INTC › analyze
INTC · Analyze from before
Date: 2026-09-22 | Price: $122.67 (Sept 21 close $121.78, +12.0%) | Market cap: $648.5B Baseline: analyze-2026-08-04.md — TRIM, conviction 4.5, fair value $55–85, written at $100.45 Price change since baseline: +22.1%. Against the baseline's own fair-value ceiling of $85: +44.3%.
0. The one-paragraph answer
The baseline's named break trigger fired, exactly as written, six days after the file was closed — and the trigger was bearish. On 2026-08-10 Intel priced an upsized $20B common stock offering at $95.00 (210,526,315 shares, closing 2026-08-12), and the underwriters' greenshoe appears to have been taken as well. The market's first verdict agreed with the baseline: INTC fell from ~$107 to $85.14 by 2026-08-24, a −20% drawdown that carried the stock into the top half of the $55–85 fair-value band. It has since risen +44% in four weeks on a foundry rumour and a sector-wide CPU rotation, neither of which is a disclosed Intel financial event. The fair value does rise — from $55–85 to $60–90 — and the reason is real: the raise retired the financing risk that was named risk #4 in the baseline, and 14A customer engagement has visibly deepened. But the arithmetic does not close. The raise cost 4.9% of dilution in a single quarter and produced a pro-forma net cash position of only ~$1.8B — it removed a liability, it did not create per-share value. Run the same bridge the baseline ran, on the new share count: complete success of the foundry turnaround is worth $64–80. Foundry breakeven plus Intel Products growing operating income another 40% is worth $88–110. The stock is $122.67. Today's price is above the value of two stacked bull cases, where in August it was above the value of one. TRIM stands, and the case for it is stronger than it was.
1. What this updates
Baseline: Output/Stocks/Semiconductors/INTC/analyze-2026-08-04.md — verdict TRIM,
conviction 4.5, fair value $55–85, entry below $55, trim 20x fwd, recheck 2026-10-22.
The file has never been refreshed. There has been no new earnings print — Q2 FY26
(reported 2026-07-23) is still the most recent, so every segment, margin and cash-flow figure
in the baseline remains the newest hard data on the name. What moved is capital structure,
price, share data and narrative.
The event list, 2026-08-04 → 2026-09-22
| Date | Event | Touches claims |
|---|---|---|
| 2026-08-10 | Intel announces a $15B common stock offering, then upsizes and prices it at $20B / $95.00 / 210,526,315 shares. 30-day option on a further 31,578,947 shares. Use of proceeds: "general corporate purposes… capital expenditures and working capital." | #9, #15, #20, #21 |
| 2026-08-12 | Offering closes. Net proceeds ~$19.7B base. BofA cuts target $160→$145; UBS cuts $121→$112. | #15, #22 |
| 2026-08-17 → 08-24 | Stock falls to $85.14 — −20% from the pre-offering level, inside the baseline fair-value band | #20, #24 |
| 2026-08-21 | Mercury Research Q2 2026 x86 share. AMD overall 34.1%, server units 34.5% (+7.3pt YoY), mobile 28.9% (+8.4pt YoY) — after Intel added mobile capacity | #13 |
| 2026-09-01 | Zinsner: 14A defect reduction is the fastest since 22nm; customer talks reported to be broadening ahead of the 2028 ramp | #3 |
| 2026-09-04 | Mizuho cuts to $92 — Neutral, post-offering, below spot | #22 |
| 2026-09-08 | Northland upgrades to Outperform, $120 (target had been suspended). UBS reported upgraded to Buy the same week (single secondary source; Yahoo's feed does not carry it) | #22 |
| 2026-09-10 | Piper Sandler initiates Neutral, $110 | #22 |
| 2026-09-15 | Tigress raises $118 → $145 | #22 |
| 2026-09-16 | SK hynix reported in exploratory talks to lease part of Intel's delayed Ohio complex, or form a JV with cloud customers, to make MEMORY chips in the US. Stock +4–5% | NEW (N1) |
| ~2026-09-18 | Tan: Intel can meet only about half of CPU demand | #13, NEW (N3) |
| 2026-09-21 | +12.0% to $121.78 on a CPU-specific rotation — Arm, AMD and INTC all double-digit on Meta's Muse AI agent | #23 |
| Announced for 2026-10-05 | ~10% PC CPU price increase — the third since late 2025, concentrated on 13th-gen Raptor Lake because AI-PC parts are selling poorly | NEW (N2) |
No new print. No signed 14A customer. No Apple or Intel confirmation. No foundry-breakeven date. The three questions the baseline handed forward are all still open, and two of them (14A commitment, Foundry external revenue) are the reason the recheck was set to 2026-10-22.
2. The break-trigger analysis — this is the centre of the pass
The baseline listed, under "What would change this verdict":
Equity issuance or a Mobileye sale to fund 2027 capex → ⬇️
and under Named Risks #4: "capex is going back above $20B in 2026 and 'significantly above' in 2027 by management's own guidance, which ends the free-cash-flow inflection and may require equity issuance." Zinsner's verbatim was "we may need to tap the capital markets."
It fired. Six days later. At the top end of the range anyone was modelling.
| Baseline expectation | What happened | |
|---|---|---|
| Form | Equity issuance or Mobileye sale | Equity. Mobileye untouched. |
| Size | Unquantified | $20B, upsized from $15B, plus a 31.6M greenshoe |
| Price | — | $95.00 — a 5.4% discount to the ~$100.45 the baseline was written at |
| Timing | "2027 is live, not hypothetical" | August 2026 — a year earlier than the baseline's own framing |
Three readings are available and the ledger records all three, because they point in different directions:
(a) The bearish reading — the one the baseline wrote. A company that cannot fund its own capex sold a fifth of a trillion dollars of stock, at a discount, three weeks after the best quarter in fifteen years. The dilution is not theoretical any more: shares outstanding went 5.043B → 5.29B in one quarter (+4.9%), and against Q2 FY25's 4.377B the base has grown +20.9% in fifteen months before the 228M DOC warrant shares that are still contracted. Every per-share number in the bull bridge got 4.7% worse.
(b) The bullish reading — and it has to be given real weight. The raise was upsized, so demand exceeded supply at $95. Pro-forma, Intel's cash and short-term investments go from $29.73B to roughly $52.3B against $50.54B of debt — from $20.8B of net debt to approximately net-cash-neutral. The FY26 >$20B and FY27 "up meaningfully" capex programmes are now pre-funded. Baseline named risk #4 and the JP Morgan Underweight thesis ("an investment cycle that could require capital raises or additional debt") are both substantially retired. That is a genuine, large improvement in durability and it deserves to move the fair value.
(c) The reading that decides it — what the market actually paid. The stock fell to $85.14. That is inside the baseline's $55–85 fair-value band, at the ceiling. The baseline's arithmetic was vindicated within fifteen trading days. What erased it was not an Intel disclosure: it was a memory-fab tenancy rumour and a sector rotation in which AMD and Arm rose as much as Intel did. The baseline's line — "you are being offered $100 by beta, not by Intel" — is now re-confirmed at $122, on stronger evidence than it had in August.
The trigger fired, the price obeyed it, and then the price stopped carrying information. That is the finding, and it is why the verdict does not soften.
3. The delta ledger
Lead with the rows that changed.
🔄 SUPERSEDED — facts changed
| # | Baseline claim | What replaced it | Forces |
|---|---|---|---|
| 9 | "+16.8% dilution in twelve months… 228M warrant shares still to come… dilution or new debt in 2027 is live, not hypothetical." | The issuance happened in August 2026, not 2027. 210,526,315 shares at $95 + a 31.6M greenshoe; shares out 5.043B → 5.29B (+4.9% in one quarter); +20.9% against Q2 FY25; fully diluted with the 228M warrant shares ≈ 5.52B, +26%. | Primary — Intel 8-K/press release 2026-08-10, corroborated by the Yahoo share-count field moving +247M. Strong. |
| 15 | Net debt $20.8B; debt/assets 25.0% and rising; "if we're super successful… we may need to tap the capital markets." | ~$22.6B of net proceeds raised. Pro-forma cash + ST investments ≈ $52.3B vs $50.54B of debt → net cash ≈ +$1.8B. The 2026–27 capex plan is funded without a further raise. This is the single most favourable change in the file. | Primary (8-K) + arithmetic. Strong. This is the row that moves the fair value up. |
| 22 | "Consensus Hold, mean target $115.27 (+14.8% upside). Not one upgrade, three cuts." | Spot is now ABOVE the consensus target. Mean target $116.37 vs $122.67 spot — i.e. −3.4% "upside." Post-offering: Mizuho cut to $92, UBS cut to $112, BofA cut $160→$145. Against that, Northland upgraded to Outperform ($120) after suspending its target in May, Tigress went $118→$145, Piper initiated Neutral $110. The ratings finally started to move — and the price still outran them. | Yahoo upgrades/downgrades feed, corroborated by press. Strong. The reported UBS "upgrade to Buy, Sept 8" is single-source and conflicts with Yahoo's Aug-12 UBS Neutral/$112 entry; flagged, not relied on. |
❌ RETRACTED — wrong when written
| # | Baseline claim | The error | Cause |
|---|---|---|---|
| 2b | "The partners it did name — SambaNova, Fortinet, Foxconn — were not characterised as wafer-foundry wins." | Fortinet was a wafer-foundry win, and was reported as such on 2026-07-21 — two days before the Q2 call and two weeks before the baseline was written. It is Intel 4's first external foundry customer, for security/firewall ASICs, and was covered as "Intel Foundry lands first named customer under CEO Lip-Bu Tan." | Reasoning, not vendor. The baseline read the earnings-call framing rather than the dedicated coverage and under-credited a real, named customer. |
| What the retraction does and does not change | It adds one named external foundry customer, on Intel 4 — a mature node, three generations behind 18A. It is a genuine credit to the "Foundry can sell wafers to strangers" claim and no credit at all to the 14A question, which is where the capex and the valuation sit. Baseline claim #1 (external revenue ~5% of Foundry revenue) and #3 (14A committed with zero signed external customers) both survive intact. |
📉 DRIFTED — moving against the thesis, and the level where it breaks
| # | Baseline claim | New evidence | Direction |
|---|---|---|---|
| 13 | "Revenue growth and market-share data point in opposite directions… reconciled by a physical CPU shortage." Mercury Q1'26: Intel x86 72.9% → 67.4%. | Mercury Q2 2026 makes this worse and removes Intel's best defence. AMD overall 34.1%, desktop 34.9%, server units 34.5% (+7.3pt YoY), and mobile 28.9%, +8.4pt YoY — after Intel added millions of units of mobile capacity in Q2. Supply relief did not restore share. The repo's own note pattern-shortage-pricing-masks-share-loss recorded its falsification condition firing on exactly this data on 2026-08-27. | Drifted against Intel. "It's capacity, not product" is no longer available as an explanation. Breaks at: x86 share below ~62% in the Q3 Mercury data, or DCAI decelerating below +20% once supply loosens. |
| 19 | Escrowed-share derivative liability $15.6B at 27 Jun 2026; GAAP EPS uninterpretable. | The liability marks higher with the stock, by design. From $100.45 → $122.67, the ~228M-share exposure implies roughly $5B of additional non-cash Q3 charge and a liability near $21B. Yahoo's EPS(ttm) of −$2.09 is this artifact, not an operating result. The DOC's in-the-money value on 241M shares at $20.00 is now ~$24.8B, up from ~$19.4B in August. | Drifted. Mechanically guaranteed to worsen while the stock rises. See pitfall-government-warrant-derivative-inverts-gaap-eps. |
🔁 REFRESHED — same claim, new number
| # | Claim | Baseline | Now |
|---|---|---|---|
| 20 | Fair value / entry / trim | $55–85 · below $55 · 20x fwd | $60–90 · below $60 · 44x fwd (§5) |
| 21a | P/S vs own 5-yr band (1.6–3.9x) | 8.88x — 2.3x the top of band | 11.4x — 2.9x the top of band, 4.2x the 2.7x mean |
| 21b | P/B | 5.79x on BVPS $17.36 | ~5.9–6.2x on pro-forma BVPS ~$19.8–20.8. ⚠️ Yahoo's 7.07x is wrong — it divides the new price by the pre-offering Q2 BVPS of $17.36 and ignores the ~$22.6B of book value the offering added. (New pitfall candidate — §7.) |
| 21c | Enterprise value | ~$561B (computed; Yahoo's $495.4B rejected) | ~$683B = $648.5B cap + $50.5B debt − $52.3B pro-forma cash + $15.6B minority interest + ~$20.7B escrowed-share liability. EV/TTM revenue 12.0x. Yahoo's $650.67B still omits the minority interest and the derivative liability — still unusable. |
| 18 | Apple foundry agreement — WSJ 2026-05-08, preliminary, unconfirmed | Still unconfirmed by either principal, and the market-moving version is now the 2026-06-18 Trump announcement that took INTC to a record $133.99 close, not the WSJ story. KeyBanc's "Apple on 18A-P for low-end M-series, production 2027" remains a sell-side assertion, not company guidance. Four months and two large rallies later, neither Apple nor Intel has said a word on the record. | The claim is more load-bearing than in August and no better evidenced. |
| 3 | 14A committed to 2028 HVM with zero signed external customers; PDK 0.9 due October 2026 | PDK 0.9 external release confirmed for October 2026. Engagement has "significantly increased" — Tan's team meeting customers weekly, discussion reportedly shifted from technical data toward capacity and supply. Zinsner: 14A defect reduction fastest since 22nm. Still zero signed. Intel's own filed risk language remains "unsuccessful to date in securing any significant external foundry customer… prospects for 14A are uncertain." | The probability improved; the evidence class did not. Priced as probability in §5, not as a customer. |
✅ CARRIED — re-tested, still true
| # | Claim | The check |
|---|---|---|
| 1 | Intel Foundry external revenue ~$293M/qtr against a ~$8.4B annualised loss — ~5% of segment revenue | No new print. Next data point 2026-10-22. |
| 4 | US government 9.9% at $20.47; DOC warrants 241M at $20.00; separate 5% warrant blocking a foundry spin-off below 51% ownership | No change. The transfer of value has grown with the price (#19). |
| 6 | Intel is a bystander in AI accelerators (~0% share) | Nothing announced. The Sept 21 rally was explicitly a CPU rotation, which is the point. |
| 7 | Evergreen: NO. Cyclical / re-rating sleeve, never evergreen | Unchanged. |
| 10 | FY26 capex >$20B; FY27 "up meaningfully"; the FCF inflection has a short shelf life | Guide unchanged. The $20B raise is the confirmation, not a contradiction — you do not sell $20B of stock to fund a capex programme you are about to shrink. |
| 11 | Q2 adjusted FCF −$8.42B vs simple FCF +$4.45B; the "turn" is definitional | No new filing. See pitfall-fcf-definition-diverges-with-jv-partner-funds. The alleged Yahoo flip did not reproduce — §7. |
| 12 | Gross margin recovery 27.5% → 40.4% is real | fin.py TTM gross margin 39%. Holds. |
| 14 | ROIC ~2–3%, at or below cost of capital for four years on ~$161B of invested capital | fin.py: ROE −11%, ROA 1%, invested capital $160.87B. Holds. |
| 16 | Q2 FY26: revenue $16.13B +25.4%, DCAI +59%, non-GAAP EPS $0.42, GAAP −$2.16 | Newest data. Unchanged. |
| 17 | Q3 guide: revenue $15.8–16.8B, non-GAAP EPS $0.38, non-GAAP GM 42.0% | Unchanged; tested 2026-10-22. |
| 23 | "You are being offered $100 by beta, not by Intel." | Re-confirmed, harder. 2026-09-21: +12.0% in a CPU-specific rotation on Meta's Muse agent — Arm and AMD moved with it, AMD approaching $1T. No Intel disclosure. |
| 24 | Beta 2.24; the name travels 20% in a week in both directions | fin.py beta 2.23. Since the baseline: $100 → $107 → $85.14 → $122.67. Holds exactly. |
| 5 | Advanced packaging (EMIB/Foveros) is the strongest remaining real asset, on a 2–4 year TSMC/Kinsus shot clock | ⏳ Not tested this pass (§7). |
| 8 | The federal 5% warrant makes a Foundry spin-off structurally near-impossible | No change. |
🆕 NEW
| # | Finding | Weight |
|---|---|---|
| N1 | SK hynix / Ohio. Reported 2026-09-16 exploratory talks for SK hynix to lease part of Intel's delayed Ohio complex, or form a JV with cloud customers, to manufacture memory chips in the US. Stock +4–5%. ⚠️ This is not an Intel Foundry logic customer, and it is not a 14A or 18A win. It is a tenancy / capacity-monetisation idea for an asset Intel has already pushed to ~2030. If signed it would convert an idle, delayed liability into rent — genuinely useful, and exactly the kind of thing that should raise a fair value modestly rather than re-rate a stock 44%. Nothing is signed. | Single-source, unsigned. Priced as probability, not revenue — principle-story-vs-revenue. |
| N2 | ~10% PC CPU price increase on 2026-10-05 — the third since late 2025, concentrated on 13th-gen Raptor Lake because AI-PC parts (Lunar Lake) are selling poorly. Supports near-term ASPs. Also an admission about mix: the pricing power is in a four-year-old node, not in the new product. | Corroborated across multiple trade outlets; Intel has not confirmed. Medium. |
| N3 | Tan: Intel can meet only about half of CPU demand. The shortage is deepening, not clearing. | Cuts both ways and the ledger records both. It defers the baseline's top monitor — DCAI normalisation — past 2027, which raises near-term earnings confidence. It also means every quarter of "growth" from here is still measuring the shortage, and Mercury Q2 shows Intel losing share anyway. |
| N4 | The "$15B more planned" figure in circulation does not reproduce. It is the initial announced size of the same offering ($15B on 2026-08-10), which was upsized to $20B before pricing. No second raise has been announced. The only outstanding issuance authority is the 31.6M-share greenshoe, which the share count suggests was already taken. | Primary (Intel newsroom / 8-K). Correcting a figure that entered this pass as an input. |
4. How the close calls were decided
(a) Does the $20B raise make the stock better or worse? — It makes the company better and the share worse, and the two do not cancel at this price.
The forces genuinely conflict. Against: 4.9% of instant dilution and a confirmation that Intel still cannot self-fund. For: a named risk retired, a capex programme pre-funded, and oversubscription at $95. The tiebreaker is arithmetic, not sentiment. The raise moved Intel from $20.8B of net debt to ~$1.8B of net cash — $0.33 per share of net cash, not $4.29. The $22.6B did not become shareholder value; it replaced borrowings and pre-paid a fab bill whose return is the entire open question. So the correct treatment is: raise the probability weight on the good scenarios (financing risk gone), and lower every scenario's EPS by 4.7% (more shares). §5 does exactly that, and the two effects nearly offset — which is why the fair value rises only ~6% while the price rose 22%.
(b) Does deeper 14A engagement earn a status change? — No, but it earns probability.
Weekly customer meetings, a conversation that has moved from "show me the data" to "how much capacity can I have", and the fastest defect-reduction curve since 22nm are all real, and they come from the CFO rather than the trade press. But Intel's own filed risk language still says it has been unsuccessful to date and that 14A prospects are uncertain, and the company has now had fourteen months to name a customer and has not. Per this command's corroboration rule, engagement described by the seller is a single-source force, not a test. Claim #3 is CARRIED, and the improvement is expressed where it belongs — in §5's probability weights, which move the breakeven scenario from ~45% to ~48% and the super-bull from ~20% to ~22%.
(c) SK hynix — why it does not change the foundry verdict.
This entered the pass framed as "SK hynix evaluating Intel Foundry." The reporting says something different and smaller: memory chips, in Ohio, by lease or JV. Intel Foundry's thesis is leading-edge logic wafers sold to fabless designers. A memory tenant in a delayed Ohio shell is a real-estate and utilisation story. It is a good one — Ohio is a stranded asset pushed to ~2030 — but it belongs against the balance sheet, not against claims #1 or #3. A single unsigned rumour that moved the stock 5% and was then extended by 12% of sector beta is precisely the shape the baseline warned about.
(d) The falsified pitfall note — does it help Intel or hurt it?
pattern-shortage-pricing-masks-share-loss had its own falsification condition fire on 2026-08-27, which sounds like a win for the bull case. It is not — it fired against Intel. The condition was "unit share stabilising while the shortage persists." What happened instead is that Intel added mobile capacity and still lost 8.4 points of mobile share year-over-year. The note was narrowed to say the guard applies to the loser in a shortage — which is Intel. This is the clearest new negative evidence in the pass, and it is the row that pulls conviction down.
(e) Where the baseline's own reasoning was put to the evidence.
The baseline concluded TRIM because "complete success of the foundry turnaround is worth $67–83, and the stock is $100.45." Put the new facts to it: is A still true? Yes — the bridge now produces $64–80 on the diluted count, essentially the same number. Is B still true? The stock is $122.67, so the gap widened from 21% to 53% above that scenario. The baseline does not need re-weighting; it needs re-pricing, and the re-pricing strengthens it.
5. Updated valuation
5.1 Measured against its own band — the decisive test, unchanged in method
| Year-end | Market cap | Revenue | P/S | BVPS | P/B |
|---|---|---|---|---|---|
| 2022 | $109.4B | $63.05B | 1.74x | $24.50 | 1.08x |
| 2023 | $212.6B | $54.23B | 3.92x | $24.96 | 2.01x |
| 2024 | $86.8B | $53.10B | 1.63x | $22.93 | 0.87x |
| 2025 | $184.3B | $52.85B | 3.49x | $22.88 | 1.61x |
| 2026-08-04 | $506.7B | $57.03B | 8.88x | $17.36 | 5.79x |
| 2026-09-22 | $648.5B | $57.03B (TTM) | 11.37x | ~$19.8–20.8 (pro-forma) | ~5.9–6.2x |
11.4x sales is 2.9x the top of a five-year band of 1.6–3.9x, and 4.2x its 2.7x mean. To return merely to the top of its own band on an optimistic FY2027 revenue of $72B implies $53/share on 5.29B shares. That number has barely moved since August, because the denominator did not move — only the price and the share count did.
⚠️ Do not use Yahoo's 7.07x P/B or its $17.36 BVPS. Both are the pre-offering Q2 figures. The $22.6B raise added roughly $4.27/share of book. Corrected pro-forma P/B is ~5.9x, and ~6.2x after the estimated Q3 escrowed-share charge. The corrected number is still about three times the top of Intel's own 0.87–2.01x historical band — so the conclusion is unchanged, but the vendor figure overstates the deterioration and would fail an audit.
5.2 Peer position
| Fwd P/E | EV/Revenue | Gross margin | Op margin | ROE | |
|---|---|---|---|---|---|
| INTC | 59.5x (Yahoo, FY27E $2.06) | 12.0x | 39% | 12% | −11% |
| AMD | ~38x | — | 53% | 14% | 8% |
| TSM | ~19x | — | 64% | 60% | 40% |
Intel's forward multiple rose from 49.3x to 59.5x since the baseline while the earnings estimate stood still. It remains the highest forward multiple of any major semiconductor manufacturer and roughly 3x TSMC's, at ~5% external foundry loading and an $8.4B annual foundry loss.
5.3 Graham — √(22.5 × EPS × BVPS)
On pro-forma BVPS of ~$20.3 (which the raise genuinely improved):
| EPS assumption | Graham IV | vs $122.67 |
|---|---|---|
| FY26E non-GAAP $1.51 | $26.3 | −79% |
| FY27E consensus $2.06 | $30.7 | −75% |
| Foundry breakeven, $3.19 | $38.2 | −69% |
| Super-bull $4.41 | $44.9 | −63% |
Weight: low, for the same reasons the baseline gave — Graham cannot price an option on a foundry and punishes a book value the escrowed-share charge is itself consuming. It is retained only for the same reason: no plausible earnings input brings it within 60% of the price.
5.4 Bogle — yield + earnings growth ± P/E change
Dividend yield 0% (suspended since Q4 2024). Starting forward P/E 59.5x. Four years.
| EPS growth | → 30x terminal | → 25x terminal |
|---|---|---|
| 20%/yr | −0.9%/yr | −4.4%/yr |
| 30%/yr | +8.7%/yr | +5.0%/yr |
| 40%/yr | +18.5%/yr | +14.5%/yr |
The multiple must now compress ~15–19% per year for four years just to reach a normal semiconductor valuation. At a very good 20% EPS CAGR the investor loses money. In August the same table needed 30%+ to be paid; it still does, and the starting multiple is worse.
5.5 DYT / DDM — N/A. No dividend, no reinstatement guided.
5.6 The bridge — the baseline's own method, on the new share count
Fully diluted shares: 5.29B outstanding + ~228M unreleased DOC warrant shares ≈ 5.52B (baseline used 5.271B — a 4.7% increase). Operating income annualised from Q2 FY26 segments.
| Scenario | Op income | Net income | EPS (5.52B sh) | @20x | @25x |
|---|---|---|---|---|---|
| Current run-rate (Foundry −$8.4B) | $12.8B | ~$10.3B | $1.87 | $37 | $47 |
| Foundry at full BREAKEVEN | $21.2B | ~$17.6B | $3.19 | $64 | $80 |
| Breakeven + Products +40% | $28.9B | ~$24.3B | $4.41 | $88 | $110 |
| Normalisation (shortage clears, DCAI +15%, Foundry −$6B) | ~$9B | ~$7.0B | $1.27 | $25 | $32 |
Add net cash of $0.33/share to each. The numbers barely move from August, because the dilution and the cash offset almost exactly.
Read the third row. In August, the price exceeded the value of one completed bull case. Today it exceeds the value of two stacked bull cases — full foundry breakeven and Intel Products growing operating income another 40% from a record quarter and a 25x terminal multiple gets you $110, and the stock is $122.67. To justify today's price you need a fourth assumption on top of the three the baseline already called "not a bull case, but three consecutive bull cases stacked."
5.7 Probability-weighted fair value
The weights move, and the reasons are named:
| Scenario | Value | Weight (Aug → now) | Why the weight moved |
|---|---|---|---|
| Normalisation / bear | ~$32 | 35% → 30% | Shortage deepening (Tan: half of demand met), Oct price increase, 3–5yr supply agreements with volume and price commitments. Normalisation is deferred past 2027. |
| Foundry breakeven | ~$72 | 45% → 48% | Financing risk retired by the raise; 14A defect curve fastest since 22nm; PDK 0.9 on schedule; Fortinet is a real (if mature-node) external customer; SK hynix Ohio optionality. |
| Breakeven + Products +40% | ~$99 | 20% → 22% | Same, plus deeper 14A engagement. |
| Weighted point estimate | ~$66 |
| Model | Weight | Output |
|---|---|---|
| Own historical P/S band (3.5–3.9x on FY27E revenue) | High | $48–53 |
| Bridge at foundry breakeven, 20–25x, diluted | High | $64–80 |
| Peer-relative (25–30x FY27E $2.06) | Medium | $52–62 |
| Probability-weighted scenarios | High | $66 |
| Graham | Low | $26–45 |
| Bogle | Context | Needs a 30%+ EPS CAGR merely to break even |
Fair value: $60–90 (baseline: $55–85)
The fair value did rise, by roughly 6%, and the reason is specific: the equity raise retired the financing risk the baseline named, and 14A engagement deepened enough to move the probability weights. It did not rise more, because the same raise diluted every per-share figure by 4.7% and the cash is committed to capex rather than to shareholders. The price rose 22% over the same period and is now 36% above the top of the widened band.
- Accumulate below $60.
- Trim: 44x forward earnings — which renders ~$90, the top of the band, on Yahoo's FY27E $2.06. ⚠️ State the basis every time. 44x sounds generous because $2.06 is the most optimistic forward estimate in the range (Wedbush FY27E was $1.53; the FY26E build from actuals + guide is $1.51). The same $90 ceiling is 60x on the FY26E basis. The multiple is set against the basis the site's price parser uses; it is not a claim that a no-free-cash-flow cyclical deserves 44x.
- Above $90 the stock is discounting outcomes that have not happened. It is at $122.67.
6. Thesis persistence, conviction delta, and the verdict
Thesis persistence
| Count | |
|---|---|
| CARRIED | 13 |
| REFRESHED | 6 |
| DRIFTED | 2 |
| SUPERSEDED | 3 |
| RETRACTED | 1 (partial — claim 2b) |
| UNTESTED | 1 |
| NEW | 4 |
Structural + Trend claims surviving as CARRIED or REFRESHED: 15 of 18 = 83%. That is high persistence — and note what it means in this specific case. High persistence against a +22% price move is normally the re-rating setup: the business held and the multiple moved. Here the business held and the multiple moved — in the direction that makes the stock worse. Nothing in the baseline's bear case broke. One row (share loss) got worse on new primary data. One row (financing) got materially better. Everything else is where it was, six weeks older, and 22% more expensive.
Conviction: 4.5 → 4.0
Named rows, per the command's rule that an unattributable move is not made:
| Direction | Rows | Effect |
|---|---|---|
| ⬆️ | #15 SUPERSEDED — net debt $20.8B → net cash ~$1.8B; 2026–27 capex pre-funded; JP Morgan's core bear point retired | +0.4 |
| ⬆️ | #3 REFRESHED — 14A defect reduction fastest since 22nm, PDK 0.9 on schedule for October, engagement moved to capacity discussions | +0.2 |
| ⬆️ | #2b RETRACTED (partial) — Fortinet is a genuine named external foundry customer on Intel 4 | +0.1 |
| ⬇️ | #13 DRIFTED — Mercury Q2: Intel added mobile capacity and still lost 8.4pt of mobile share. "It's capacity, not product" is gone as a defence | −0.5 |
| ⬇️ | #9 SUPERSEDED — +4.9% dilution in one quarter, +20.9% in fifteen months, ~26% fully diluted, and the company has demonstrated it will issue into strength | −0.4 |
| ⬇️ | #19 DRIFTED + #22 SUPERSEDED — the Treasury's claim grew to ~$24.8B; spot now sits above the consensus target with a post-offering cut to $92 on the tape | −0.3 |
Net −0.5. Conviction 4.0. This is a conviction score on the name, not on the verdict. Intel is more solvent and no more competitive than it was in August.
Verdict: TRIM — unchanged in direction, stronger in degree.
1. Is this a good business? Better financed than in August, and no better competitively. The credit is real and larger than it was: the balance sheet is fixed for this capex cycle, 18A is in high-volume manufacturing ahead of TSMC's N2, 14A defect reduction is the fastest since 22nm, PDK 0.9 ships in October, and Fortinet is a real named external customer. Against that: consolidated operating margin 12%, ROE −11%, ROIC ~2–3%, adjusted FCF −$8.42B last quarter, external foundry revenue ~$293M against a $2.09B quarterly foundry loss, and unit share still falling in every segment even where Intel added capacity. Grade: C+, and the improvement this quarter was financial, not competitive.
2. Has the market already priced it in? It has priced in materially more than in August, on materially less news. 11.4x sales against a 1.6–3.9x own band. 59.5x forward earnings — 3x TSMC. Spot is above the consensus price target. And the arithmetic that settles it: foundry breakeven is worth $64–80, breakeven plus 40% Products growth is worth $88–110, and the stock is $122.67.
Four-box test: a C+ business at a considerably worse price than the one the baseline already called too expensive. The trim recommendation from 2026-08-04 stands and applies with more force — and the interval since has shown it can be executed: the stock traded at $85.14 on 2026-08-24, inside the old fair-value band, five weeks ago.
What would change this verdict
| Trigger | Direction |
|---|---|
| A signed 14A tier-1 customer with disclosed volume — PDK 0.9 lands October 2026, decisions guided H2 2026 → H1 2027 | ⬆️ Materially. Still the missing evidence, now fourteen months after the 2025 condition was set |
| Apple or Intel confirming the foundry agreement on the record | ⬆️ A large part of two rallies rests on a deal neither principal has ever acknowledged |
| Intel Foundry external revenue > $1B in a quarter (now $293M), or the quarterly loss narrowing below −$1.0B | ⬆️ Materially |
| SK hynix Ohio lease or JV actually signed, with disclosed rent or capital | ⬆️ Converts a stranded, delayed asset into cash flow |
| Advanced packaging (EMIB) reaching disclosed multi-$B annual revenue | ⬆️ The most credible near-term upside |
| Price below $60 with the run-rate intact | ⬆️ To ACCUMULATE |
| x86 unit share below ~62% in the Q3 Mercury data | ⬇️ Confirms erosion is product, not capacity — the Q2 data already points this way |
| DCAI decelerating below +20% as the shortage clears | ⬇️ Still the top monitor, now deferred past 2027 by Tan's "half of demand" comment |
| A further equity raise, or the FY27 capex number rising again | ⬇️ The company has now demonstrated it will issue into strength at a discount |
| Terafab volume slipping past 2027 | ⬇️ |
| TSMC's Kinsus EMIB-equivalent reaching production | ⬇️ |
🎯 The falsification test — what would prove this verdict wrong
This verdict is wrong if the foundry stops being an option and becomes a business before the multiple compresses. Concretely, any two of the following by the 2027-Q1 report would break it and force a re-rate to HOLD or better:
- A named tier-1 14A customer with disclosed volume or a disclosed prepayment — not "engagement", not an analyst's channel check, not a Presidential announcement. Intel or the customer, on the record.
- Intel Foundry external revenue above $1.0B in a single quarter (3.4x the current $293M), or the Foundry operating loss below −$1.0B/quarter (currently −$2.09B). Either would put the breakeven scenario on a guidable path rather than an analyst's assumption.
- x86 unit share stabilising at or above 67% in two consecutive Mercury quarters while the shortage persists — which would mean the share loss really was capacity, and the Q2 mobile data was an artifact of a single quarter's allocation.
- DCAI growth above +30% in a quarter where server CPU ASPs are flat or falling — growth that survives the removal of shortage pricing.
And the single cleanest disproof: FY2027 adjusted free cash flow guided positive. The entire bear case is that Intel spends more than it earns and funds the difference by issuing stock. A guided positive adjusted FCF for 2027, with capex still above $20B, would falsify it outright.
Conversely, this verdict is confirmed if the 2026-10-22 print shows Foundry external revenue below ~$400M, the Foundry loss still worse than −$1.8B, and no named 14A customer — in which case the stock at $122.67 is paying 11.4x sales for a fourteen-month-old promise.
7. What this pass did NOT test
Recorded so the next run is honest about what it inherits.
- ⏳ Claim #5 — advanced packaging (EMIB/Foveros). The baseline called it the strongest remaining real asset, with a 2–4 year shot clock from TSMC's Kinsus equivalent. Not re-tested. No new EMIB revenue disclosure, no update on the Kinsus timeline. This is the second consecutive pass in which the most credible near-term upside on the name went unexamined — flag it as a priority for 2026-10-22.
- ⏳ Governance (baseline §4.7). The baseline explicitly flagged the Reuters 2026-07-29 related-party story (Intel supplying chip technology to a startup led by a Tan co-investor) "for follow-up before the 2026-10-22 recheck." It was not followed up. A prior UNTESTED that stays UNTESTED across two passes is exactly what this command is supposed to surface.
- ⏳ Terafab. No progress check. Small-batch AI5 production was guided "late 2026" — the window is open now and nothing was verified.
- ⏳ Mobileye. Still ~77% owned, ~$11.5B, the last large monetisable asset. Not re-valued.
- ⏳ 18A yields. Intel still publishes nothing. Unchanged and unknowable.
- ⚠️ Single-source conclusions flagged: the SK hynix Ohio talks (N1); the ~10% October CPU price increase (trade press, not Intel); the reported UBS upgrade to Buy on 2026-09-08, which conflicts with Yahoo's own feed showing UBS at Neutral/$112 on 2026-08-12 and is not relied on in §5.
- ⚠️ The greenshoe is inferred, not confirmed. Shares outstanding moved 5.043B → 5.29B (+247M) against a base offering of 210.5M, which is consistent with the 31.6M option being exercised plus routine RSU vesting. Intel has not been seen to confirm it. If the option was not taken, the dilution is ~0.6pt smaller than modelled — immaterial to the verdict.
Data quality — two vendor findings
| Item | Vendor | Correction | Status |
|---|---|---|---|
| P/B and BVPS | Yahoo: 7.07x on BVPS $17.36 | $17.36 is the pre-offering Q2 figure. The $22.6B raise added ~$4.27/share of book. Pro-forma BVPS ~$19.8–20.8 → P/B ~5.9–6.2x | 🚩 New trap. After any large primary equity offering, vendor BVPS/P/B lags until the next 10-Q, and the error runs in whichever direction the raise went. Pitfall-note candidate. |
| Enterprise value | Yahoo: $650.67B | Omits the $15.6B minority interest and the ~$20.7B escrowed-share derivative liability; understates by ~$33B. Computed EV ~$683B | 🚩 Still unusable, as the baseline found |
| EPS(ttm) −$2.09 / no P/E | Yahoo | The escrowed-share mark-to-market, not an operating result. See pitfall-government-warrant-derivative-inverts-gaap-eps | ✅ Known trap, correctly handled |
| "Yahoo FCF flipped +$4.87B → −$4.95B for the same fiscal year" | Reported as an input to this pass | Did not reproduce. fin.py returns FY2025 FCF = −$4.95B today, and the baseline's own table recorded FY2025 = −$4.95B on 2026-08-04. The two figures that differ are FY2025 (−$4.95B) and TTM-through-Q2 (+$2.83B) — a period difference, not a vendor flip. | ⚪ Not a vendor error. The real, documented trap on this name is definitional: pitfall-fcf-definition-diverges-with-jv-partner-funds. Intel's own adjusted FCF of −$8.42B remains the number to cite, per principle-primary-source-beats-vendor. |
| Consensus "Recommendation: buy" | Yahoo | Yahoo's aggregate key moved to "buy" while its own mean target ($116.37) sits below spot and the underlying distribution (51 analysts, avg ~$108) is Hold | ⚠️ Use the target and the distribution, not the key |
Differential re-analysis by Financebot — Fundamentals · Moat · Valuation · Sentiment · Portfolio Strategist. Baseline TRIM 4.5 @ $100.45 → TRIM 4.0 @ $122.67. Fair value $55–85 → $60–90. 13 carried · 6 refreshed · 2 drifted · 3 superseded · 1 partially retracted · 1 untested · 4 new. Recheck 2026-10-22 (Q3 FY26 print, date not yet confirmed by Intel).