Intel CEO buys shares 32% cheaper as Foundry posts $1.85 billion Q3 loss

Lip-Bu Tan's second reported 2026 family-trust purchase came after weaker guidance and a share-price selloff. Intel's filings show most Foundry revenue is internal, making outside customers and cash returns the measures to watch.

On November 2, after Intel's October 29 outlook and October 30 selloff, chief executive Lip-Bu Tan bought 150,000 Intel shares through a family trust. It was his second roughly $10 million purchase since August, this time at a price almost 32% lower. Intel still needs to turn manufacturing progress into external customers and cash returns.

What the second purchase shows

Intel's November 4 Form 4 reports transaction code P and a weighted-average price of $64.7312, with executions between $64.705 and $64.735. Multiplying 150,000 shares by that price yields $9,709,680, excluding unreported transaction costs. The Rule 10b5-1 box was unchecked, which establishes only that the purchase was not reported as executed under such a trading plan.

In August, Tan's trust bought 105,263 shares at $95 for $9,999,985. Together the two purchases total 255,263 shares for $19,709,665; November's price was 31.9% below August's. The earlier purchase did not identify a near-term low. The evidence supports repeated buying more clearly than successful market timing.

The trust held 1,464,669 shares after November's transaction; the filing separately lists 16,471 directly owned shares and 500 in a 401(k), totaling 1,481,640 across those reported holdings. The form does not disclose the trust's source of funds or Tan's reasons for buying.

Foundry's sales are mostly to Intel itself

Intel's third-quarter 10-Q reports $6.697 billion of Foundry segment revenue, up 58%, and a $1.853 billion operating loss. Only $411 million was external revenue. Dividing $411 million by $6.697 billion gives 6.1% external revenue, with approximately $6.286 billion internal. Intel eliminated about $6.3 billion of intersegment sales on consolidation, so segment revenues cannot simply be added to derive group revenue.

Intel attributed most of the increase in external Foundry revenue to a nonrecurring contract. The $411 million is therefore an imperfect baseline for repeatable third-party manufacturing, much less leading-edge wafer sales. A customer evaluating a process or design kit is at a different commercial stage from one placing binding, revenue-producing wafer orders.

The product businesses show a different split. Intel's Q3 results and 10-Q show Client Computing and Physical AI revenue down 5% to $8.097 billion, with client CPU volume down 14% and average selling prices up 9%.

Data Center and AI revenue grew 78% to $7.318 billion, while server CPU unit volume grew 10% and average selling prices 54%, predominantly because of premium mix. Those server measures cover only part of the segment, but they expose growth weighted toward pricing and premium mix rather than a corresponding rise in units.

IDC forecasts an 11.3% contraction in 2026 PC shipments, and analyst Jean Philippe Bouchard expects memory shortages to persist through 2027. That is a market outlook, not a forecast of Intel's sales or market share.

Profits and liquidity need separate tests

Intel reported $16.513 billion of consolidated Q3 revenue, up 21%, and $0.60 in GAAP diluted earnings per share against $0.52 non-GAAP. The 10-Q also records a $1.005 billion noncash Q3 gain on escrowed government shares, so reported earnings alone do not measure the economics of running or expanding fabs.

Q3 operating cash flow was $5.082 billion and gross capital expenditure was $4.219 billion, leaving $863 million by simple subtraction. Intel reported adjusted free cash flow of $1.639 billion after adjustments that include government incentives, net partner contributions and finance-lease payments. These calculations answer different questions, and one quarter cannot establish the cash economics of future manufacturing investment.

At September 26, Intel had $54.005 billion of cash and short-term investments and $50.134 billion of total debt. Its August public share offering raised $22.617 billion net, improving liquidity through the issuance of new shares.

On the earnings call, CFO David Zinsner put 2026 gross capex near $20 billion, expected a strong increase in 2027, and said 2027 probably would not deliver positive free cash flow. Tan's trust purchase changes his reported beneficial exposure but provides Intel no new capital.

Physical progress still needs customers

On Intel's earnings call, management said Intel 3 costs had fallen about 15% and Intel 18A costs about 40% versus Q1, while yields improved. A SemiAnalysis teardown documented Intel 18A structures in manufactured Panther Lake silicon, meaningful technical evidence that does not independently establish fab-wide yields, wafer costs or external-customer margins.

Intel also released a version 0.9 design kit for 14A to selected customers in October, an evaluation milestone rather than a contracted production volume. Bank of America Securities analyst Vivek Arya said on October 30 that Intel "has not communicated enough traction with external foundry customers." His criticism concerns external commercialization, while leaving the status of undisclosed customer discussions open.

What to watch in the next quarters

Intel guided Q4 revenue to $15 billion to $16 billion, below Q3's $16.513 billion, and GAAP gross margin to about 37.1%, against 43.0% in Q3.

Management's preliminary estimate included approximately $300 million of Q4 revenue and 200 basis points of gross margin impact tied to the Fab 28 incident; it said the full financial impact was not yet reasonably estimable. A single quarter's miss or beat therefore needs to be separated from that disruption and the underlying client and server trends.

Over the next three months, the most likely outcome is another loss-making Foundry quarter while server mix supports Products, assuming no further major disruption. Through November 2027, an upside case requires repeatable external wafer revenue, narrowing manufacturing losses and better cash conversion despite increasing spending; a downside case combines persistent PC weakness with slow customer qualification. By November 2029, 14A execution, contracted volumes and returns on invested capital should be more informative than any 2026 insider purchase.

How to test Foundry's progress

The immediate test is whether Foundry's external sales become a larger, repeatable business as operating losses narrow; ultimately those revenues must support Intel's investment. With medium-high qualitative confidence, we expect Intel's Q4/FY2026 results and filings available by a March 1, 2027 desk cutoff to show a quarterly Foundry operating loss and external sales below 25% of Foundry segment revenue.

The 25% threshold tests a substantial shift in revenue mix, not break-even. Nonnegative Foundry operating income or an external share at or above 25% would falsify the call; missing segment detail would leave it unresolved, and passing would not prove lasting profitability.