Intel INTC

Reference table

1. What the business is, and what the vendor file gets wrong about it

Intel is an integrated device manufacturer: it designs chips and it owns the fabs. Since the 2024 reorganisation it reports those as separate segments with internal transfer pricing "intended to approximate market pricing," so that, in the 10-Q's words, "the Intel Products operating segments and Intel Foundry are meant to reflect separate fabless semiconductor and foundry companies, respectively."

The reportable segments as of Q2'26, per the 10-Q, are:

• Intel Products, comprising two operating segments: CCPG (Client Computing and Physical AI Group — renamed from Client Computing Group; PCs and edge devices) and DCAI (Data Center and Artificial Intelligence — x86 server CPUs, accelerators, NICs, IPUs and custom ASICs).
• Intel Foundry — technology development, manufacturing, assembly, test and advanced packaging. "At present, substantially all of our Intel Foundry business supports internal manufacturing for Intel Products."
• All Other (not a reportable segment) — Mobileye, IMS, start-ups, and historical Altera results.

The vendor's seg_prod block does not describe this company, and it is the most misleading single field in the file. For FY2025 it reports "Client Computing Group $32.228B, Data Center Group $16.919B, Intel Foundry Services $17.826B, Other $3.563B, Intersegment Eliminations −$17.683B." The arithmetic sums correctly to the reported $52.853B, so the totals are not wrong — but three of the labels are:

• "Intel Foundry Services" is not a segment. Intel Foundry is, and the $17.826B is the whole segment including intersegment sales to Intel Products. External foundry revenue was $53M in the first half of 2025 and $467M in the first half of 2026. A reader taking the vendor field at face value would conclude Intel has an $17.8B third-party foundry business. It does not.
• "Data Center Group" has been DCAI since 2024; "Client Computing Group" became CCPG in 2026.
• The FY2024 row contains an "Internet Of Things Group" of $5.842B — a segment Intel does not report.

Segment results from the filings, which supersede the vendor entirely:

Q1'26 figures are derived by subtracting the 10-Q's stated Q2 figures from its stated half-year figures and are labelled as such. The gap between segment operating income and consolidated operating income is corporate unallocated expenses — restructuring, share-based compensation and acquisition-related costs, which are deliberately excluded from segment results — plus intersegment eliminations.

Geography. seg_geo for FY2025 lists United States $15.757B, Singapore $9.535B, Taiwan $7.672B, Other Countries $7.195B — summing to $40.159B against reported revenue of $52.853B, a $12.694B hole. China, which was $15.532B in FY2024, has vanished from the FY2025 row entirely. Either it was reclassified into "Other Countries" (which rose only $611M) or the row is incomplete. Treated as unreliable and used for nothing.

Non-equity tripwire — checked and passed. INTC is common stock on NASDAQ: price $100.94 (not par-like), beta 2.187 (not bond-like), 117.1M shares traded today, a 52-week range of $19.50 to $140.94 (a 623% band), and no fixed distribution — the dividend has been suspended since FY2025. This is unambiguously the common equity.

2. The single most important thing to understand: Intel now loses money when its stock goes up

Q2'26 operating income was PLUS $1.981B. Q2'26 net income was MINUS $11.033B. The bridge is one item.

On 22 August 2025 Intel entered into a Warrant and Common Stock Agreement with the Department of Commerce. In exchange for the acceleration of $5.7B of remaining CHIPS Act Direct Funding Agreement disbursements and $3.2B of Secure Enclave disbursements, Intel agreed to issue the government up to 433 million shares — 275 million on closing and 159 million into escrow, releasable at a $20.00 per share rate as Secure Enclave money arrives — plus warrants over a further 241 million shares at $20.00, exercisable only if Intel ceases to own at least 51% of its foundry business.

The common stock and the warrants sit in permanent equity. The escrowed shares do not. After an SEC accounting consultation that concluded in Q4'25 and forced a revision, Intel classifies them as a derivative liability, remeasured to fair value each quarter through interest and other, net. The 10-Q states the consequence in plain terms:

> "In Q2 2026 ... we recognized a $12.5 billion loss related to the net change in the fair value of shares of our common stock released from escrow during the quarter and remaining in escrow at the end of Q2 2026, driven by an increase in our stock price. The fair value of the related derivative liability was $15.6 billion at the end of Q2 2026."

The liability was $2.7B at 27 December 2025 and $15.6B at 27 June 2026. For all of FY2025 the mark cost $1.8B. In the first half of 2026 alone it cost $13.6B.

What this does to every trailing metric:

Three further consequences that are easy to miss.

First, the loss is genuinely non-cash and genuinely real at the same time. No cash left the company. But 143 million escrowed shares are still outstanding and will be issued at a $20.00 reference price against $3.2B of disbursements; 71 million of them are already in the basic share count. Half of any that remain unreleased at the end of the eligibility period go to the government for nothing, and half are cancelled.

Second, it runs in reverse. If the shares fall to $60, the derivative liability shrinks and Intel will report a multi-billion-dollar gain. A trailing price-to-earnings ratio computed on such a quarter would look spectacular and would mean exactly as little.

Third, it is a hard constraint on the most obvious value-unlocking transaction. The 241 million warrants at $20.00 become exercisable if Intel ceases to own at least 51% of its foundry business. Spinning or selling down the foundry writes the U.S. government an option on 4.7% of the company at a fifth of today's price. Any "separate Products from Foundry and re-rate" thesis has to pay that toll, and no vendor field anywhere in this file mentions it.

Therefore: this dive values Intel on segment operating economics, on forward non-GAAP earnings estimates, and on sum-of-parts. GAAP net income and every ratio derived from it is reported for completeness and used for nothing.

3. Growth and margin trajectory — real, and lower-quality than it looks

The revenue reacceleration is genuine and it is the best in five years.

Five consecutive quarters of sequential gross-margin improvement from 27.5% to 41.5%, and a return to positive operating income. This is a real operational recovery and it should not be understated.

But the composition matters, and the 10-Q is unusually candid about it:

> "These increases were primarily driven by ASP increases, the majority of which resulted from a higher mix of premium products sold, with demand-based pricing actions contributing to a lesser extent, in part to offset higher input costs."

and

> "In Q2 2026 and YTD 2026, market demand exceeded our available product supply due to capacity constraints at our factories and industry-wide supply constraints. We expect these industry-wide shortages of substrates, memory and other critical components to persist into next year."

That is price and mix in a shortage, not units. The knowledge base independently corroborates both the fact and its fragility from the same speaker on the same day (Section 7): Intel and AMD are "fully sold out on CPUs, have sent customers price-increase notices, and are no longer competing on price" (conviction 70) — and "the CPU shortage gives Intel short-term profits and they'll do better, but it doesn't save the company long-term" (conviction 50).

Cost reduction is doing real work. R&D and MG&A were $4.5B in Q2'26, down 6% year over year, and 28.2% of revenue against 37.5% a year earlier. But R&D specifically has been cut from $17.528B (FY2022) to $13.774B (FY2025), a 21.4% reduction, at a company that is behind on process technology. Headcount is 85,100. Two restructuring plans, from 2024 and 2025, are substantially complete. Some of the margin improvement is therefore borrowed from the future — a semiconductor company that intends to lead on process cannot indefinitely spend less on it.

A vendor-versus-filing note on gross profit. The vendor reports Q2'26 gross profit of $6.694B, which against Q2'25's $3.542B is +89.0%. The 10-Q says gross profit increased "$3.0 billion, or 84%." That implies roughly $6.5B, a $177M difference. The filing wins; we treat the 41.5% gross margin as approximate and describe it as "roughly 40–41%."

4. The Foundry question — and why the improvement is not what it appears

Intel Foundry's operating loss is the single most important number in this company, because roughly half the share price depends on it going to zero and then positive.

The headline reads well: a $1.1B improvement year over year in the quarter, $1.0B in the half. The 10-Q's own decomposition says otherwise.

> "Operating loss was $4.5 billion in YTD 2026, compared to an operating loss of $5.5 billion in YTD 2025, primarily driven by $1.8 billion of lower period charges, including the absence of $797 million of non-cash asset impairment and accelerated depreciation charges recognized in Q2 2025 ... as well as $391 million of lower inventory charges ... These benefits were partially offset by $830 million of lower product profit in YTD 2026, driven by an increased mix of higher-cost wafers manufactured on our Intel 18A process node."

Read the arithmetic. The entire improvement, and more, is the non-recurrence of charges taken a year ago. The underlying manufacturing profit went BACKWARDS by $830M, because the leading-edge node the whole thesis rests on costs more to run than the nodes it replaces. That is normal for an early ramp and it is exactly what you would expect. It is also the opposite of "the losses are narrowing on schedule."

The external customer question is worse still. External foundry revenue was $293M in Q2'26, up $271M year on year. The 10-Q attributes the increase "primarily due to Altera's transition to an external customer following the deconsolidation of Altera in Q3 2025" — that is, a business Intel itself owned until 12 September 2025 and now sells wafers to. The same filing discloses that total revenue from Altera as a customer was $181M in Q2'26. So genuinely third-party, non-Altera foundry revenue is at most about $112M in the quarter — call it $450M annualised against a segment losing roughly $9B a year.

And the FY2025 10-K, filed 23 January 2026, states the position without hedging:

> "We have been unsuccessful to date in securing any significant external foundry customers for any of our nodes and our prospects for securing a significant external foundry customer for Intel 14A are uncertain."

> "If we are unable to secure a significant external customer for our Intel 14A node, we may pause or discontinue development of Intel 14A and subsequent next generation leading-edge nodes ... any decision to pause or discontinue our pursuit of Intel 14A and successor leading-edge process technologies may be effectively irreversible."

Six months later the 10-Q records a material change, and it is the strongest single piece of good news in this file:

> "During Q2 2026, we committed to completing development of Intel 14A, with a number of future Intel products designed to utilize the node and manufacturing expansion projects underway for production of products on the node. We also made continued progress towards meeting performance and design milestones for potential significant customers to evaluate Intel 14A for their future products. We intend to accelerate various of our manufacturing expansion projects."

That is a real reversal — from "we may discontinue" to "we committed to completing" — and it is management's own framing, so it carries half weight. It is also carefully hedged in the very next clause: "though the scale and pace of our manufacturing expansion projects will ultimately be dictated by the amount of committed demand for Intel 14A that we are able to obtain." No customer is named. And the cash tells a different story from the language: the 10-Q states that first-half 2026 capital expenditure was LOWER than the first half of 2025. A company genuinely accelerating fab construction spends more, not less.

Process status, from the filings: 18A entered high-volume production at the start of 2026 (Intel Core Ultra Series 3), introducing gate-all-around transistors and backside power delivery; 18A-P entered risk production in June 2026; 14A is in development; EMIB-T adoption scales from 2026, Foveros-B and Foveros-R target high-volume manufacturing in 2027, Foveros Direct hybrid bonding on 18A-PT in 2028. A 12nm mature-node platform with UMC is expected to begin production in 2027.

5. Balance sheet, share count, and the enterprise-value rebuild

The most recent complete balance sheet available in this file is FY2025 (2025-12-27). The extracted 10-Q text contains the narrative but not the financial statement tables, so the Q2'26 balance sheet cannot be rebuilt from the filing. This is stated as a gap, not estimated around.

At 2025-12-27, per bal_a:

• Cash and equivalents $14.265B · short-term investments $23.151B · total cash and short-term investments $37.416B
• Short-term debt $2.499B · long-term debt $44.086B · capital lease obligations $0 · total debt $46.585B (the components sum correctly; there is no double-count here, unlike the known vendor failure mode)
• Total assets $211.429B · total liabilities $85.069B · total equity $114.281B (including non-controlling interests) · goodwill $23.912B · intangibles $2.772B

Defect 1 — the vendor reports net debt of $32.320B; the correct figure is $9.169B; we use $9.169B. The vendor's $32.320B is total debt ($46.585B) less cash and equivalents ($14.265B) and nothing else — it discards the entire $23.151B of short-term investments. Both the 10-K and the 10-Q define the company's liquidity as "total cash and short-term investments", and the 10-Q confirms Intel funded the Apollo transaction "through a combination of existing cash and cash equivalents, short-term investments and a $6.5 billion term loan facility." The vendor's convention overstates net debt by $23.151B, or 252%.

Defect 2 — the vendor's enterprise value of $546.804B implies net debt of $37.663B, which reconciles to nothing in the filings, and it omits a $15.6B liability that does exist. $546.804B less the $509.141B market capitalisation is $37.663B. That is neither the FY2025 reported net debt ($32.320B) nor the corrected figure ($9.169B). It appears to be a Q2'26 figure computed on the same defective convention: Q2'26 total debt should be approximately $46.585B + $6.5B of new senior notes − $1.5B of February notes settled − $1.0B of May notes settled = ~$50.6B, which less an implied $12.9B of cash and equivalents gives $37.7B. Short-term investments are excluded again.

Meanwhile the vendor omits the $15.6B escrowed-shares derivative liability entirely — a real, quantified, filing-disclosed claim payable in Intel stock.

So the vendor's enterprise value is wrong in two large and roughly offsetting ways. Correcting only the short-term-investment omission would reduce EV by perhaps $10–23B (the Q2'26 short-term investment balance is not disclosed in the extracted text; it was $23.151B at year-end and was partially consumed by the $14.2B Apollo payment). Adding the escrowed-shares liability would increase it by $15.6B. The headline figure of ~$547B may therefore be approximately right by accident, while both of its components are wrong. Every EV-derived multiple in the file — evToSalesTTM 9.59x, evToEBITDATTM 148.9x, evToFreeCashFlowTTM 193.1x, netDebtToEBITDATTM 10.26x, enterpriseValueMultipleTTM 148.9x — is rejected, both because the inputs are unreliable and because EBITDA is not a coherent measure for a company whose largest income-statement item is a mark on its own equity.

Defect 3 — non-controlling interests are material and the vendor handles them inconsistently. bookValuePerShareTTM is $20.208 while shareholdersEquityPerShareTTM is $17.152 — a gap of ~$15.4B on 5.104B shares, which is the minority interest in Arizona SCIP (Brookfield) and Mobileye. Price to book is 4.99x on total equity and 5.89x on Intel-only equity. We use 5.89x, the Intel-only figure, because that is what a common shareholder owns. The vendor's own priceToBookRatioTTM of 5.885 confirms this is the basis it uses for that specific field.

The Apollo transaction — the second-largest event of the quarter. On 8 April 2026 Intel reacquired Apollo's 49% minority interest in Ireland SCIP (Fab 34) for $14.2B cash, funded from existing cash, short-term investments and a $6.5B 364-day term loan at 4.79% that was repaid in Q2 from a $6.5B senior-note issue (4.65% 2031, 5.00% 2033, 5.30% 2036, 6.13% 2056, 6.20% 2066). The accounting is worth noting: only $142M of non-controlling interest was eliminated and a $532M derivative liability extinguished; the residual $13.5B was charged directly to capital in excess of par value. It does not appear in the income statement, but it is $13.5B of shareholders' equity spent to buy back an interest in a single fab — and the coupons on the replacement debt, up to 6.20%, tell you what the market charges Intel for 40-year money.

Arizona SCIP remains outstanding, with construction-in-progress of $6.0B and in-service assets of $13.8B on Intel's balance sheet that "are not available to us as they can be used only to settle obligations of the VIE." Intel is contractually required to operate it at minimum production levels or pay volume-related damages.

Share count — the dilution nobody discusses.

Up 19.3% from FY2024. The sources are documented: 275M shares to the Department of Commerce, 87M to SoftBank at $23.00, 215M to Nvidia at $23.28, and 16M of escrowed shares released at a $20.00 reference. The 2026 proxy records 5,021,010,228 shares outstanding at the 16 March record date.

Still outstanding and in no vendor share count: 143M escrowed shares (2.8%) and 241M warrants at $20.00 (4.7%). Fully released and exercised, that is another 7.5% of dilution — struck at a fifth of today's price.

The honest framing: the U.S. government, SoftBank and Nvidia bought a fifth of this company at $20.00–$23.28 in the second half of 2025. It is $100.94 today. Existing shareholders financed a recapitalisation at prices that, in hindsight, were extraordinarily generous to the incoming holders. That is not a criticism of the transactions — Intel needed the money and, per the knowledge base, needed roughly $50B of it — but it is a large part of why the equity has quintupled, and it is not repeatable.

6. Cash flow and capital returns — four years of burn

$89.390B of capital expenditure over four years, during which revenue fell from $63.054B to $52.853B and cumulative free cash flow was minus $44.501B. That is the long-run fact against which every forward projection has to be argued.

FY2025 is materially better, with the burn down to −$4.949B, and capex at $14.646B is now only 0.978x depreciation — essentially maintenance level. freeCashFlowYieldTTM of 0.556% and priceToFreeCashFlowRatioTTM of 179.8x are arithmetically correct on a trailing basis and economically uninformative; they are reported, not used.

The tension worth flagging. Capex has fallen to maintenance, and the 10-Q says first-half 2026 capex was lower still — yet management says it "intends to accelerate various of our manufacturing expansion projects." Committed capital expenditure was $9.1B for 2026 with $3.7B committed long-term as of the FY2025 balance sheet date, plus $2.2B of other 2026 purchase obligations and $4.5B long-term. Either the acceleration has not started, or the guidance and the cash flow are describing different futures. Watch this in the Q3 print.

Capital returns are zero. No buyback in any of the four years. The dividend was suspended in FY2025 ($1.599B paid in FY2024, $0 in FY2025; dividendYieldTTM is 0). profile.lastDividend of 0.5 is a historical artefact and dividendPayoutRatioTTM of −1.270 is meaningless — both rejected.

Government incentives distort the picture in ways worth naming. Of FY2025's total capital-related incentives, $6.1B was recognised as a non-cash investing activity ($3.3B in FY2024, $1.1B in FY2023), some collectible in cash and some settleable as tax credits. $2.3B of CHIPS Act incentives had been received under grant accounting before the August 2025 agreement; everything received under Secure Enclave since is equity, not grant. Reported capex is therefore net of substantial government contribution, and the gross investment in the fab base is larger than the cash-flow line shows. The 10-Q notes first-half 2026 saw lower proceeds from capital-related government incentives than the prior year.

Also relevant to tax: a $9.9B valuation allowance was first established against U.S. deferred tax assets in Q3'24, and both filings confirm Intel "was not able to benefit from our current year domestic loss before taxes due to the domestic valuation allowance." Intel pays little domestic cash tax and gets no book benefit from domestic losses. effectiveTaxRateTTM of −13.6% and taxBurdenTTM of 1.146 are artefacts of this and are not usable.

7. Valuation — priced in or room?

At $100.94 (market cap $509.141B, ~5.104B diluted shares):

Analyst counts in italics. Estimates are non-GAAP; reported EPS is GAAP. The gap is enormous and structural here — FY2025 consensus EPS was $0.345 against GAAP of −$0.059, and FY2026 will be far wider because of the derivative mark. **All forward multiples in this dive are struck on the non-GAAP epsAvg basis and are not comparable to the GAAP line.

est.ebitdaAvg and est.ebitAvg are corrupt across the entire forward strip and are rejected. FY2027E shows EBITDA $18.143B, EBIT $4.281B and net income $10.727B; FY2028E shows $20.515B / $4.841B / $15.718B; FY2029E shows $24.011B / $5.666B / $20.703B. Net income cannot be three to four times EBIT at a company with ~$50B of debt. Forward valuation runs on epsAvg, exactly as the data contract prescribes.

Estimate coverage note. The near years are well covered (21–32 analysts) and the far years are not: FY2029 and FY2030 EPS rest on six analysts each. The FY2030 revenue estimate of $122.306B implies +29.2% growth on FY2029 — the fastest year in the entire forecast, six years out, from a six-analyst sample. No conclusion in this dive rests on FY2029 or FY2030 except to demonstrate what today's price requires.

Peer context. The vendor peer set is only partly useful: ADI ($185.2B), AMAT ($434.0B), APH ($210.8B), ARM ($299.6B), KLAC ($255.3B), LRCX ($397.4B), NOW ($122.1B), QCOM ($170.8B), SONY ($131.2B), TXN ($259.0B). No peer multiples are supplied, so no peer-multiple comparison is drawn. But two capitalisations are used as scale references below, and both are from this batch's own price table: KLAC $255.3B and LRCX $397.4B.

7a. What today's price assumes (the inversion)

At $100.94 — 50.2x FY2027 consensus non-GAAP EPS, 8.93x trailing sales and 5.89x book on a business with a 0.14% trailing operating margin — the price embeds roughly the following falsifiable claims:

• Intel Foundry is worth approximately $258B. (Our arithmetic, from the payload and the filings.) Value Intel Products at a fabless multiple — take the midpoint of FY2025's actual $12.7B and H1'26's annualised $17.8B of segment operating income, call it $15B, tax at 15% to $12.75B, apply 22x for a low-growth but currently pricing-powerful x86 franchise: $280B. Add $10B for Mobileye, IMS and the rest. Deduct estimated net debt of ~$23B and the $15.6B escrowed-shares derivative: $251B, or $49.3 a share. The market pays $100.94. The residual — $258B, or half the share price — is Intel Foundry, a segment that lost $2.1B last quarter and has at most ~$112M a quarter of non-Altera external revenue. For scale, $258B is more than KLA Corporation's entire $255.3B market capitalisation.
• The full FY2030 consensus is delivered, and the market still pays 25x for it. (Consensus-derived, 6 analysts.) FY2030E EPS of $6.42 at a 25x multiple is $160.50 in late 2030; discounted 4.4 years at 11% that is $100.8 — today's price to within a dollar. At 20x it is $80.6. The price is not discounting the 2030 case; it IS the 2030 case at a premium multiple.
• The current CPU shortage is a permanent repricing rather than a cycle. (Filing-derived.) Q2'26 growth was ASP and mix while "market demand exceeded our available product supply"; consensus models revenue rising from $52.853B (FY2025) to $71.541B (FY2027), +35% in two years, on a company whose unit volumes are supply-constrained and whose all-time revenue peak is $79.024B (FY2021).
• The escrowed-shares liability stops mattering. (Our reading.) At $15.6B and growing with the share price, it is a claim that no screen counts and that grows precisely when the equity is doing well.
• The most fragile assumption, named as such: that a significant external customer signs for Intel 14A. The January 10-K says Intel has never secured one on any node in four years and that failing to secure one may make 14A uneconomic, with the decision to stop being "effectively irreversible." Roughly half of today's price rests on a sentence in the filings that says the opposite.

7b. The return bridge (why the multiple moves)

Expected return over 12–24 months decomposes as: non-GAAP EPS growth (consensus +36% from FY2026E $1.476 to FY2027E $2.012) + multiple drift + shareholder yield (ZERO — no dividend, no buyback, and a share count growing 19% in eighteen months, so the yield is NEGATIVE).

Our fair value assumes multiple COMPRESSION, and we say so plainly. At $68 the base pays 33.8x FY2027E rather than the 50.2x on offer today. The reason is not that the earnings are worse than consensus — we use consensus — but that the multiple currently being paid is not a multiple on earnings at all; it is a price on the foundry option, and we mark that option at $96B rather than $258B.

Note that forward-multiple rolldown is not compression: 68.4x FY2026E becoming 50.2x FY2027E at an unchanged price is mechanical. The compression we assume is a genuine re-rating of the foundry option, which requires either an adverse 14A development or simply the passage of time without a customer. Conversely, the bull case at $105 requires no compression at all — it requires the market to keep paying today's multiple, which is why the bull case and the current price are almost the same number.

Most of the potential return here is not earnings growth and not yield. It is re-rating. And with the street target at $110.41 — 9.4% above spot after one session — there is very little of it left on the table even if you agree with the crowd.

7c. Variant perception (where we differ, what would surprise)

• We differ from the street on what the foundry is worth, and that is the whole disagreement. The consensus target of $110.41 implicitly marks Intel Foundry at roughly $305B. We mark it at $96B in the base case. We are not more pessimistic than the street about Intel Products — we use consensus revenue and consensus EPS. Watchable number: external foundry revenue excluding Altera, currently at most ~$112M a quarter.
• We differ on the direction of the foundry trend. The market read "loss narrows from $3.2B to $2.1B" as progress. The 10-Q says $1.8B of the half-year improvement was lower period charges and $830M was negative product profit. On the underlying line the foundry got worse. Watchable number: whether the Q3'26 10-Q attributes any improvement to product profit rather than charges.
• We differ on whether today's move is information. Nothing in the payload or in any filing explains a 10.92% session. Q2'26 was reported twelve days earlier and the stock was $90.04 a week after it (per an insider tax-withholding Form 4) and $91.00 the day before this dive. The market did not reward the beat; it rewarded the sector, and Intel's 2.187 beta did the rest.
• Where we do NOT differ: the CPU shortage is real, pricing power is real, DCAI at +59% is real, 18A shipping in volume is real, and the 14A commitment in Q2'26 is a genuine and material improvement on the January language. We are not arguing Intel is broken. We are arguing it is priced as though it is already fixed.
• Positive surprise that would force a re-rate upward: a named significant external customer committing volume to 14A. This is binary and it is the only thing that matters. Secondary: foundry product profit turning positive; capex actually accelerating in the Q3 cash-flow statement; DCAI holding above +40% growth once the shortage eases.
• Negative surprise that would break the thesis further: the shortage normalising and ASPs giving back the gains; a 14A milestone slip; a further Mobileye impairment (the 10-Q discloses a 1% higher discount rate would have cost another $682M against $4.3B of remaining goodwill); or — the risk that is genuinely uninsurable — damage to the Israeli fab. The 10-Q states that in late March 2026 Iran published a target list of U.S. companies with Middle East operations "with Intel being near the top of that list", that a substantial portion of current and anticipated revenue comes from Intel 7 products made there, and that "we are not insured for business interruptions resulting from war or political violence" and the assets are self-insured.

Synthos fair values

All three anchors are sum-of-parts built on Intel Products segment operating income, cross-checked against forward consensus non-GAAP EPS. Shares outstanding 5.104B.

• Bear ~$42 — Products at 18x post-tax through-cycle segment income ($230B) + $8B other − $39B of net debt and derivative liability + $15B for Intel Foundry. Cross-check: 20.9x FY27E EPS, 14.1x FY28E, 2.4x Intel-only book. The scenario: the shortage normalises through 2027, ASPs give back, 14A fails to land a customer and the option is written off. −58%.
• Base ~$68 — Products at 22x post-tax through-cycle segment income ($280B) + $10B other − $39B of net claims + $96B for Intel Foundry. Cross-check: 33.8x FY27E EPS, 22.8x FY28E, 15.9x FY29E, 4.0x Intel-only book, 6.1x trailing sales. Note that this base case still values a loss-making foundry with no external customer at $96B — roughly a third of ARM's market capitalisation. It is not a pessimistic number. −32.6%.
• Bull ~$105 — Products at 26x ($331B) + $12B other − $39B + $212B for Intel Foundry. Cross-check: 52.2x FY27E EPS, 35.1x FY28E, and — the cleanest version — the full FY2030 consensus EPS of $6.42 at 25x, discounted 4.4 years at 11%, gives $100.8. The scenario: a significant external customer signs for 14A, the shortage persists through 2027, and the market keeps paying a strategic-asset multiple. +4%.

Base is 32.6% BELOW spot. The bull case is the current price. The asymmetry is roughly 14:1 against (58% down, 4% up). That is an Avoid, and it would be an Avoid even if we agreed with every operational claim management makes.

8. Knowledge base — the heaviest triage in the batch

Search: INTC Intel foundry "Pat Gelsinger" "Lip-Bu Tan" across 51,928 distilled claims. The search header reports 132 term matches; 103 distinct claim records were returned. 23 are used. 80 are discarded or demoted. This is by a wide margin the dirtiest raw result set in the batch, and the triage is the point.

How the 103 break down:

The match: text discards deserve specific mention because two distinct collisions are at work. First, the ordinary one: "Intel" invoked as a cautionary analogy inside claims about other companies — "margins get crushed like Intel" (about Nvidia), "unless you innovate you become Intel", "ARM is a little screwed too". Second, and more insidious: the word "intel" meaning intelligence. Two 2026 claims matched on "DoD and intel agencies" and "proprietary on-the-ground information (intel)" — neither has anything to do with the company. Likewise "foundry" collides with Azure AI Foundry, Palantir Foundry and Diamond Foundry, which between them account for most of the 22 "foundry" term matches. None of these are name-level conviction on Intel Corporation and none is used.

"Pat Gelsinger" returned zero direct term matches; "Lip-Bu Tan" returned one.

The structural-bear lane — the deepest and oldest, and it is specific about mechanism.

> 2025-02-04 · bearish · conviction 70 · speaker Dylan Patel (independent) · entities: INTC

> "Intel led manufacturing 20+ years but its culture has broken — mismanaged Fabs and designs, said no to iPhone; it can't get leading-edge to high-yield last-mile production despite still-brilliant PhDs in Oregon."

Repeated by the same speaker on 2025-02-05 (conviction 63) and 2025-02-11 (conviction 55), the latter adding "even as its 50,000 engineers remain excellent." Three separate statements of the same mechanism from the same well-informed independent voice: the talent is fine, the last mile is not. That claim is precisely what the Q2'26 filing shows in the foundry line — 18A is shipping, and it costs $830M more in product profit than the nodes it replaces.

> 2025-09-30 · bearish · conviction 80 · source all_in · entity: Intel

> "Intel missed mobile and under-invested in EUV a decade ago; once behind in chips, the compounding cycle makes recovery very difficult."

> 2024-12-06 · bearish · conviction 62 · speaker Ian King (journalist), channel lisa_su_mgmt · entity: INTC

> "Intel took roughly five years off and made bad decisions; after the worst share decline in ~40 years and 15,000 job cuts, it might never catch up again."

(Attribution note: the speaker is journalist Ian King; the channel is a management-voice lane. We attribute to the speaker and note the channel separately, per house policy.)

> 2025-10-16 · bearish · conviction 70 · source invest_like_the_best · entity: Intel

> "America's apex manufacturers — Boeing, Intel, Detroit/Tesla — are mostly declining, evidence the US isn't relearning hard manufacturing."

> 2026-01-09 · bearish · conviction 55 · source business_breakdowns · entity: Intel

> "Intel announced layoffs escalating to ~2400 people in Oregon amid disappearing state tax revenue."

The constructive lane — cyclical and policy-driven, and notably more recent.

> 2026-04-07 · bullish · conviction 70 · speaker Dylan Patel (independent) · entities: INTC, AMD

> "Intel and AMD are fully sold out on CPUs, have sent customers price-increase notices, and are no longer competing on price — just maximizing how many they can make and sell."

> Test: "Intel/AMD CPU prices keep rising near-term"

This is the single most valuable claim in the lane and it is independently confirmed by the 10-Q filed three months later ("ASP increases ... demand-based pricing actions ... market demand exceeded our available product supply"). It is also, from the same speaker on the same day, explicitly bounded:

> 2026-04-07 · neutral · conviction 50 · speaker Dylan Patel (independent) · entity: INTC

> "The CPU shortage gives Intel short-term profits and they'll do better, but it doesn't save the company long-term — AMD, Amazon and others catch up on capacity."

> 2026-01-05 · neutral · conviction 45 · speaker Gavin Baker (independent) · entity: INTC

> "Mixed emotions on the US taking a 10% Intel stake, but calling Intel a national asset matters; Lip-Bu Tan understands what's needed and this administration will pressure US chipmakers to route business to Intel."

An earlier version of the same view (2025-10-22, conviction 35) adds "Lip-Bu is a legendary operator and Gelsinger was treated unfairly — if Intel gets its act together some Nvidia chips could be made there." Note the conviction levels: 35 and 45. Even the constructive lane is held weakly.

> 2025-09-22 · neutral · conviction 55 · speaker Dylan Patel (independent) · entity: INTC

> "Intel needs ~$50B; the $5B Nvidia, $2B SoftBank, $10B government stakes are still relatively small but should boost investor confidence for a larger capital-markets raise."

> 2025-09-20 · bullish · conviction 75 · speaker Jordi Visser (independent), channel anthony_pompliano · entities: INTC, NVDA

> "The Intel-Nvidia deal (plus Tesla-Samsung) signals inference rising and the arrival of NPUs for humanoids, cars and on-device thinking — the analog/physical world meeting the cloud."

(Attribution: speaker Visser, channel Pompliano. Attributed to Visser; the channel carries low weight and the distinction is made explicitly.)

> 2025-09-17 · bullish · conviction 80 · source all_in · entities: TSMC, Intel

> "Tariffs are reshoring chip manufacturing to the US; TSMC investing hundreds of billions, Intel deal puts government balance sheet to work."

One match: text hit is retained as supporting colour because it is genuinely about Intel's foundry demand rather than an analogy:

> 2026-04-07 · bullish · conviction 70 · speaker Dylan Patel (independent) · entity tagged: TSM · match: text

> "AI chips are buying up all TSMC 3nm and 2nm capacity, squeezing Apple, Qualcomm and MediaTek off leading-edge nodes and toward Intel foundry."

What the lane establishes. The structural claims are bearish, older, more numerous and more mechanistic; the constructive claims are newer, cyclical or political, and held at markedly lower conviction (35, 45, 50, 55). Both are visible in the filings simultaneously — a booming Products segment on shortage pricing, and a foundry whose underlying profit is deteriorating with no external customer after four years. The knowledge base and the 10-Q are telling the same story: the near term is genuinely good and the durable thing has not been fixed. Net conviction: negative-moderate. The disagreement between the lanes is about duration, and the price is paying for the durable version.

Why Moderate-Negative rather than strongly negative. Eleven source lanes is good breadth. But the deepest and most specific voice, Dylan Patel, appears on both sides — bearish on the franchise in early 2025, bullish on the cycle in April 2026 — and no claim in the entire set addresses Intel at $100.94, or indeed at any price above roughly $25. The knowledge base has a view on the business. It has no view on this valuation. That limitation is stated rather than papered over.

9. Data integrity — what we rejected from the vendor file and why

Twelve material items. This section is the point of the exercise, and Intel is the name where it matters most.

1. netDebt of $32.320B (FY2025) omits $23.151B of short-term investments — the vendor reports $32.320B, the balance sheet supports $9.169B, we use $9.169B. Total debt of $46.585B less cash of $14.265B and short-term investments of $23.151B. Both filings define liquidity as "total cash and short-term investments", and the 10-Q confirms Intel funded the $14.2B Apollo payment from "existing cash and cash equivalents, short-term investments and a $6.5 billion term loan." Overstated by 252%.

2. enterpriseValueTTM of $546.804B implies net debt of $37.663B, which reconciles to no figure in any filing — and it omits the $15.6B escrowed-shares derivative liability that the 10-Q discloses. The implied figure appears to be Q2'26 total debt (~$50.6B after $6.5B of issuance and $2.5B of settlements) less cash and equivalents only, excluding short-term investments again. Two errors in opposite directions that partially cancel. All EV multiples — evToSalesTTM 9.59x, evToEBITDATTM 148.9x, evToFreeCashFlowTTM 193.1x, evToOperatingCashFlowTTM 36.6x, netDebtToEBITDATTM 10.26x, enterpriseValueMultipleTTM 148.9x — rejected.

3. Every trailing profitability ratio is a measurement of the share price, not the business — REJECTED as a class. Q2'26 net income of −$11.033B against operating income of +$1.981B is bridged by a $12.5B non-cash mark on the escrowed-shares derivative liability, which the 10-Q says was "driven by an increase in our stock price." Rejected: priceToEarningsRatioTTM −47.84x, priceToEarningsDilutedRatioTTM −47.84x, returnOnEquityTTM −10.76%, returnOnAssetsTTM −5.58%, returnOnTangibleAssetsTTM −6.29%, netProfitMarginTTM −19.79%, pretaxProfitMarginTTM −17.28%, earningsYieldTTM −2.19%, incomeQualityTTM −1.380, priceToEarningsGrowthRatioTTM 0.0040, forwardPriceToEarningsGrowthRatioTTM −1.316.

4. ratios_ttm contains three mutually inconsistent margin figures. ebitMarginTTM −15.27%, operatingProfitMarginTTM +0.137%, ebitdaMarginTTM +6.44%. The first has the derivative loss leaking into "EBIT". Only the segment operating figures from the 10-Q are used.

5. est.ebitdaAvg and est.ebitAvg are corrupt across the ENTIRE forward strip — REJECTED. FY2027E: EBITDA $18.143B, EBIT $4.281B, net income $10.727B. FY2028E: $20.515B / $4.841B / $15.718B. FY2029E: $24.011B / $5.666B / $20.703B. FY2030E: $31.017B / $7.319B / $31.176B — net income equal to EBITDA. Net income cannot exceed EBIT by 3–4x at a company with ~$50B of debt and ~$12B of annual depreciation. All forward valuation runs on epsAvg.

6. seg_prod labels do not describe the company — the vendor reports "Intel Foundry Services $17.826B" for FY2025; the 10-Q shows external foundry revenue of $467M for the first half of 2026 and $53M for the first half of 2025; we use the filing. This is the most materially misleading field in the file: a reader would infer a $17.8B third-party foundry business. The $17.826B is the whole Intel Foundry segment including intersegment sales to Intel Products. Additionally, "Data Center Group" has been DCAI since 2024, "Client Computing Group" became CCPG in 2026, and the FY2024 row contains an "Internet Of Things Group" of $5.842B — a segment Intel does not report. The totals reconcile to reported revenue; the labels do not describe reality.

7. seg_geo FY2025 is incomplete — China has vanished and $12.694B is unaccounted for. The FY2025 row sums to $40.159B against reported revenue of $52.853B. China (including Hong Kong) was $15.532B in FY2024 and does not appear in FY2025; "Other Countries" rose only $611M. Rejected as unreliable; no geographic conclusion is drawn.

8. Vendor Q2'26 gross profit of $6.694B implies +89.0% year over year; the 10-Q says gross profit rose "$3.0 billion, or 84%" — a ~$177M difference; the filing wins. The 41.5% gross margin is described in this dive as "roughly 40–41%."

9. profile.lastDividend of 0.5 contradicts dividendYieldTTM of 0 and a FY2025 cash-flow statement showing no dividend paid — REJECTED. The dividend was suspended in FY2025 ($1.599B paid in FY2024). dividendPayoutRatioTTM of −1.270 is meaningless and is rejected.

10. tech and quote disagree on the 52-week range. tech.hi52 $140.94 vs quote.yearHigh $142.35; tech.lo52 $19.50 vs quote.yearLow $19.60. We use the tech block throughout, for consistency with the moving averages, drawdown and relative-return figures computed from the same series.

11. interestCoverageRatioTTM of 0.068x is a trailing artefact of the same distortion. On the 10-Q's Q2'26 operating income of $1.981B, coverage in the current quarter is not 0.068x. Coverage is genuinely thin — the new senior notes carry coupons up to 6.20% — but the specific figure is not a statement about Q2'26 and is not used. grahamNumberTTM is null and grahamNetNetTTM is −$11.81; both noted, neither usable.

12. Two structural claims on the equity appear in NO vendor field. 143 million escrowed shares (2.8% of the count) not yet released, and warrants over 241 million shares (4.7%) at $20.00 exercisable if Intel ceases to own 51% of its foundry. Neither is in any share count, book-value-per-share figure or enterprise-value calculation in this file. They are sourced from the 10-K and 10-Q and used explicitly.

Two gaps in the source material, stated rather than estimated around.

• The 8-K dated 2026-07-23 contains only the cover page. Exhibit 99.1 — the Q2'26 earnings press release carrying the reconciliations and the Q3'26 outlook — is not in the extracted text. This dive therefore contains no company revenue, margin or EPS guidance of any kind. Earnings-call transcripts are unavailable on the current data plan, so management Q&A is also absent.
• The financial-statement tables are absent from the extracted 10-K and 10-Q text. Only narrative figures survive. The Q2'26 balance sheet therefore cannot be rebuilt from the filing; the most recent complete balance sheet available is FY2025 (2025-12-27), and the Q2'26 net-debt and short-term-investment positions are estimated with the estimation labelled. The FY2025 full-year Intel Foundry operating loss is also not recoverable from the extracted text; only the quarterly and half-year figures for 2025 and 2026 are.

Not defects, correctly reported: longTermDebt ($44.086B) and capitalLeaseObligations ($0) are not double-counted in totalDebt ($46.585B) — the known vendor failure mode does not occur here. capexToDepreciationTTM of 0.978 is consistent with the cash-flow statement. stockBasedCompensationToRevenueTTM of 4.20% ties to $2.434B of SBC on $52.853B of FY2025 revenue.

10. Technicals — and what today's move does to the entry

• Price $100.94, +10.92% on the day from a previous close of $91.00. Opened $95.32, ranged $94.32–$101.375, closed near the high on 117.1M shares.
• −28.4% from the 52-week high of $140.94, and +417.6% above the 52-week low of $19.50.
• Still 9.8% BELOW the 50-day moving average of $111.90, even after a 10.92% session. +49.3% above the 200-DMA of $67.63.
• RSI 48.4 — neutral, not overbought. MACD −6.17 — negative.
• Relative performance: 12-month +422.7% vs SPY +24.3% and QQQ +30.8% — a ~400-point outperformance. 6-month +106.8% vs SPY +11.1%. 3-month +5.4% vs SPY +7.6% — the only window in which it lags.
• Beta 2.187, the highest in the peer set.

Today's move and what it does to the entry

Nothing in the payload or in any filing explains a company-specific reason for today's 10.92% move, and the honest answer is to say so.

Q2'26 was reported on 2026-07-23, twelve days ago. The market did not reward it. An insider Form 4 records shares withheld for tax at $90.04 on 2026-07-30, a week after the print; the stock closed at $91.00 on 2026-08-03. There is no 8-K after 2026-07-23. Between the earnings beat and yesterday's close, the stock went nowhere.

What did happen today was a violent sector session across the whole batch: MRVL +12.81%, INTC +10.92%, LRCX +7.85%, MU +7.62%, AMD +7.00%, KLAC +6.95%, AVGO +6.6%, AMAT +5.48%, ASML +4.22%. The group average is roughly +7.4%. INTC at +10.92% is about 1.47x the group — which is very close to what a 2.187 beta produces on a sector move of that size. AMD, which reports after today's close and which shares the x86 pricing thesis Dylan Patel describes, rose 7.00% into its own print. The most parsimonious explanation, and the only one the data supports, is that today's move in Intel is sector beta plus x86 sentiment, not information about Intel.

Three things follow for the entry.

First, the crowd's headroom was consumed in one session. The street's price-target consensus is $110.41 — 9.4% above spot. Before today it was 21.3% above. A single day removed more than half the street's implied upside without a single estimate changing.

Second — and this is the one point in Intel's favour — this is NOT a gap to new highs. The stock is still 28.4% below its 52-week high and 9.8% below its 50-day moving average, with a negative MACD. It is a bounce inside a drawdown, not a breakout. A reader who liked this business would be buying a stock that is cheaper than it was in June, not chasing an all-time high. That is a materially better technical setup than several other names in this batch.

Third, none of that changes the answer, because the objection here is not technical. Our base fair value of $68 is 32.6% below spot; it was 25.3% below yesterday's $91.00 and it would still be below spot at $60. This is not a name we would buy on a retest; it is a name whose valuation we cannot reconcile at any price near the current one. We are therefore explicitly not asking the reader to wait for a pullback and then buy. We are asking them not to own it. If the stock round-trips to $91 or to $80, the analysis does not change — only at roughly $68 does it become a Watch, and only with evidence on 14A does it become anything more.

The dispersion in the street's own targets is the fairest summary of the situation: high $200, low $60, against a spot of $100.94. A 3.3x spread between the extremes is not a consensus; it is a coin flip with a mean attached.

11. Insiders — compensation mechanics only, and one useful price marker

All eight transactions in the file are equity-compensation mechanics. There is not a single open-market purchase and not a single open-market sale.

The reading, stated carefully. Restricted-stock vests and the automatic share withholding that funds the tax on them are not decisions. They tell you nothing about what management thinks of the price. What is notable is the absence on both sides: in a twelve-month period in which the stock rose 422.7%, no insider in this file sold a single share on the open market — and none bought one either.

The one genuinely useful datum is the withholding prices, which are actual market prints on specific dates: $109.82 on 2026-06-01 and $90.04 on 2026-07-30. Together with today's $100.94 and yesterday's $91.00, they establish that this stock has traded between roughly $90 and $110 for the last two months and that today's close sits squarely in the middle of that range, not at a breakout.

Note also that the officer whose vest is most recent carries the title GM Foundry — the segment on which the entire valuation turns — and holds 224,121 shares. That is a real but not large personal stake in the outcome.

12. Verdict, kill criteria and flip conditions

Avoid.

Let us be precise about what is being said and what is not. Intel's operating business is better than it has been in three years. Revenue grew 25.5% year over year in Q2'26, the best print since FY2021. DCAI grew 59%. Gross margin has improved in five consecutive quarters from 27.5% to roughly 41%. Intel Products earned $4.8B of segment operating income in the quarter and $8.9B in the half. Non-GAAP EPS of $0.42 doubled the $0.21 estimate. 18A is in high-volume production, 18A-P is in risk production, and in Q2'26 the company committed to completing 14A, reversing the January 10-K's contemplation of abandoning it. The U.S. government owns 275M shares, Nvidia owns 215M and is co-developing products, and the free cash burn has fallen from −$15.656B to −$4.949B. None of that is in dispute.

What is in dispute is the price. At $100.94 the market capitalisation is $509.141B. Value Intel Products generously — 22x post-tax segment operating income taken at the midpoint of FY2025 actual and H1'26 annualised — add the rest, deduct the debt and the $15.6B derivative liability, and you get roughly $251B. The residual $258B is Intel Foundry: a segment that lost $2.1B last quarter, whose underlying product profit fell $830M year on year, whose genuinely third-party revenue is at most ~$112M a quarter, and about which the company's own annual report says it "has been unsuccessful to date in securing any significant external foundry customers for any of our nodes." $258B is more than KLA Corporation's entire market capitalisation.

Our base fair value of $68 still ascribes $96B to that foundry. It is 32.6% below spot. The bear case is $42. The bull case, $105, is essentially the current price and requires the full six-analyst FY2030 consensus to be delivered and re-rated at 25x. The asymmetry is roughly 14:1 against.

And the crowd has just caught up. The street target of $110.41 is 9.4% above spot after one session that moved the stock 10.92% on no company-specific news. Even a reader who disagrees with everything above has almost no consensus upside left.

What you would be accepting. A 2.187 beta; negative GAAP earnings that are structurally negative because the company books a loss whenever its own shares rise; four consecutive years of negative free cash flow totalling −$44.501B; 19.3% dilution in eighteen months executed at $20.00–$23.28; R&D cut 21.4% since FY2022 at a company that is behind on process; no dividend, no buyback, negative shareholder yield; a C- vendor quality rating with an overall score of 1; and an Israeli fab that Iran has publicly named as a target and that is self-insured for war damage.

Pre-registered KILL criteria — what would confirm this call:

• Intel Foundry product profit failing to turn positive by the FY2027 close, or a Q3/Q4'26 10-Q that again attributes loss narrowing to charge non-recurrence rather than product profit.
• A 14A milestone slip, or explicit language in a filing walking back the Q2'26 "committed to completing" statement. The January 10-K's own words: such a decision "may be effectively irreversible."
• ASPs normalising as the substrate and memory shortages ease — the 10-Q expects them to persist only "into next year." Watch DCAI and CCPG revenue growth decomposed into price and units.
• Capital expenditure failing to accelerate despite the stated intention, or accelerating without a committed external customer — the second is worse than the first.
• Any further Mobileye impairment against the remaining $4.3B of goodwill.
• The escrowed-shares derivative liability exceeding $20B, which would mean the stock has risen further and the reported loss with it.

Pre-registered FLIP conditions — what would make us wrong, in order of importance:

• A named significant external customer committing volume to Intel 14A. This is binary, it is the entire thesis, and it would justify a complete re-underwriting rather than an adjustment. It would most likely arrive as an 8-K.
• Intel Foundry product profit turning positive on the underlying line, disclosed as such in an MD&A, for two consecutive quarters.
• External foundry revenue excluding Altera exceeding roughly $500M a quarter — the first evidence in four years of a real third-party business.
• A price near $68 with the operating trajectory intact, at which point this becomes a Watch rather than an Avoid.
• Resolution of the warrant obstacle to a foundry separation — an amendment or waiver of the 241M-share warrant condition would remove a structural barrier to the single most obvious value-unlocking transaction.

Where INTC fits in the Synthos Framework Portfolio: nowhere, at this price. Suggested weight 0%. For readers who want exposure to the same buildout, this batch offers three genuinely different expressions and Intel is the one where the least of what you pay for is currently working: AMD is the x86 share-gainer with the same pricing tailwind and none of the fab burden; the equipment names (LRCX, KLAC, AMAT, ASML) sell to every foundry regardless of which one wins; and MRVL is the custom-silicon expression. Intel is the only one of the group whose valuation requires a specific, named, four-years-overdue event that its own annual report says has never happened. Logged as a tracked Synthos call (Avoid) as of 2026-08-04 at $100.94.

Single biggest risk to owning it: that there is never a significant external foundry customer. Roughly half the share price rests on the FY2025 10-K's sentence — "we have been unsuccessful to date in securing any significant external foundry customers for any of our nodes and our prospects for securing a significant external foundry customer for Intel 14A are uncertain" — being superseded. If it is not, the foundry is a $9B-a-year cost centre attached to a good x86 business, and the equity is worth what Intel Products is worth.

Most fragile assumption in OUR view: that the market will ever again value Intel on segment economics rather than on national-security scarcity. The U.S. government owns 8.6% of this company outright, holds escrowed shares and warrants over 7.5% more, and has an explicit contractual interest in Intel retaining at least 51% of its foundry. A strategic asset with a sovereign shareholder can trade above any private-market valuation for a very long time, and the knowledge base's constructive lane is built on exactly that observation — "calling Intel a national asset matters" and "this administration will pressure US chipmakers to route business to Intel." If that is the regime, an Avoid at $100.94 will look wrong for years before it looks right. We accept that risk explicitly. We are not willing to underwrite $258B of foundry value on it.

Provenance & disclosures

• Traceability: 103 distinct knowledge-base claim records returned from a search of INTC Intel foundry "Pat Gelsinger" "Lip-Bu Tan" across 51,928 records (search header reports 132 term matches). 23 used as name-level conviction; 80 discarded or demoted. Breakdown: 64 match: entity and 39 match: text; of the entity hits, 5 named a different "Foundry" (Foundry the company ×3, Azure AI Foundry, Diamond Foundry) and were discarded; 21 pre-dated June 2024 and were demoted as stale relative to a company that has since changed chief executive, taken a U.S. government equity holder, sold 51% of Altera and shipped 18A; 11 were Jensen Huang's own voice (competitor chief executive — Intel is named but the claim argues for GPUs) and were demoted to colour; 6 were topically about something else despite the entity tag (private credit, meme-regime financing, EU semiconductor sovereignty, Japan–China export controls, enterprise-software de-rating). All 39 text matches were discarded except one, retained as supporting colour. The text-match collisions were of two kinds: "Intel" as a cautionary analogy in claims about other companies ("margins get crushed like Intel", "unless you innovate you become Intel"), and the word "intel" meaning intelligence ("DoD and intel agencies", "proprietary on-the-ground information (intel)"). "foundry" collided with Azure AI Foundry, Palantir Foundry and Diamond Foundry. "Pat Gelsinger" returned zero direct matches; "Lip-Bu Tan" returned one. Sources used (11 lanes): dylan_patel, gavin_baker, jordi_visser, all_in, invest_like_the_best, business_breakdowns, no_priors, odd_lots, chinatalk, plus two channel-attributed claims where speaker and channel differ — Ian King (journalist) on a management-voice channel, and Dylan Patel on the bill_gurley channel. Where speaker and thinker differ, attribution is to the speaker and the channel is noted separately, per the four-lane policy. Net conviction: negative-moderate — the structural claims are bearish and mechanistic, the constructive claims are cyclical and policy-driven and are held at notably low conviction (35, 45, 50, 55). No claim in the set addresses Intel at any price above roughly $25. All quotes are verbatim from stored claim text.
• Data as-of: income statement 2026-06-27 (Q2'26, reported 2026-07-23, twelve days before this dive; 10-Q filed 2026-07-24) · balance sheet and cash flow 2025-12-27 (FY2025) — no Q2'26 balance sheet is available (see gaps below) · estimates 2026-08-04 · prices 2026-08-04 (quote timestamp 1785873601 = 2026-08-04T20:00:01Z; close $100.94, +10.92%; 50-DMA $111.90; 200-DMA $67.63; RSI 48.4; MACD −6.17) · knowledge-base claims 2026-08-04. Filings used: 10-K filed 2026-01-23 (FY2025), 10-Q filed 2026-04-24 (Q1'26), 10-Q filed 2026-07-24 (Q2'26), 8-Ks 2026-04-30, 2026-05-15 and 2026-07-23. No figure comes from memory, recall or external retrieval.
• Data rejections (detailed in Section 9): netDebt corrected from $32.320B to $9.169B (omits $23.151B of short-term investments); enterpriseValueTTM of $546.804B flagged as reconciling to no filing figure and as omitting the $15.6B escrowed-shares derivative liability, with all EV multiples rejected; all trailing profitability ratios rejected as measurements of the share price rather than the business (P/E −47.84x, ROE −10.76%, ROA −5.58%, net margin −19.79%, earnings yield −2.19%, income quality −1.380, PEG 0.0040, forward PEG −1.316); three mutually inconsistent margin fields flagged (EBIT −15.27% vs operating +0.137% vs EBITDA +6.44%); est.ebitdaAvg/ebitAvg rejected across the entire FY2026–FY2030 strip as arithmetically impossible; seg_prod labels rejected as not describing the current segment structure, with "Intel Foundry Services $17.826B" identified as the single most misleading field in the file; seg_geo FY2025 rejected as incomplete ($12.694B unaccounted, China missing); Q2'26 gross profit reconciled to the 10-Q's "84%" rather than the vendor's implied 89%; lastDividend 0.5 and dividendPayoutRatioTTM −1.270 rejected against a suspended dividend; tech 52-week range used in preference to quote; interestCoverageRatioTTM 0.068x flagged as a trailing artefact; 143M escrowed shares and 241M warrants at $20.00 noted as absent from every vendor field.
• Basis note — the most important one in this dive: Intel's GAAP net income is dominated by a non-cash mark on a derivative liability representing its own shares held in escrow for the U.S. government, which rises when the share price rises. Q2'26: a $12.5B loss, taking a $1.981B operating profit to an $11.033B net loss. First half 2026: $13.6B. FY2025: $1.8B. Liability $2.7B at 2025-12-27, $15.6B at 2026-06-27. All GAAP-earnings-derived metrics are rejected. This dive values Intel on segment operating economics, on non-GAAP forward consensus (epsAvg), and on sum-of-parts. Consensus estimates are non-GAAP; reported EPS is GAAP; the two are not comparable and every forward multiple here is on the non-GAAP basis.
• Estimate coverage: strong in the near years (21 analysts on FY2026 EPS, 26 on FY2027, 27–32 on revenue) and thin in the far years (10 analysts on FY2028 EPS, 6 on FY2029, 6 on FY2030). The FY2030 revenue estimate of $122.306B implies +29.2% growth on FY2029, the fastest year in the whole forecast, six years out. No conclusion rests on FY2029 or FY2030 except to demonstrate what the current price requires.
• Peer note: the vendor peer set (ADI, AMAT, APH, ARM, KLAC, LRCX, NOW, QCOM, SONY, TXN) contains no true integrated-device-manufacturer comparable, and no peer multiples are supplied, so no peer-multiple comparison is drawn. KLAC ($255.3B) and LRCX ($397.4B) are used only as scale references for the implied foundry valuation, from this batch's own price table.
• Fair-value caveat: the $42 / $68 / $105 anchors are sum-of-parts arithmetic — Intel Products segment operating income taken at the midpoint of FY2025 actual ($12.7B) and H1'26 annualised ($17.8B), tax-effected at 15%, at 18x / 22x / 26x, plus $8–12B for Mobileye, IMS and other, less estimated net debt of ~$23B and the $15.6B escrowed-shares derivative liability, with the residual being the implied value of Intel Foundry ($15B / $96B / $212B). Cross-checked against consensus non-GAAP EPS at 20.9x / 33.8x / 52.2x FY2027E. Stated arithmetic, not a discounted cash flow. The Q2'26 net-debt input is estimated because the Q2'26 balance sheet is not available; the estimation is labelled throughout. The base case still values a loss-making foundry with no external customer at $96B and is not a pessimistic construction.
• Gaps, stated rather than estimated around: (1) The 8-K dated 2026-07-23 contains only the cover page — Exhibit 99.1, carrying the Q2'26 press release and the Q3'26 outlook, is absent, so this dive contains no company guidance of any kind; (2) earnings-call transcripts are unavailable on the current data plan, so management Q&A is absent; (3) the financial-statement tables are absent from the extracted 10-K and 10-Q text, so the Q2'26 balance sheet cannot be rebuilt and the FY2025 full-year Intel Foundry operating loss is not recoverable — only the quarterly and half-year figures for 2025 and 2026 are; (4) seg_geo FY2025 is incomplete.
• Half-weight management framing: the Q2'26 commitment to completing 14A, the stated intention to accelerate manufacturing expansion, and the expectation that industry-wide shortages persist "into next year" are all management's own statements in the 10-Q and are discounted accordingly and labelled as such wherever used. The knowledge-base claim sourced from a management-voice channel (2024-12-06) is attributed to its actual speaker, a journalist, with the channel noted separately.
• Timing: Q2'26 was reported 2026-07-23, twelve days before this dive, and the 10-Q was filed 2026-07-24 — the most recent quarter is fully reflected. The next print is 2026-10-22, 79 days away. Today's 10.92% move occurred twelve days after the print, on a sector-wide session, with no intervening company filing.
• Not investment advice. Independent research, educational and informational only, never personalised. No recommendation to buy, sell or hold any security is made to any person.
• Version: 2026-08-04-full.