1/2024 Jabil reorganized “to focus on speed, precision, and solutions” into three reportable segments (FY2025 10-K, Note 14):

The single most important structural fact in this table: segment margins are essentially flat at ~5.4% across all three segments, regardless of end-market. Whether Jabil is building an AI server rack, a continuous glucose monitor, or a consumer device, it converts revenue to operating income at roughly the same thin rate. That uniformity is the financial signature of an absence of pricing power (developed in).

• Regulated Industries — automotive & transportation, healthcare & packaging, renewables & energy infrastructure. The slowest-growing but, in healthcare, the highest-quality (FDA-qualified, long product cycles, sticky). Includes the Pii pharmaceutical CDMO acquisition.
• Intelligent Infrastructure — “the modern digital ecosystem including AI infrastructure”: capital equipment (semicap test — a high performer), cloud & data center infrastructure (server/storage/rack integration), and networking & communications (InfiniBand/Ethernet switching, silicon photonics/optics). Now the #1 segment and the entire growth story. Houses the 16%-of-revenue anchor customer (“Customer A”) and the second and third hyperscaler programs.
• Connected Living & Digital Commerce — connected devices, warehouse automation/robotics, retail, digital commerce. Mixed consumer environment; digital commerce is one of the higher-margin end markets.

Customer base & concentration. FY2025: the five largest customers were ~36% of net revenue; 87 customers accounted for ~90%; one customer (“Customer A,” primarily Intelligent Infrastructure) was 16% (FY2025 10-K). This is materially less concentrated than peers — Celestica’s top-3 are ~58%, Fabrinet’s Nvidia+Cisco ~46%, TTM’s two largest ~23%. Jabil historically carried heavy mobility/Apple concentration; the BYD divestiture and the AI ramp diluted it. The watch-item is re-concentration toward the new 16% AI anchor.

Business model mechanics. Jabil runs “customer-centric business units,” most serving a single customer with dedicated equipment, engineers, planners, and buyers (FY2025 10-K). Revenue is program-based and non-recurring in the SaaS sense — performance obligations satisfied over time, programs “generally completed in less than one year.” There is no contractual annuity; the “stickiness” is qualification and relationship, not contract.

Footprint & balance-sheet shape. ~100 sites, ~30 countries, 135,000 employees (Asia 71k / Americas 49k / Europe 15k). FY2025 revenue by geography: US $7,444M (jumped on domestic datacenter), Mexico $5,689M, China $4,196M (declining from $5,868M FY23), Malaysia $3,644M; foreign revenue 75% (down from 86% FY23 as US datacenter scaled). Asset-light: long-lived assets just ~$3.3B on $29.8B revenue (~11% of sales); capex held to 1.5–2% of revenue; a large receivables-securitization program ($4.15B sold FY2025) and customer inventory deposits fund working capital.

Verdict: A scaled, diversified, asset-light, single-customer-business-unit EMS with the cleanest customer diversification in its peer group. Revenue is thin-margin, uniform-margin (~5.4% across segments), and program-based. The three-segment recast is an honest re-labeling that correctly surfaces Intelligent Infrastructure as the growth engine — but it does not change the underlying economics, which are those of an assembler.

3. Industry Dynamics

Where EMS sits in the value chain. Electronics manufacturing services occupy the lowest-rent tier of the electronics value chain: silicon/IP owners (Nvidia, Broadcom, Arista — 40–65%+ gross margins, 40–60% operating margins) → system designers/OEMs/hyperscalers → EMS/ODM contract manufacturers (3–6% operating margins) → component suppliers. Rent accrues to scarce silicon and to system design and brand; the assembly tier earns a thin conversion fee on flow-through materials. Jabil’s ~5.4% segment margin is dead-center in that band and confirms the structural read.

Competitive set. The industry is fragmented and intensely cost-competed:

• Hon Hai / Foxconn (~$210B+ revenue) — the scale leader, Nvidia’s principal GB200 rack integrator; roughly 7x Jabil’s revenue.
• Flex (~$28B) — Jabil’s closest scale-and-mix peer (and, per the factor model, its #1 statistical twin); currently restructuring/spinning units.
• Celestica (~$12–19B market cap) — the AI-EMS darling, with its Connectivity & Cloud Solutions (CCS) segment and Hardware Platform Solutions earning higher (~8%) margins on a more concentrated, design-led hyperscaler book.
• Sanmina, Benchmark Electronics, Wistron/Quanta/Wiwynn (ODMs) — Wiwynn’s revenue rose ~164% in 2025 on AI servers, illustrating how fast ODM capacity floods the AI-assembly pool.

Jabil’s own 10-K lists the competitive factors candidly, with cost first: “cost; accelerated production time-to-market; higher efficiencies; global locations; rapid scaling of production; advanced technologies; quality; and improved pricing of components.” An industry whose first-named competitive factor is cost is, by definition, not a moat industry.

Greenwald lens — is there a barrier to entry? Largely no. The three genuine advantage types in Greenwald’s taxonomy are supply/cost advantages, demand-side customer captivity, and economies of scale tied to captivity. Jabil has a scale/cost competency (procurement leverage, global footprint, utilization) but it is a competency, not a barrier — Foxconn is far larger, and new ODM capacity enters AI assembly continuously. There is some program-specific customer captivity (qualification, NPI ramp, dedicated tooling and business units, the cost/risk of re-qualifying a new manufacturer), but it is shallow and resets every product generation, and customers deliberately multi-source. The clincher is the financial test: if Jabil had real captivity it would show up as segment-margin dispersion and pricing power; instead margins are flat at ~5.4% everywhere. Greenwald verdict: no durable moat; a strong operator in a no-moat industry.

Marathon capital-cycle lens. AI/datacenter manufacturing is a textbook capital-cycle subject: extraordinary returns are attracting extraordinary capital (Foxconn, Quanta, Wiwynn, Celestica, Jabil all adding rack/server/networking capacity simultaneously). High returns invite supply, which historically mean-reverts margins. The nuance specific to EMS: because the model is asset-light (capex 1.5–2% of sales, customer-funded materials), the classic “denominator blow-up” (over-investment crushing ROIC) is muted — Jabil is not building $10B fabs. The real capital-cycle risk here is a volume air-pocket against a fixed-cost (labor, facility, SG&A) base, plus pricing compression as ODM capacity floods in, plus customer in-sourcing in a downturn — exactly the dynamic that produced Jabil’s lifetime −90% max drawdown across prior cycles.

Verdict: Structurally bad industry — fragmented, low-margin, customer-power-dominated, cyclical, no barriers to entry. The current AI/datacenter capex supercycle is a powerful cyclical tailwind sitting late in a capital-flooding cycle. Jabil is one of the best operators in a bad neighborhood; that is a real distinction, but it is not a moat.

4. Competitive Position

The actual edge. Jabil’s competitive position rests on three things, none of which is a durable barrier:

1. Scale + global footprint + procurement leverage — a genuine cost/efficiency competency that lets Jabil bid competitively and serve large, multi-region programs. But Foxconn out-scales it ~7x, and Flex is comparable, so this is table stakes among the top tier, not a differentiator.
2. Program-specific switching costs — qualification, NPI engineering, dedicated business units, regulatory certification (especially in healthcare/auto). Real but shallow; they raise the cost of switching mid-program but reset at each product generation, and customers structurally multi-source.
3. Diversification + breadth of capability — the ability to offer “one-stop” integration across compute, storage, networking, power, cooling, and rack-level integration. Management’s stated AI strategy is to “enter where we have a capability the customer needs, deliver with quality, and then expand” — precisely how the second and third hyperscaler relationships grew. This is a real commercial advantage in winning share, but it is a sales motion, not an economic moat.

Head-to-head. Against Celestica, Jabil is cheaper on EV/sales and EV/EBITDA, more diversified by customer, but lower-margin — Celestica’s CCS earns ~8% vs Jabil’s Intelligent Infrastructure ~5.4%, implying Celestica’s AI book is more design-led/higher-content. Against Flex, Jabil is the closer twin; Flex’s higher-value “CPI” (Communications/Enterprise/Cloud) work runs ~9.2% margins — again above Jabil’s II segment, again implying Jabil’s AI mix is weighted to lower-content assembly. Against Foxconn, Jabil simply cannot match scale in racks. The uncomfortable read: on the very segment driving the re-rating, Jabil earns lower margins than the two peers most comparable — evidence it is capturing volume, not the highest-content design rent.

Is the AI win durable? Partly. The wins are real (a 16% anchor customer, second and third hyperscalers, Mikros liquid-cooling and Hanley modular-power capability, silicon photonics). But the work is commoditizable: open standards (Ethernet, OCP), hyperscaler in-house silicon, and a crowded field of Foxconn/Quanta/Wiwynn/Celestica all bidding for rack and server integration. Hyperscalers explicitly multi-source and design to standards precisely to keep their assemblers interchangeable. The durability lever is whether Jabil moves up the value stack (power, advanced liquid cooling, optics, rack-level integration, services/recurring via Hanley) fast enough to earn content rent rather than assembly spread — the single most important thing to watch.

Verdict: Crowded market, weak differentiation, no durable moat. Jabil’s relative strength is diversification and operational excellence, not depth of advantage. The AI franchise is genuine and well-executed but commoditizable, and — tellingly — currently lower-margin than the comparable franchises at Celestica and Flex.

5. Growth History and Forward Opportunities

Historical growth. Revenue: FY2020 $27.3B → FY2021 $29.3B → FY2022 $33.5B → FY2023 $34.7B → FY2024 $28.9B → FY2025 $29.8B. The FY23→FY24 decline is not organic deterioration — it is the Mobility/BYD divestiture (~$6B of mobility revenue exited) plus soft 5G/EV/renewables end-markets. Underneath, the mix rotated hard toward Intelligent Infrastructure (II rose from 32% of revenue FY23 to 41% FY25). TTM revenue through Q2-FY26 had re-accelerated to ~$32.7B, and management now guides FY2026 to ~$35B (+17%).

The AI ramp (the whole story). AI-related revenue (a Jabil-defined cross-segment metric, not a GAAP segment) grew from ~$9B (FY2025) to a guided ~$13.6B (FY2026) — +50%, +$4.6B — and management expects a “similar percentage” growth rate in FY2027 off the larger base, implying roughly ~$20B. Within Q3-FY26, Intelligent Infrastructure grew +21% YoY (networking +>50% on an India ramp), and Q4 II is guided +32% YoY. Drivers: a third hyperscaler win (expected ~$200M FY27 → $1B+ FY28), the second hyperscaler’s storage ramp, semicap test (chip-technology churn driving test demand), new capacity in North Carolina/Memphis/India, and the higher-margin Hanley power and Mikros cooling capabilities.

The Adani India alliance. Announced June 2026: a non-definitive strategic alliance with Adani Enterprises to build a multi-gigawatt AI-datacenter manufacturing platform in India (next-gen liquid-cooled racks, servers, storage, networking, plus power distribution/transformers/switchgear/thermal). Management is explicit it is not yet a definitive framework and a meaningful contributor only in FY2028. Potentially large, but speculative and capacity-gated.

Quality of the growth — the crux. The growth is organic and real, but high-volume rather than high-margin. Intelligent Infrastructure earns the same ~5.4% segment margin as the rest of the company; against Flex CPI (~9.2%) and Celestica CCS (~8%), Jabil’s AI mix is evidently lower-content. So Jabil is capturing the AI build-out’s throughput, amplified at the per-share level by the ~27% share-count reduction — but it is not (yet) capturing the design rent. The highest-quality growth prize is arguably not the contested AI assembly pool but the under-penetrated healthcare/CDMO outsourcing opportunity (long cycles, sticky, higher margin, immature outsourcing), plus digital commerce automation/robotics.

Verdict: Mixed-to-good quality growth. The top-line growth is genuine, organic, and well-executed, and per-share compounding is amplified by buybacks. But the AI growth driving the re-rating is volume not rent (same ~5.4% margin), and the margin-accretion thesis (“6%+ in FY27”) is the unproven fulcrum. High-volume growth at a thin, undifferentiated margin does not, by itself, justify a 99.6th-percentile price-to-sales multiple.

6. Financial Quality

Margin structure. Thin but stable and modestly rising on mix: GAAP gross margin ~8–9% (8.9% FY25), GAAP operating margin ~4.8–5.0%, core operating margin ~5.8% (Q3-FY26) and guided to ~6.4% in Q4-FY26 (seasonally strongest) and “above 6%” for FY2027. Core operating income Q3-FY26 was $504M (5.8%) vs GAAP $445M (5.1%). The margin trajectory is genuinely improving with the datacenter/power/cooling mix — but off a low base, and the improvement is incremental (10 bps at the FY level FY25→FY26 guide).

The GAAP-vs-core gap (quality-of-earnings). Jabil reports company-defined “core” operating income, core EPS, and adjusted FCF that exclude amortization of intangibles, stock-based compensation (~$107M FY25), restructuring/severance, and divestiture/impairment items, taxed at a normalized core rate. The gap is large and persistent: FY2025 GAAP diluted EPS was $5.92 vs core ~$9.0 — roughly a $3/share (~50%) uplift. FY2026 core EPS is guided to $12.70 on a much lower GAAP base. Two flags: (1) restructuring is a perennial “adjustment” — −$181M FY25, −$296M FY24, −$57M FY23 — so calling it non-recurring is generous (the same EMS quality-of-earnings flag flagged in the Flex/Celestica work); (2) FY2024 GAAP EPS ($11.17) is a trap — it was inflated by the ~$0.9B Mobility/BYD divestiture gain and is not run-rate. Analysts and the comp section both anchor on core, so the body uses core for valuation while flagging the gap.

Cash generation & capital intensity. This is the genuine strength. Adjusted FCF rose $274M (FY21) → ~$1,172M (FY25), guided to >$1.4B FY26. Capex was cut from ~4% to ~1.6% of sales — the asset-light model lets revenue and FCF grow without a capital drag. EV/adjusted-FCF is ~21–22x at the current price.

Working capital. Structurally favorable but a downside risk: accounts payable (~$7.9B) funds inventory (~$4.7B); the cash-conversion cycle is ~13 days; customers pre-fund materials via inventory deposits ($1.2B). Q3-FY26 inventory days were elevated (84 gross, ~68 net of deposits, above the 55–60 target) on AI shipment timing — management expects normalization. In a demand air-pocket this negative-working-capital structure reverses violently (payables unwind faster than inventory clears), a classic EMS cash risk.

Returns on capital. ROIC ~20–21% (FY22–FY25) — genuinely high and genuinely real, but it is an asset-light artifact: a tiny invested-capital denominator (long-lived assets ~$3.3B, customer-funded materials, $4.15B receivables securitized) divided into a thin-margin numerator. It is not evidence of a barrier to entry — it is evidence of a capital-efficient operating model. ROE (~10.8% reported, but mechanically distorted) and the eye-popping P/B (29.8x) are buyback artifacts: equity has been hollowed to ~$1.5B (treasury stock ~$7.9B), book value per share is just ~$12.49, and tangible book is near zero. P/B is therefore meaningless here (like a heavily-bought-back name) and should be ignored in favor of P/S and forward core P/E.

Balance sheet. Investment-grade; net debt ~$2.6B; total debt ~$4.4B; cash ~$1.4B (Q3-FY26); debt/core EBITDA ~1.3x. Sound and unstretched.

Verdict: A high-quality cash machine built on a low-quality margin. The FCF generation, capital efficiency, and ~20% ROIC are real and admirable. But economics improve only modestly with scale/mix (not dramatically), the GAAP-to-core gap and perennial “restructuring” warrant skepticism, ROE/book are hollowed by buybacks, and the negative-working-capital structure is a downside cash risk. Good business; not a great one.

7. Capital Allocation

Capital allocation is Jabil’s strongest qualitative pillar — and also where the honest knocks live.

The buyback (the central per-share story). Jabil has repurchased roughly $5.11B of stock over FY2021–FY2025 ($428M / $696M / $487M / $2,500M / $1,000M), cutting diluted shares from 152M to 111M (−27%) — the dominant driver of per-share compounding. In Q3-FY26 it repurchased ~$291M under a $1B authorization it intends to complete in Q4. The critique: the buyback is valuation-insensitive. The cheap repurchases (FY21–FY23, ~$50–130) were excellent capital allocation; but the largest buybacks (~$4.5B+ across FY24–FY26) were executed at progressively higher prices (~$110 → ~$380), and the marginal dollar today is being spent at the richest valuation in company history. Buying back ~all of free cash flow regardless of price is a real, if not thesis-breaking, concern — it converts a low-multiple-era strength into a high-multiple-era risk.

Dividend. Token: ~0.1% yield, ~5% payout. Not part of the thesis.

M&A scorecard — net intelligent.

• The Mobility/BYD SALE (Dec 2023, ~$2.1B net cash, ~$0.9B gain) is the best decision of the period. Jabil exited a concentrated, low-margin, single-customer (Apple-adjacent) mobility business near a peak and redeployed into AI/datacenter, healthcare, and buybacks. A genuinely well-timed, value-creating divestiture.
• Bolt-on acquisitions (~$385M FY25) are disciplined, capability-led tuck-ins up the value chain: Mikros Technologies (liquid cooling), Pharmaceutics International / Pii ($309M, pharma CDMO), Hanley (modular power distribution + services/recurring, double-digit margin, tracking ahead of plan), Intelligent Audio/Retronix. These move Jabil toward higher-content, higher-margin capabilities — exactly the right strategic direction.

Insider behavior — a negative tell. Across the trailing Form 4 corpus (165 filings scanned since Sep-2024, spot-checked back to 2021), there are zero open-market purchases (code P) by any insider. Activity is grants, tax-withholding, gifts, and sales only. Executive Chairman Mark Mondello sold ~$39M+ in FY2025 and continued into 1H-FY26 (Dec-25 $11.3M + $4.2M; Jan-26 multiple tranches ~$14.5M), a mix of 10b5-1 and discretionary (un-marked) sales. CEO Mike Dastoor sold ~$5.2M FY25 plus Jan-26 ($3.7M) and Apr-26 ($2.7M) discretionary tranches. The signal: insiders net-sold into the 7x melt-up with no conviction buying. Mitigant: Mondello still holds ~1.1M shares; Dastoor’s holdings are modest (~33K). This is a soft-negative on insider conviction, not a red flag of distress.

Incentive design. Annual bonus metrics = corporate core operating income, core operating margin, and free cash flow; PBRSUs = 3-year core-EPS growth + relative TSR (vs S&P Composite 1500 Tech Hardware). The FY23–FY25 PBRSU cumulative core-EPS target ($26.00) was beaten ($26.87 → 135% vesting). The concern, consistent with the broader cohort: pay is anchored on the same “core”/adjusted EPS and operating income that exclude intangible amortization and SBC and flatter M&A — with no explicit ROIC hurdle. The structure rewards growth and adjusted profitability, not returns on capital. CEO Dastoor FY25 total comp ~$16.1M; Exec Chairman Mondello ~$33.3M (larger than the CEO); pay ratio ~1,325:1 (a global low-wage manufacturing base inflates this); say-on-pay only ~72% FOR (Jan-2025) — a governance amber flag, likely reflecting Mondello’s outsized pay and the CEO-transition mega-grants.

Other governance/structural items (8-K timeline): CEO transition Mondello→Dastoor (Oct-2024, Mondello to Executive Chairman); FY2025 three-segment reorganization; a December-2024 Amazon stock-purchase warrant for up to ~1.16M shares (hyperscaler alignment but future dilution); two $1B buyback authorizations; no litigation, restatement, or material-weakness disclosure in the five-year corpus (a clean compliance record).

Verdict: Intelligent capital allocation on net — the divestiture timing, the early-cycle buybacks, and the capability-led bolt-ons are genuinely well done. The honest knocks: buying back ~all FCF with zero valuation sensitivity at record prices; comp anchored on adjusted core EPS with no ROIC gate; and an insider/governance picture (zero buys, heavy chairman selling, ~72% say-on-pay) that leans mildly negative on conviction and alignment.

8. Changes and Headwinds — Last Two Years

Strategic changes (mostly strengthening the per-share story, neutral-to-mixed on quality):

• Mobility/BYD divestiture (Dec 2023) — the pivot: shed ~$6B low-margin concentrated revenue near peak; redeployed to AI/healthcare/buybacks. Strengthens mix and diversification.
• Three-segment reorganization (FY2025) — surfaced Intelligent Infrastructure as the #1 segment and re-told the equity story around AI. Presentational, but it correctly reframes where growth is.
• CEO transition (Oct 2024) — Mark Mondello → Mike Dastoor (CFO-turned-CEO); Mondello to Executive Chairman with comp still exceeding the CEO’s — a governance watch-item.
• Capability-led M&A — Mikros (cooling), Hanley (power + services), Pii (pharma CDMO): the right direction (up the value chain), modest in size.
• AI-revenue inflection — $9B → $13.6B → ~$20B path; third hyperscaler win; Amazon warrant; Adani India alliance (FY2028). The re-rating engine.

Headwinds / risks accumulating:

• Valuation re-rating itself — at the richest multiple in company history, the bar for continued outperformance is now high; the June-17 “sell-the-news” reversal is an early warning that good news is priced in.
• Customer re-concentration toward the 16% AI anchor (Customer A) — reverses the diversification benefit.
• Component shortages — management flagged high-bandwidth memory, high-density-interconnect PCBs, DDR4-and-below as tight (a gating factor on the FY27 ramp, partly mitigated by hyperscalers’ allocation priority).
• Ramp inefficiency — new capacity (NC/Memphis/India) phases in through early CY2027 with associated ramp drag on margins (management’s own caveat behind the “6%+” FY27 framing).
• Cyclical/macro — auto remains volatile; consumer/Connected Living mixed; the entire AI-capex cycle is the dominant macro dependency.

Verdict: The last two years’ changes have strengthened the business (better mix, diversification, value-chain position) and amplified the per-share story (buyback), which is largely why the multiple re-rated. But they have simultaneously raised the risk by lifting the valuation to a level that prices in continued perfection and by re-concentrating toward a single AI anchor.

9. Risk Analysis (Risk Matrix)

Net read: The dominant risks are valuation/cycle-driven, not solvency-driven. A total or catastrophic loss is remote (sound balance sheet, diversified, clean record). The realistic downside is a de-rating + cyclical earnings reset — a 5–6%-margin assembler re-priced from a 99.6th-percentile P/S back toward its long-run range as AI growth decelerates — which, given beta 1.54 and ~31% idiosyncratic vol, could be a 40–60% drawdown without any business catastrophe.

10. Valuation Discussion (Embedded Expectations)

No price target, no recommendation — embedded expectations and scenarios only.

Where the multiple sits. At ~$372 (~$28.4B market cap / ~$31B EV):

• AZI own-history percentiles: GAAP P/E 50x (93rd), P/B 29.8x (99.6th, ignore — buyback-hollowed equity), P/S 1.24x (99.6th), composite 97.4th — the richest valuation in the stock’s own history.
• EV/sales ~0.95x; EV/EBITDA ~13x; EV/adjusted-FCF ~21–22x.
• On core (adjusted) earnings: ~29x FY26 core EPS (~$12.70); ~24x a plausible FY27 (if core EPS grows ~20% to ~$15–16). Historically Jabil traded ~8–12x core EPS.

Comp table (approximate; multiple sources and dates):

Jabil screens cheaper than Celestica on most metrics (Celestica is the extreme of the AI-EMS re-rating) but more expensive than Flex (its closest twin) and far more expensive than Foxconn. The key own-history fact dominates: this is the most Jabil has ever cost relative to its own sales and earnings.

Embedded-expectations math. At ~$31B EV against ~$1.4–1.5B adjusted FCF (~21–22x EV/FCF) and ~24x FY27 core EPS, the price embeds a triple-compound bet:

1. AI revenue keeps compounding ~30–50%/yr from $13.6B (peak-cycle hyperscaler capex does not roll over);
2. Core operating margin rises from 5.8% to ~6.5–7% (Jabil captures content/rent, not just volume); and
3. The buyback keeps shrinking the float at elevated prices.

For the stock to “grow into” a normalized EMS exit multiple (~16–18x core EPS) without a de-rate, core EPS would need to roughly double from ~$12.70 toward ~$22–25 over 3–4 years — feasible only if all three legs hold. What the market is pricing correctly: the AI-content wins are real, ROIC is ~20%, diversification is best-in-class, the balance sheet is sound, and FCF is robust. What it is likely pricing incorrectly: extrapolating peak-cycle hyperscaler capex as a secular constant; crediting AI growth as high-margin when Jabil earns the same ~5.4% on it (volume, not rent); and assigning a 99.6th-percentile price-to-sales multiple to an assembler whose customers can in-source and whose scale lags Foxconn.

Scenarios (illustrative, no price target):

• Bear: Hyperscaler-capex air-pocket and/or in-sourcing; revenue stalls ~$30–32B, core OM slips to 5.0–5.3%, core EPS ~$10–12; multiple re-rates toward EMS norm (~10–12x). Large drawdown driven by multiple more than earnings.
• Base: AI grows +20–30% then decelerates; other end-markets normalize; core EPS ~$15–17 by FY28; multiple normalizes toward ~16–18x. Modest upside-to-flat from here as multiple compression offsets earnings growth.
• Bull: AI sustains +40–50% (Adani contributes from FY28), core OM reaches 6.5–7% (real content capture), core EPS ~$20–24; multiple holds ~22–25x. Requires the capital cycle not to revert — the heroic assumption.

Verdict: Jabil is priced for a best-case continuation of a peak-cycle AI capex boom and a margin inflection it has not yet demonstrated and continued buyback accretion at record prices — all at the 99.6th percentile of its own price-to-sales history. The asymmetry is unattractive at this level: the bull case requires the capital cycle to defy gravity; the bear case needs only a normalization of either the cycle or the multiple.

11. Variant Perception

Consensus. The Street view (broadly Buy/Outperform, multiples expanding) is that Jabil is a secular AI-infrastructure winner with an asset-light model, improving margins, a fortress of free cash flow, and an aggressive buyback — a “picks-and-shovels” way to own the AI build-out at a lower multiple than the silicon names. The June raise reinforced this.

Strongest bull case. Jabil is the diversified, asset-light, lower-risk way to play a multi-year AI-infrastructure supercycle. It is winning hyperscaler share (2nd and 3rd customers), moving up the value chain into power (Hanley), cooling (Mikros), and optics, with a massive India optionality (Adani) from FY2028. ~20% ROIC, >$1.4B and growing FCF, and a 27%-and-counting share-count reduction compound per-share value. At ~24x FY27 core EPS for a company growing core EPS ~20–25%, it is “not expensive” on growth-adjusted terms, and far cheaper than Celestica.

Strongest bear case. Jabil is a no-moat, capital-cycle assembler capturing AI volume not margin (Intelligent Infrastructure earns the same ~5.4% as everything else, below Flex and Celestica), priced at the richest multiple in its history (P/S 99.6th percentile) on the extrapolation of peak hyperscaler capex. The recent doubling already prices the good news (the June-17 raise produced an intraday ATH and a same-day reversal). Insiders sold ~$50M+ with zero open-market buys; say-on-pay was only ~72%; comp is anchored on adjusted core EPS with no ROIC gate. With beta 1.54, ~31% idiosyncratic vol, and a lifetime −90% max drawdown, the downside in a cycle turn is severe — and needs only a multiple normalization, not a business catastrophe.

Factor-positioning read (FactorsToday, 2026-06-18). JBL is statistically a high-beta AI-cyclical, not a compounder: realized beta 1.54, LowVolatility −0.55 (strongly anti-defensive), Quality only +0.10, Value −0.12 (anti-value), Momentum +0.34, a dual Industrials +0.29 / Technology +0.29 straddle, plus AI/semicap custom-factor loadings. Idiosyncratic vol ~30.6% (R² ~0.46 — overwhelmingly a single-name story). Relative strength 89th percentile, alpha +0.25, sitting just −3.6% off its peak (at the highs, not a falling knife). Leaderboard (annualized): 3y +55%/Sharpe 1.30, 1y +82%/Sharpe 1.93 — but lifetime max-drawdown −90%, lifetime Sharpe 0.28 (the EMS down-cycle tail). The #1 factor-neighbor is Flex (0.944), then a cluster of high-beta AI/3x-leveraged ETFs. The model corroborates the framing: a crowded, high-beta, peak-cycle momentum trade that has already run — the risk is a momentum reversal / de-rate, not a value trap.

The 3–5 assumptions that matter most:

1. Does hyperscaler/AI-datacenter capex keep growing (not just stay high) through FY2027–2028?
2. Can Jabil push core operating margin durably above ~6% (rent capture, not just volume)?
3. Does Jabil hold/gain share against Foxconn and the ODMs — and contain Customer-A concentration?
4. Does the multiple hold (~22–25x) or normalize (~16–18x) toward EMS norms?
5. Is the buyback accretive or value-destructive at record prices?

Falsification tests.

• Bull case is falsified by: hyperscaler capex cuts or AI-revenue growth decelerating below ~20%; core OM stuck at ≤5.5%; or — the most immediate signal — raised guidance that fails to lift the stock (June 17 was already a warning shot).
• Bear case is falsified by: AI revenue compounding 30%+ through a visible capex deceleration (proving share gains and content capture) combined with a durable 6.5%+ core operating margin that proves the rent — not just the volume — is being captured.

12. Fact vs. Interpretation Table

13. Open Questions

1. What exactly is Jabil’s AI content mix? How much of the $13.6B is low-content rack assembly vs higher-content power/cooling/optics? This determines whether the 6%+ margin target is achievable. (Not disclosed at this granularity.)
2. Who is “Customer A” (16%)? Likely a hyperscaler/cloud anchor in Intelligent Infrastructure — but its identity, contract terms, and in-sourcing risk are undisclosed.
3. Will the Adani India alliance become definitive, and on what capital/JV structure? Management is explicit it is not yet a framework; the economics and Jabil’s capital exposure are unknown.
4. How durable is the margin inflection through the FY27 capacity ramp? Management guides “6%+” but flags ramp inefficiency through CY2027.
5. What is the true normalized FCF in a down-cycle, once the negative-working-capital tailwind reverses?
6. Why did say-on-pay fall to ~72%, and will the board respond (Mondello’s pay > CEO’s; no ROIC metric)?

14. What Must Be True

Bull case — what must be true:

• Hyperscaler/AI-datacenter capex must keep growing (not merely stay elevated) through FY2027–2028, sustaining 30–50% AI-revenue growth off a ~$13.6B base.
• Core operating margin must break durably above ~6% (toward 6.5–7%), proving content/rent capture rather than thin assembly spread.
• Jabil must hold/gain share against Foxconn and the ODMs while containing Customer-A concentration, and convert the third hyperscaler and Adani optionality into material revenue.
• Falsification test: AI-revenue growth decelerates below ~20%, OR core OM remains stuck at ≤5.5%, OR a raised guide fails to move the stock (the multiple has already de-rated in investors’ minds). The June-17 intraday reversal is the first crack in this case.

Bear case — what must be true:

• The AI-capex cycle proves to be a peak that decelerates or air-pockets (Marathon capital-cycle reversion), and/or customers in-source/multi-source away commoditizable assembly.
• The 99.6th-percentile price-to-sales multiple normalizes toward the stock’s long-run EMS range as growth slows.
• Margin stays stuck at ~5.4–5.8% (volume not rent), and the buyback at record prices proves value-neutral.
• Falsification test: AI revenue compounds 30%+ through a visible industry capex deceleration (proving genuine share/content gains) and core operating margin holds durably above 6.5% — together demonstrating Jabil has escaped the assembler’s economics. Absent both, the bear case stands.

The analysis body takes no investment position and contains no price target; the only position expressed in this article is the clearly-labeled Claude's Take block at the top, which is the author’s own subjective view.

15. Source Appendix

See the accompanying source appendix (JBL_source_appendix.md) for the full, dated citation list. Primary sources include: Jabil Inc. FY2021–FY2025 Forms 10-K and FY2026 Forms 10-Q (SEC EDGAR, CIK 0000898293); Q3-FY2026 earnings call transcript (2026-06-17); DEF 14A proxy statements (FY2021–FY2025); Form 4 insider filings (2021–2026); AZI valuation-index and price data (accessed 2026-06-19); ROIC.ai fundamentals, ratios, and enterprise value (accessed 2026-06-19); FactorsToday factor model (accessed 2026-06-18); and published research on comparable companies on Celestica (CLS), Flex (FLEX), Fabrinet (FN), and TTM Technologies (TTMI).

APPENDIX A — Standard Diligence Questionnaire

Supplemental to the analysis. Fact/Interpretation/Assumption labels where it matters. Report date 2026-06-19.

General

What thoughtful questions have other investors asked about this company?

• Is the AI/Intelligent-Infrastructure revenue high-margin content or low-margin assembly volume? (The crux — Jabil’s II segment earns the same ~5.4% as everything else.) [Interpretation]
• Is hyperscaler capex durable or peaking? ~46% of revenue is now AI-related. [Fact/Open Question]
• Can core operating margin durably exceed 6% (management’s FY27 target)? [Open Question]
• Is buying back ~all FCF at record-high prices smart capital allocation? [Interpretation]
• What is the right multiple for a 5–6%-margin EMS — and is the current 99.6th-percentile P/S sustainable? [Interpretation]

Cyclicality & Earnings Nature

Are earnings at a cyclical high or low? Near a cyclical high, driven by the AI-datacenter capex supercycle. AI revenue $9B→$13.6B (FY25→FY26); core EPS guided ~$12.70 FY26 vs ~$9 FY25. [Fact] The risk is that this is a peak being extrapolated. [Interpretation]

Driven by external environment or internal actions? Both. Externally: the AI capex boom (the dominant driver). Internally: the Mobility divestiture (cleaner mix), the 27% share-count reduction (per-share leverage), and capability-led M&A. [Fact]

How stable are revenues? Program-based, non-recurring (no contractual annuity); historically cyclical (lifetime −90% max drawdown). Diversification across 87+ customers and three segments dampens but does not eliminate cyclicality. [Fact]

Outlook for products/services & market size. EMS market ~$600B+ (2024) heading toward ~$1T by early-2030s. [Interpretation] AI-infrastructure, healthcare CDMO, and digital-commerce automation are the growth vectors; mobility and parts of consumer are mature/shrinking. Global, with US datacenter revenue surging (US revenue $7.4B FY25). [Fact]

Business Quality & Competitive Moat

Is the industry getting more or less competitive? More — ODM capacity (Foxconn, Quanta, Wiwynn) is flooding into AI-server assembly (Wiwynn +164% revenue 2025). [Fact/Interpretation]

How profitable is the business (ROIC, ROE)? ROIC ~20–21% (genuinely high, but an asset-light artifact of a tiny invested-capital denominator). ROE ~10.8% reported but distorted by buyback-hollowed equity (BVPS ~$12.49; near-zero tangible book). [Fact]

How profitable is the industry / barriers to entry? Structurally low — 3–6% operating margins, fragmented, no real barriers to entry, customer power dominant. [Fact/Interpretation] Greenwald verdict: no durable moat.

Can it be easily understood? Yes — a contract manufacturer that builds to spec for a thin conversion fee, plus a buyback. [Fact]

Can it be undermined by foreign low-cost labor? It is foreign low-cost labor (135k employees, 71k in Asia; 75% foreign revenue) — that is the model. The threat is the reverse: scale leaders (Foxconn) and ODMs underbidding, and customers in-sourcing. [Fact/Interpretation]

Do brands matter? Nature of competition? No consumer brand; competition is on cost, scale, footprint, time-to-market, quality. The 10-K lists “cost” first among competitive factors. [Fact]

Customers’ switching costs? Real but shallow — qualification, NPI, dedicated business units, tooling, regulatory certification (deepest in healthcare/auto). Reset each product generation; customers multi-source by design. [Interpretation]

Financial Condition & Balance Sheet

Assets not fully recognized? Customer relationships and qualification status are off-balance-sheet intangibles. The $4.15B securitized receivables and customer inventory deposits flatter working capital. [Fact]

Off-balance-sheet liabilities? The receivables-securitization program; operating leases (in ROU assets). No undisclosed material items in the 5-yr corpus. [Fact]

How conservative is the accounting? Mixed. GAAP is clean (no restatement/material weakness), but the company-defined “core” metrics exclude perennial restructuring (−$181M/−$296M/−$57M FY25/24/23), SBC, and amortization — a ~50% uplift to EPS that flatters M&A and is also the comp metric. [Fact/Interpretation]

How CapEx-hungry? Very light — 1.5–2% of revenue (cut from ~4%). The core attraction of the model. [Fact]

Capital Allocation & Management

How much FCF, and how is it used? Adjusted FCF $274M→$1,172M (FY21→FY25), >$1.4B guided FY26. Used overwhelmingly for buybacks (~$5.1B FY21–25, −27% shares) plus bolt-on M&A and a token dividend (~0.1% yield). [Fact]

Significant acquisitions recently? Capability-led tuck-ins: Mikros (cooling), Pii ($309M CDMO), Hanley (modular power + services), Intelligent Audio. The headline divestiture (Mobility→BYD, ~$2.1B, Dec-2023) was the marquee transaction and well-timed. [Fact/Interpretation]

Buying back shares? Yes, aggressively — but valuation-insensitive, with the largest buybacks at the highest prices. [Fact/Interpretation]

Issuing shares to insiders? SBC ~$107M/yr; plus a Dec-2024 Amazon warrant (up to ~1.16M shares). Net share count is down 27% (buyback > dilution). [Fact]

Compensation policy. Bonus = core operating income / core operating margin / FCF; PBRSU = 3-yr core-EPS growth + relative TSR. No ROIC hurdle. CEO Dastoor ~$16.1M; Exec Chairman Mondello ~$33.3M; pay ratio ~1,325:1; say-on-pay ~72%. [Fact]

Motivations of management. Insiders net-sold ~$50M+ into the melt-up with zero open-market buys — low signaled conviction at current prices. [Fact/Interpretation]

Valuation & Market Data

ADR / MLP / K-1? No — US common stock, NYSE-listed, standard 1099. [Fact]

Dividend policy. Token quarterly dividend (~0.1% yield, ~5% payout); not a thesis driver. [Fact]

How profitable / is net income diverging from CFO? Thin GAAP net margin (~2.2% FY25); CFO is robust and exceeds net income (D&A, working-capital benefit), the normal EMS pattern. The divergence is favorable (cash > accounting earnings), but reverses in a downturn. [Fact]

Risks & Downside

What would cause the stock to decline? A multiple de-rating from the 99.6th-percentile P/S; a hyperscaler-capex air-pocket; customer in-sourcing; margin failing to exceed 6%; a raised guide failing to lift the stock (already observed June 17). [Interpretation]

Risk of catastrophic loss? Low — investment-grade, net debt/EBITDA ~1.3x, diversified, clean compliance record. [Fact]

Chance of total loss? Very low. The realistic downside is a large drawdown (de-rate + cyclical reset), not impairment — but given beta 1.54 and ~31% idiosyncratic vol, a 40–60% drawdown is plausible without any solvency event. [Interpretation]

Recent News & Events

Has the business environment changed recently? Yes — the AI-infrastructure ramp is the defining change: AI revenue raised to ~$13.6B FY26, a third hyperscaler won, and the (non-definitive) Adani India alliance announced (June 2026). Q3-FY26 (June 17) beat and raised, but the stock reversed intraday. [Fact]

Significant acquisitions / accounting changes / new markets? Hanley/Mikros/Pii bolt-ons; FY2025 three-segment reorganization; new capacity in North Carolina, Memphis, and India; CEO transition (Mondello→Dastoor, Oct-2024). No accounting-policy changes of note. [Fact]

APPENDIX B — Source Appendix

Report date 2026-06-19. Primary sources prioritized. All facts in the article trace to one of the public sources below.

Primary — SEC Filings (EDGAR, CIK 0000898293)

Primary — Earnings Call Transcript

Quantitative Data Sources (third-party; reconciled to filings)

Comparable Companies Referenced

Public-company comparables used for industry-structure and valuation cross-read (each issuer’s own SEC filings and disclosures).

Notes on Source Reliability

• GAAP vs. core: Jabil’s “core” operating income/EPS and “adjusted” FCF are company-defined non-GAAP measures excluding amortization, SBC, restructuring (perennial), and divestiture/impairment items. The memo uses core for valuation comparability while flagging the ~50% gap to GAAP and the FY24 divestiture-gain distortion.
• AZI P/B percentile (99.6th) is mechanically meaningless for JBL — equity is hollowed by ~$7.9B of treasury stock (buybacks); read P/S and forward core P/E instead.
• “AI-related revenue” is a Jabil-defined cross-segment metric, not a GAAP reportable segment; do not equate it to the Intelligent Infrastructure segment.
• Third-party quantitative sources (AZI, ROIC.ai, FactorsToday) are aggregated/estimated data, not primary; for US-filer facts, EDGAR filings are authoritative and were used to reconcile material figures.