What Drives Intel (INTC) Shares: Fabs, x86 and Foundry Customers

Few large-cap technology companies are harder to read from a single headline number than Intel. Any useful Intel stock analysis has to separate three things that moved in different directions over the past year: an operating business that has returned to double-digit growth, a contract manufacturing arm that is still losing billions, and a reported bottom line distorted by an accounting treatment of shares the company owes the US government. In the quarter ended 27 June 2026, reported on 23 July 2026, Intel (INTC) posted its fastest revenue growth in more than fifteen years and an $11 billion net loss at the same time. Understanding how both of those statements are true is most of the work.

Two businesses under one roof

Intel designs chips and it manufactures chips, and since 2024 it has reported those activities as separate segments with internal transfer pricing between them. That structure is the single most important thing to grasp before looking at the numbers.

The product side sells into two main end markets. Client Computing covers the x86 processors that go into notebooks and desktops, a business tied to PC refresh cycles and to Intel's share against Advanced Micro Devices (AMD) and, increasingly, Arm-based designs. Data Center and AI covers server processors plus the accelerator and networking products aimed at AI infrastructure, where Nvidia (NVDA) dominates the high end of training silicon.

Intel Foundry is the manufacturing network: the fabs, the process nodes, the packaging. Today most of its revenue comes from making Intel's own chips, which is why its reported revenue is large while its external customer base is still small. The strategic bet is that Foundry eventually sells meaningful capacity to outside customers, the way Taiwan Semiconductor Manufacturing (TSM) does. That is the hinge on which the whole investment case turns, and it has not swung yet.

What the Q2 2026 numbers actually showed

Revenue in Q2 2026 was approximately $16.1 billion, up roughly 25% year on year and around $1.8 billion above the midpoint of the company's own guidance. Intel described it as the strongest revenue growth in more than fifteen years.

The segment detail matters more than the total:

• Data Center and AI: revenue of about $6.3 billion, up 59% year on year. This is the line that changed the story, driven by server CPU demand alongside AI buildouts.
• Client Computing: revenue of $8.9 billion, up 13%. Still the largest segment, and still the cash engine.
• Intel Foundry: revenue of $5.8 billion, up 31%, but with an operating loss of roughly $2.1 billion in the quarter.

Margins moved sharply. GAAP gross margin came in at 40.4%, against 27.5% in the comparable quarter a year earlier. On a non-GAAP basis gross margin was 41.8%, about 280 basis points above guidance, which management attributed to higher revenue, better manufacturing yields and higher average selling prices. Non-GAAP earnings were $0.42 per share against a consensus nearer $0.21, and non-GAAP net income was $2.2 billion. Operating cash flow was $7.0 billion in the quarter and $8.1 billion across the first half.

For context on how recent that turn is: full-year 2025 revenue was about $52.9 billion, essentially flat on 2024, and Intel Foundry recorded 2025 revenue of $17.8 billion against a full-year operating loss of $10.3 billion per the 10-K segment disclosure. Two consecutive quarters of growth do not erase that, but they do change the starting point.

The accounting line that turned a profit into a loss

Intel's GAAP net loss for Q2 2026 was approximately $11.0 billion. Almost all of it traces to a single non-cash item of roughly $12.5 billion.

Under its August 2025 CHIPS Act Secure Enclave agreement with the Department of Commerce, Intel issued shares into escrow that are released to the government as the company performs under the agreement and receives cash proceeds. Accounting rules require Intel to carry that obligation as a derivative liability measured at fair value, with changes running through interest and other, net. When Intel's share price rises, the value of the shares it owes rises, and the increase is booked as a loss. The derivative liability stood at $15.6 billion at the end of Q2 2026, up from $2.7 billion in December 2025.

The practical implication for anyone doing their own Intel stock analysis is that GAAP net income and GAAP EPS are, for now, close to useless as a measure of operating performance, and they will move inversely to the share price. The charge is non-cash and does not touch operating cash flow. It is also not nothing: it reflects a real obligation to hand over real equity. The honest reading is that the operating business earned about $2.2 billion in the quarter on a non-GAAP basis, while the reported figure tells you something about the stock price rather than the business.

The balance sheet and the capital bill

Leading-edge manufacturing is the most capital-hungry activity in technology, and Intel's balance sheet reflects a company funding it mid-turnaround. At 27 June 2026 Intel held $12.9 billion in cash and equivalents plus $16.9 billion in short-term investments, roughly $29.7 billion between them. Against that sat long-term debt of $48.5 billion and short-term debt of $2.0 billion, so total borrowings of about $50.5 billion.

Management guided to more than $20 billion of capital spending in 2026. Three funding events in the past eighteen months are worth knowing about, because together they explain much of the change in the share count and the debt load:

• August 2025: the US government converted roughly $8.9 billion of CHIPS Act and Department of Defense grants into Intel equity, initially about a 10% holding, reduced to around 8.2% after a subsequent share sale. The government holds no board seat. SoftBank separately bought 87 million shares at $23.00 each, about $2 billion.
• December 2025: the Federal Trade Commission approved Nvidia's $5 billion equity investment in Intel, which came alongside an agreement to jointly develop custom processors.
• April and August 2026: Intel paid $14.2 billion to buy out Apollo's 49% interest in the Fab 34 joint venture in Ireland, funded with cash and a $6.5 billion bridge loan later refinanced with notes. In August 2026 it priced an upsized $20 billion equity offering at $95.00 per share, selling 210,526,315 shares for net proceeds near $19.7 billion. Estimates of the resulting dilution cluster around 4% to 5% of the pre-offering share count, varying with the share-count base used.

Intel pays no dividend. The payout was suspended and remained at zero through mid-October 2026, with a trailing twelve-month dividend of $0.00. Income investors who once held Intel for its yield no longer have that reason to, and all of the current capital is being directed at the fab roadmap rather than returned to holders.

Process nodes, customers and what the market is paying

The product roadmap and the foundry case both run through Intel's process nodes. Panther Lake, the first high-volume product on the 18A node, is in production, and third-party analyst estimates of 18A yields have improved through 2026 from the mid-60s toward the 80s in percentage terms. Those yield figures come from analysts rather than company disclosure, so treat them as directional. The 18A-P variant is the one aimed at external customers.

The commercially decisive node is 14A. Intel has said it is engaged with multiple prospective external customers and expects them to make 14A decisions in the second half of 2026 and into the first half of 2027. No large external customer has a 14A product shipping in volume today. For a foundry, a signed commitment years before first silicon is the entire basis of a capacity plan, which is why announcements here move the stock more than any single quarter's revenue.

On valuation, the market is clearly pricing something other than trailing earnings. Shares closed at about $89.85 on 16 October 2026 and traded near $95 on 19 October. Forward price-to-earnings estimates ranged from roughly 46 to 63 times depending on the data provider and the earnings base used, and price-to-sales sat around 9.5 to 10.6 times, against a historical median closer to 3 times according to UBS strategists. A forward P/E measures price against expected earnings, so a high reading can mean either that expectations are elevated or that the earnings base is still depressed relative to what the assets could produce. With Foundry losing money at the operating line, both interpretations are live, and the spread between data providers is itself a signal that the denominator is unsettled.

What to watch from here

Third-quarter 2026 results are scheduled for 29 October 2026. Guidance was for revenue of $15.8 billion to $16.8 billion with non-GAAP gross margin near 42% at the midpoint, so the first question is whether the Data Center and AI momentum held and whether margin continued to build.

Beyond the quarter, four things would meaningfully change the picture. The first is the Foundry operating loss: it has to narrow on a path toward breakeven, and the rate of improvement matters more than the absolute number. The second is a named external 14A customer with volume attached, which would convert the foundry thesis from strategy to contract. The third is capital discipline, specifically whether capital spending stays near the guided level or expands again, and whether further equity issuance follows the August raise. The fourth is server and PC share against AMD, visible in segment revenue growth rather than in any single product review.

A reasonable Intel stock analysis treats the operating recovery and the structural question as separate judgments. The Q2 2026 numbers show a products business with real momentum and margins recovering from a low base. The Foundry losses, the $50 billion debt load, the absent dividend and the unproven external customer pipeline are all still in place. The derivative charge, meanwhile, will keep making reported earnings look stranger than the business is, in either direction, for as long as the escrow obligation sits on the balance sheet. Those are the lines to follow rather than the headline figure.

Broadcom (AVGO) sits in the same competitive set and comes at AI infrastructure through custom silicon and networking rather than general-purpose processors, which makes it a useful contrast to Intel's product strategy. On the equipment side of the same capital cycle, Lam Research (LRCX) supplies the etch and deposition tools that leading-edge fabs buy, so its business reads across to whether spending plans like Intel's are being executed.