Intel

Phase A — Understand the business

Lens 1 · Company Overview

Intel is an integrated device manufacturer (IDM) — it both designs and manufactures logic chips, the rare combination in an industry that has otherwise bifurcated into fabless designers (Nvidia, AMD, Qualcomm) and pure-play foundries (TSMC). The business now reports in three segments plus an "all other" bucket:

• Client Computing Group (CCG) — PC/edge CPUs (Core Ultra, Core, the flagship consumer brands). FY2025 revenue $32,228M, segment operating income $9,317M. This is the cash engine.
• Data Center and AI (DCAI) — Xeon server CPUs and the (struggling) AI-accelerator line. FY2025 revenue $16,919M, operating income $3,422M. Accelerating: Q1 2026 DCAI revenue $5,052M, op income $1,542M, vs $4,126M / $575M a year prior — a +22% YoY revenue print management called AI-CPU-demand-driven.
• Intel Foundry — the manufacturing arm, selling wafers internally to Intel Products and (nascently) to external customers. FY2025 revenue $17,826M (substantially intersegment), operating loss $(10,318)M. External third-party foundry/assembly revenue was just $307M in 2025 ($159M in 2024) — i.e. the foundry is ~98% an internal cost center today.
• All Other (Mobileye, IMS Nanofabrication, start-ups; Altera through Sept 11, 2025) — FY2025 revenue $3,563M, op income $264M.

Customer & contract structure. Highly concentrated: the three largest customers were 19% / 12% / 12% = 43% of FY2025 net revenue. These are the PC/server OEMs and distributors (Dell, HP, Lenovo and the channel — the filing anonymizes them as Customers A/B/C). Revenue is transactional product sale, not take-or-pay or recurring; there is no subscription cushion.

The thing to understand: Intel is two businesses fused at the hip — a profitable fabless-style products company (CCG + DCAI generated $12,739M segment operating income in FY2025) bolted to a $10.3B-loss-a-year leading-edge foundry it is trying to turn into the only US-based alternative to TSMC. The consolidated result of fusing them: an operating loss of $(2,214)M and net loss attributable to Intel of $(267)M, diluted EPS $(0.06) for FY2025.

Lens 2 · Supply Chain

Intel is unusual: it owns most of its own upstream. Map the chain and name the stakeholders:

• Upstream equipment — Intel's leading-edge ambitions ride on ASML (sole-source EUV/High-NA EUV lithography), plus Applied Materials, Lam Research, KLA, Tokyo Electron for deposition/etch/metrology. The High-NA EUV dependency on ASML is a genuine single-source chokepoint for the 18A/14A roadmap.
• Internal fabs — Intel runs its own wafer fabrication (Oregon, Arizona, New Mexico, Ireland's Fab 34, Ohio under construction). This is the capital sink: Intel's net inventory alone was $11.6B at year-end 2025 (5.5% of total assets).
• External foundry — critically, Intel is also a TSMC customer — it outsources a meaningful share of its own leading-edge tiles (e.g. compute tiles on recent client products) to TSMC even while building Intel Foundry to compete with them. This is the strategic contradiction at the heart of the company.
• Co-investment partners (the financing layer of the "supply chain") — Intel has sold minority stakes in its fab capacity to fund it: Apollo owns 49% of Ireland SCIP (Fab 34) for ~$11.0B net proceeds (2024); Brookfield historically funded the Arizona SCIP (49%, $9,106M non-controlling interest at YE2025, up from $3,888M after $5,108M of partner contributions in 2025). These are off-balance-sheet-ish ways to fund capex without diluting common equity — but they bleed minority interest out of consolidated income.
• Downstream — sells through distributors and directly to OEMs (Dell, HP, Lenovo, Supermicro) → PCs and servers → enterprises and consumers. Hyperscalers (the Customer A/B at ~19%/12%) buy Xeon directly.
• End-customer foundry prospects — Microsoft, Amazon (AWS) and now reportedly Google are the names that matter for the 18A external thesis.

Chokepoints: ASML High-NA (single source), and Intel's own yield on 18A — the supply chain is mostly internal, so execution risk is concentration risk.

Lens 3 · Competitive Advantages (moats)

Intel's moat is real but eroding on three of four edges:

• x86 architecture franchise (durable). Intel + AMD effectively duopolize x86; the installed base, software/driver ecosystem and enterprise validation create high switching costs in PCs and traditional servers. This is the moat that still pays — CCG's $9.3B and DCAI's $3.4B segment operating income come from it. The Nvidia partnership (below) is, in part, a vote that x86 still matters for AI head-nodes.
• Manufacturing scale / US-based leading-edge (contested, subsidized). Intel is the only company attempting leading-edge logic manufacturing on US soil at scale. That has become a geopolitical moat: it is why the US government took equity and why CHIPS money flowed. But it is a moat propped up by subsidy, not yet by economics — Foundry loses $10B+/yr.
• Process leadership (lost, fighting back). Intel ceded the process crown to TSMC over 2018-2024. The entire bull case rests on 18A clawing it back. 18A entered production ramp in H2 2025; Panther Lake (Core Ultra Series 3) is the first product; Clearwater Forest is the server variant. CEO Lip-Bu Tan initially wanted to skip 18A as an external node and bet on 14A, then reversed after yields improved. This is a moat-in-recovery, not a moat-in-hand.
• Bargaining power (weak and weakening). Customer concentration at 43% means the OEMs have leverage. Against suppliers (ASML) Intel has none. Against its foundry competitor TSMC, Intel is simultaneously a customer — the opposite of bargaining power.

Net: the products moat funds the company; the manufacturing "moat" is a subsidized bet that may or may not become a real one. Ground: kb/hardware/wiki/positioning.md, bottlenecks.md (commercial layer).

Lens 4 · Segments

FY2025 segment detail, all ``:

Geography (billing location, ``): US $15,757M (29.8%, and growing — was $12,994M in FY24), China $12,694M (24.0%, shrinking from $15,532M — tariff/export-control drag), Singapore $9,535M, Taiwan $7,672M. The US-up / China-down shift is the geopolitical realignment showing up in the revenue line.

Read: the segment story is "DCAI accelerating, CCG margin eroding, Foundry the $10B anchor." The consolidated loss halved YoY mostly because FY2024 carried a $6,970M restructuring charge and an $8,023M tax-valuation-allowance hit; the underlying operating loss is structural to Foundry. Q1 2026 confirms the trajectory: DCAI op income $1,542M vs $575M, Foundry still −$2,437M for the quarter.

---

Phase B — Measure performance

Lens 5 · Earnings Result (latest print: Q1 2026, period ended 2026-03-28)

unless noted:

• Net revenue $13,577M, +7.2% YoY (vs $12,667M). Beat consensus by 9.4% ($12.4B expected).
• Gross profit $5,347M → GM 39.4%, up from 36.9% a year ago — a real margin improvement.
• Operating loss $(3,136)M — but this is masked by a $4,070M restructuring/impairment charge (primarily a Mobileye goodwill impairment). Ex-charge, the operating line was roughly breakeven-to-positive.
• GAAP net loss attributable to Intel $(3,728)M, diluted EPS $(0.73). Adjusted EPS $0.29 vs ~$0.02 consensus — a ~1,350% surprise on the non-GAAP line.
• Guidance/tone: this was Intel's sixth consecutive quarter of beating guidance; management flagged AI-CPU demand and DCAI strength. Tone has shifted decisively more confident under Tan.
• Balance-sheet flags: share count exploded — weighted diluted shares 5,083M in Q1 2026 vs 4,343M a year earlier (+17%) — the direct fingerprint of the US-government, Nvidia and SoftBank equity issuances (Lens 8). Net inventory $11.6B (YE2025) remains an audit-flagged risk. Non-controlling-interest contributions of $2,064M into Arizona SCIP in Q1 alone show capex is still being part-funded by partners.
• Market reaction: stock +24% the day after the print, its largest single-day gain since 1987 — though that move was off a ~$66 base. A dozen-plus sell-side targets were raised within hours (Bernstein $62→$78, Wells Fargo $58→$80).

Unusual vs. own history: a +24% reaction to a GAAP loss quarter tells you the market is trading Intel on the narrative (margin recovery + adjusted beat + foundry validation), not the GAAP P&L.

Lens 6 · Earnings Calls (sentiment trend)

No transcripts on disk (transcripts/ empty), so ``-grounded. Trajectory across the last several calls:

• Late-Gelsinger era (2024): defensive, "IDM 2.0," repeated apologies on yields and timelines; the $18.8B FY2024 loss (worst since 1986) dominated.
• Tan transition (2025): reset — flatten the org, integrate NEX into CCG/DCAI, divest Altera (51% to Silver Lake), spin focus to 18A discipline and cost cuts.
• Q1 2026 call: notably upbeat — six straight guidance beats, AI business +40% YoY, 18A "yields improving," external-customer "inbound interest" language.

Phrases that appeared: "disciplined capital," "18A ramp," "external foundry interest," "AI head-node." Phrases that disappeared: "five nodes in four years" (the Gelsinger mantra), unconditional capex-growth guidance. The arc is from contrition to cautious confidence — but the confidence is leading the financials, not following them.

Lens 7 · Comps

Peer multiples are ``, dated June 2026. The index carried no hardware peers, so peers pulled manually (AMD, NVDA, TSMC, QCOM).

The killer observation: the entire complex has melted up (AMD +347% YoY, TSMC +132% YoY, INTC ~+500% YoY off the ~$19 low). But Intel is the only name in the table carrying growth-stock multiples (fwd P/E ~133x, EV/Sales ~8x) on a company with a negative trailing EPS and a $10B/yr Foundry loss. NVDA at ~31x earnings and QCOM at ~22x are cheaper than Intel on forward earnings while being structurally more profitable. Intel's valuation is entirely a bet on the 18A/Foundry option, not on the current earnings stream.

Lens 8 · Stock-Price Catalysts (>5% moves, ~last 2 years — the recapitalization story)

This is the most important lens for Intel right now. The stock went from ~$19 (52-wk low) to ~$118-135 in roughly five months. The catalyst chain, all ``:

1. Dec 2024 — Gelsinger ousted. CEO abruptly retired after the $18.8B FY2024 loss. Stock bottomed in this window.
2. Mar 2025 — Lip-Bu Tan named CEO. Reset narrative.
3. Aug 2025 — US government takes ~9.9% stake. Trump administration converted ~$8.9B of CHIPS grants/secure-program funds into 433.3M primary shares at $20.47/share. The single biggest re-rating catalyst — a sovereign backstop.
4. Sept 18, 2025 — Nvidia invests $5B at $23.28/share, plus a co-development deal: Intel builds custom x86 CPUs for Nvidia AI platforms (NVLink-connected) and x86 SoCs with RTX GPU chiplets for PCs. Strategic validation from the AI kingmaker.
5. 2025 — SoftBank invests $2B (~2% stake).
6. Sept 2025 — Altera 51% divested (to Silver Lake), simplifying the story.
7. Apr 23, 2026 — Q1 beat, +24% (best day since 1987).
8. June 2026 — reported Google ~3M AI-chip / foundry deal drove an ~11% pop. (Specifics thinly sourced — treat as ``.)

What the pattern reveals: for two years Intel traded on survival and strategic-validation events (who is willing to put capital/credibility behind it), not on earnings. The market reacts to government/strategic-partner endorsements and foundry-customer signals far more than to the GAAP P&L. That makes the stock a sentiment/option vehicle — wonderful on the way up, brutal if a marquee customer walks.

---

Phase C — Judge people & books

Lens 9 · Management

• CEO: Lip-Bu Tan (since Mar 2025). Track record: legendary semiconductor investor (Walden International) and the operator who turned around Cadence Design Systems (EDA) into a compounding machine — deep credibility with the design ecosystem and foundry customers. Was on Intel's board, resigned in 2024 over strategy disagreements, then returned as CEO — a strong "I'll only take it if I can run it my way" signal.
• Tenure & skin in the game: ~15 months in. Has been buying/holding; the bigger alignment is the strategic-investor syndicate (Nvidia, SoftBank, US gov) now on the cap table.
• Capital-allocation history (Intel's, pre-Tan, is poor): years of buybacks at higher prices, missed the mobile and AI-accelerator waves, over-built capex into a process roadmap that slipped. ROE/ROIC has been negative-to-marginal (net loss in 2024 and 2025). Tan's allocation is different in kind: sell minority fab stakes (Apollo/Brookfield SCIPs), divest non-core (Altera, partial Mobileye via the 23% NCI), cut opex (R&D down to $13,774M in FY25 from $16,546M ), and fund capex with partner + sovereign capital rather than common equity alone — though the equity issuances still diluted holders ~17%.
• Red flags: the government as a ~10% shareholder is a double-edged sword — political capital and a backstop, but also political constraints on plant siting, export policy and capital returns. Heavy reliance on non-controlling-interest financing makes consolidated earnings quality murkier (minority interest absorbed $293M of income in FY25, and a $(553)M swing in Q1 2026 driven by the Mobileye impairment).
• Archetype: professional turnaround operator with founder-level conviction and an investor's eye for the cap table. The right profile for this moment — if 18A yields hold.

Lens 10 · Forensic Red Flags

Income statement, balance sheet, cash flow — all `` unless noted:

• Restructuring as a recurring "one-time" item. $2,191M (FY25), $6,970M (FY24), and $4,070M in Q1 2026 alone. When "non-recurring" charges recur every period, non-GAAP "adjusted" EPS (the $0.29 the market cheered) systematically flatters the picture. Watch the gap between GAAP $(0.73) and adjusted $0.29.
• Mobileye goodwill. Goodwill was $23.9B at YE2025, of which $8.2B sits in the Mobileye reporting unit — and the auditor (E&Y) flagged it as a Critical Audit Matter because impairment indicators appeared in Q4 2025. The Q1 2026 $4,070M charge confirms the write-down began. Further Mobileye impairment is a live risk.
• Inventory valuation. $11.6B net inventory, 5.5% of assets, also a Critical Audit Matter — excess/obsolete-reserve judgment on new-node products with "limited historical data". In a demand air-pocket this is where a surprise charge would come from.
• Non-controlling interests / SCIP structures. The Apollo (Ireland) and Brookfield (Arizona) co-investment vehicles are consolidated (Intel is primary beneficiary) but route a growing slice of economics to minorities — NCI on the balance sheet rose to $12,079M at YE2025 from $5,762M. Earnings attributable to Intel can diverge from consolidated net income; read the "attributable to Intel" line, not the headline.
• Interest & other swing. FY2025 "interest and other, net" was a +$3,257M gain (vs +$226M FY24) — much of it non-operating (the Altera divestiture gain and equity-stake dynamics). This is what flipped FY25 to a near-breakeven income-before-tax of $1,557M despite the $(2,214)M operating loss. Quality-of-earnings: the "improvement" to roughly breakeven is partly financial-engineering and divestiture gains, not operations.
• SBC: $2,434M FY25 — meaningful but down from $3,410M; not egregious for the sector.
• Cash flow vs earnings: capex remains enormous; the company is funding it through asset sales, partner contributions and sovereign/strategic equity rather than operating cash flow — the tell that the core can't yet self-fund the foundry build.

Regulatory findings (required sub-section):

• SEC Litigation Releases / AAERs: None. "No material regulatory or legal findings — verified via SEC EDGAR EFTS (LR, AAER) as of 2026-06-17".
• Non-SEC enforcement (web): EU antitrust — the long-running rebate case ended with a €376.36M fine reimposed in Sept 2023 (down from the original €1.06B, which was annulled). Largely a closed, immaterial-to-cash legacy item.
• Material litigation (Item 3 / web): VLSI Technology patent dispute — exposure once exceeded $3B across multiple patents; Intel won a key Texas jury verdict in May 2025 (jury found Fortress controls VLSI, undercutting damages), but an appeals court reinstated part of the case, so it remains live. This is the one genuinely material contingency.
• Verdict: accounting is audited clean (unqualified E&Y opinion, effective ICFR ); the risks are aggressive non-GAAP add-backs and impairment/inventory judgment, not fraud. Legal tail = VLSI.

---

Phase D — Project & stress-test

Lens 11 · Forward Projection

Building bottom-up from FY2025 actuals + Q1 2026 run-rate, into FY2026/27/28 EPS paths. All outputs `` with arithmetic shown; this is a breadth-loop dive, so no Brier forecast is logged (per --watchlist rule).

Anchors: FY2025 revenue $52,853M, GM 34.8%, op loss $(2,214)M, EPS $(0.06). Q1 2026 annualized run-rate ≈ $54.3B revenue, GM ~39%, with restructuring distorting GAAP. Diluted share count now ~5.0–5.1B.

• Base (FY2026E): Revenue ~$55–57B (CCG flat, DCAI +15–20%, Foundry external still de minimis); GM recovers toward ~40% as 18A internal mix improves and restructuring annualizes down; opex normalizes. Adjusted EPS ~$0.90–1.20. GAAP likely still thin/near-breakeven given recurring charges.
• Bull (FY2026E): DCAI AI-CPU demand + first external 18A wafer revenue + further GM lift → revenue ~$60B, GM ~42%, adjusted EPS ~$1.50–1.80.
• Bear (FY2026E): PC demand air-pocket, China export drag worsens, another Mobileye/inventory impairment, Foundry loss fails to narrow → revenue ~$50B, GM ~36%, GAAP EPS negative again, adjusted ~$0.40.
• FY2027–28E: the entire upside is external Foundry revenue inflecting as 18A-P / 14A win customers (14A customer decisions expected H2 2026–H1 2027 ). If Foundry's loss halves from −$10.3B toward −$5B and external revenue scales past ~$2–3B, FY2028 adjusted EPS could reach $2.00–2.50. If 18A external traction disappoints, FY2028 looks like FY2026 — sub-$1 GAAP, perpetually "next year." This binary is the whole story.

Sensitivity: at ~$118/share and 5.05B shares ($588B cap), the stock prices in something like the FY2028 bull EPS today — a ~50x multiple on a not-yet-earned number. There is little margin of safety if the Foundry option doesn't pay.

Lens 12 · Bull vs Bear

Bull case. Intel is the only Western leading-edge logic manufacturer, and the world (US government, Nvidia, SoftBank, Microsoft, Amazon, Google) has now put capital and credibility behind making it succeed. 18A is ramping with improving yields; Panther Lake/Clearwater Forest put Intel back on a competitive process for the first time since ~2018. DCAI is inflecting on AI-CPU demand (+22% rev, +142% op income YoY ). Margins are recovering (GM 34.8% FY25 → 39.4% Q1 2026 ). The cap table is now de-risked: sovereign backstop + strategic partners mean bankruptcy/dilution-spiral risk is largely off the table. If Foundry converts even two or three marquee external customers at volume, Intel re-rates from "IDM in distress" to "the second-source foundry the entire non-China supply chain needs" — a structurally scarce asset. Capital allocation under Tan is finally disciplined (asset sales, opex cuts, partner-funded capex).

Bear case (2–3 things that permanently impair or de-rate):

1. Valuation is the bear case. A barely-profitable company at ~133x forward earnings / ~8x EV-sales is priced for a turnaround that is announced but not delivered. Foundry still lost $10.3B in FY2025 and external revenue was $307M. The gap between price and proof is enormous.
2. 18A could underdeliver on yield/cost. The entire thesis is one node. If 18A/18A-P yields or costs disappoint, external customers (Apple, Nvidia, hyperscalers) stay with TSMC, and Foundry never reaches the scale economics that justify the build. Intel has slipped roadmaps before.
3. Structural demand erosion + dilution. CCG margin already fell 20% YoY; PC is mature, AI compute is going to GPUs/custom ASICs (Intel's accelerator line is a non-factor vs Nvidia). Share count is up 17% YoY and could rise further. Government ownership adds political constraints on capital returns and strategy.

Pre-mortem (18 months out, thesis broke): 18A external traction stalled (a marquee customer publicly chose TSMC's A14), a PC down-cycle hit CCG, and another Mobileye/inventory impairment landed — the GAAP losses returned, the "validation" narrative deflated, and a ~133x-forward stock compressed toward a 20-30x normalized multiple, i.e. a 50%+ drawdown even with the business intact.

Contrarian view (what the market refuses to see): the market is treating government + Nvidia + SoftBank investment as proof the turnaround will work. It is actually proof the turnaround needs external capital to be attempted — these are bets on optionality, not dividends on success. The same names also cap the equity: a government 10% holder will resist the kind of brutal capital discipline (closing fabs, slashing capex) that a pure-profit-maximizer might run.

Lens 13 · Devil's Advocate (short-seller)

Dismantling the bull case:

• Revenue concentration: 43% of revenue from three customers; lose one OEM relationship or a hyperscaler's Xeon allocation to AMD EPYC/Arm, and the products engine that funds everything cracks.
• The moat is weaker than bulls think: x86 is being attacked from both sides — AMD has out-executed Intel on the same architecture for years, and Arm (Nvidia Grace, Amazon Graviton, Ampere) is eating data-center share. Intel's AI-accelerator effort (Gaudi) is irrelevant next to Nvidia.
• Most dangerous competitor: TSMC — and the tell is that Intel is TSMC's customer. If your own foundry can't make your best chips cost-competitively, why would Apple or Nvidia trust it with theirs?
• Worst capital-allocation history: a decade of buybacks-at-highs and missed transitions; the current "fix" relies on selling minority stakes in the crown-jewel fabs (Apollo/Brookfield) and the government converting grants to equity — financial engineering, not earned cash flow.
• What must hold for $118: 18A external revenue must inflect to multiple billions by FY2028, Foundry losses must halve, GM must hold ~40%+, and the AI-compute thesis for x86 head-nodes must be real. Miss any one and the multiple is indefensible.
• If growth disappoints 20-30%: revenue $42-44B, GM back toward mid-30s, GAAP losses return — the stock has no earnings to fall back on and re-rates on book value ($110B equity, P/B already 5.3x), implying substantial downside.
• Single scenario that permanently impairs: a public, named marquee customer (e.g. Apple or a hyperscaler) chooses TSMC A14 over Intel 14A in the H2-2026/H1-2027 decision window — that would brand 18A/14A as "not good enough," collapse the external-foundry option, and leave Intel a subscale IDM carrying TSMC-scale capex. Plausibility: moderate, and it's a binary the market is currently pricing at near-zero.

Lens 14 · Management Questions (ordered by information value)

1. On 18A and 18A-P, what are the defect-density and cost-per-wafer curves versus TSMC N2/A16, and at what external volume does Foundry reach operating breakeven?
2. Which external customers have signed binding 18A/14A wafer agreements (not LOIs/PDK evaluations), and what is the committed dollar backlog and revenue-recognition timing?
3. What is the explicit plan and timeline to take Intel Foundry from a $(10.3)B operating loss to breakeven — and what external-revenue run-rate does that require?
4. How does the US government's ~10% ownership constrain capital allocation — could you close or sell a fab, cut capex sharply, or resume buybacks without political friction?
5. Given recurring multi-billion "restructuring" charges three years running, what should investors model as the normalized GAAP operating margin, and when does GAAP catch up to adjusted?
6. Will you continue using TSMC for leading-edge tiles, and on what timeline does Intel manufacture 100% of its own flagship products internally?
7. What is the residual impairment risk in the $8.2B Mobileye goodwill and the broader $23.9B goodwill balance under current assumptions?
8. How do you defend CCG margins (op income −20% YoY) against AMD and Arm-based PC silicon?
9. What is the realistic AI-accelerator (Gaudi/successor) strategy, or is DCAI's AI growth purely a host-CPU story attached to Nvidia/AMD GPUs?
10. How should we think about further equity dilution — is the current ~5.05B share count the ceiling, or will more capital be raised to fund the Ohio/Arizona build-out?
11. What are the unit economics of the Apollo/Brookfield SCIP structures, and how much of consolidated Foundry economics ultimately accrues to minorities vs Intel holders?
12. What is the China revenue trajectory under tightening export controls (already −18% YoY to $12.7B), and how much is structurally at risk?
13. What gross and operating margin does the Nvidia x86-custom-CPU partnership carry, and how material is it to FY2027-28 revenue?
14. What is the contingency plan if a marquee customer publicly selects TSMC over Intel 14A in the upcoming decision window?
15. What capital-return policy should long-term holders expect over the next three years, given the competing claims of capex, deleveraging, and the government stake?

---