Intel Corp. INTC

Key data

The business

Intel designs computer chips and, unlike almost everyone else, still builds them in its own factories. Three engines matter. Client Computing, the processors inside laptops and desktops, is the cash cow that pays the bills. Data Center and AI sells server chips, and here Intel has been losing ground to Nvidia's accelerators for years. Intel Foundry is the swing: the plan to rent out those factories to other chip designers, the way Taiwan's TSMC does, so the enormous fixed cost of a leading-edge fab gets spread across many customers instead of one. That is the whole story. A modern fab costs north of $20 billion and only pays off when it runs full. Intel's do not run full, because for now the only big customer is Intel. The moat, if it arrives, is being the one leading-edge foundry on American soil that a hyperscaler or a defense program can trust. Today it is a promise, not a moat.

The numbers

The top line inflected hard last quarter and the bottom line got worse at the same time.

Q3 2025's profit was the Altera divestiture gain, not operations. What matters is Q2 2026: revenue up 25% year over year and 19% sequentially, and operating income swung to a positive $1.8 billion from a $3.2 billion loss a year earlier. The $11 billion net loss sits entirely below that line, in goodwill write-downs (goodwill fell $3.4 billion in six months) and a tax charge. The operating turn is real; the headline loss is cleanup. On the adjusted basis the Street tracks, Intel earned $0.42 against a $0.21 estimate, a clean double.

Revenue has fallen by a third since 2021 and profit has evaporated. The compounding math the stock is asking you to believe runs the other way: from a $483 billion market cap, at 31 times a hoped-for $3.08 of 2028 earnings, the price already banks a recovery that the segment file does not yet show. The gating item is cash, not the income statement.

Capex is down and free cash burn is narrowing, but it is still burn, and Q2 alone drained $4.4 billion of cash to $12.9 billion. The variant view: the Products operating recovery is real and under-priced, while the Foundry story that justifies the multiple is unproven and the cash clock is running. The print that settles it is Foundry external revenue plus the quarterly cash balance.

Management

The signal that carries weight is executive Lip-Bu Tan buying $10 million of stock with his own money on August 11, the single largest insider action in a year and a rare thing for a chief executive at this price. Set against it, three officers sold about $7.5 million combined, plan status not disclosed. Buybacks have been zero since 2022, the correct call while burning cash, and the dividend was already gone. Pay is tied to margin and operating-income targets per the incentive plan, which at least points management at the number that actually turned. The record here is a founder-operator putting cash in while the house is on fire, which is either conviction or a signal he can see the demand the filings cannot yet.

How it fails or surprises you

Foundry wins a marquee customer (right tail). If 18A lands a hyperscaler or a large fabless name at volume, external Foundry revenue inflects from a rounding error and the half-empty fabs start earning their cost. First proof: a step-change in the Foundry segment line.

The cash runs short. Burn of $4.4 billion in one quarter against $12.9 billion of cash, with $46.6 billion of long-term debt as of year-end 2025, means another year of this forces dilution, more debt, or deeper government dependence. The tell is operating cash flow and the cash balance in the next two prints.

The revenue jump was pull-forward. Q2's 25% surge may reflect customers building inventory ahead of tariffs, and the 10-Q flags customer cash received in the quarter. If so, the best quarter in years reverses next print, and the operating turn proves temporary.

Closing thoughts

The shape of this is: two numbers tell you which story is true, and both are disclosed every quarter, so you find out fast. The fat left tail is the balance sheet: sustained cash burn against a heavy debt load, where the permanent loss lives and where survivability, not upside, is the only question that matters. The right tail is genuinely large and genuinely unpriced in the segment numbers, a single external Foundry win that re-rates the whole thesis. Between them sits a stock that has already run four-fold off its low, so the market is paying today for a recovery that must show up in Foundry revenue and free cash to be real. Judgment, not arithmetic: the downside is better defined than the upside is proven.

The bet is still that Intel finally learns to make chips as well as Taiwan does, and that the world's biggest tech companies pay Intel to build their chips in its American factories. What breaks it is the fab that never fills while the cash drains. Watch two numbers together, Foundry external revenue and the quarter-end cash balance; if external revenue is still a rounding error and cash is below $10 billion next print, the recovery the price assumes has not arrived.