Analysis
Intel delivered one of the cleanest earnings beats of the current reporting season and got punished for it anyway. The company earned $0.42 per share against a Wall Street estimate of just $0.21 -- roughly double consensus -- on revenue of $16.1 billion, beating estimates by $1.7 billion and posting its best top-line growth in more than fifteen years. Shares initially popped after hours, then reversed hard: by the end of trading Friday, Intel was down 11% on the day, wiping out roughly $90 billion in market capitalization accumulated since Thursday afternoon.
The disconnect comes down to one number: capex. CFO David Zinsner told analysts Intel is raising its 2026 capital expenditure guidance from $18 billion to more than $20 billion, and warned that 2027 spending will move "significantly above" that level. Zinsner also said Intel could tap capital markets for additional funding if needed -- language that immediately raised dilution concerns among investors who had spent the past year watching Intel's turnaround story stabilize under CEO Lip-Bu Tan. A stock that had rallied roughly 327% over the prior year, per pre-earnings coverage, was primed for exactly this kind of a profit-taking reaction to any spending surprise.
The capex debate is really a referendum on Intel's foundry bet. The Foundry division is still unprofitable, but Zinsner said losses are narrowing every quarter as external revenue climbs, and the company's ASIC business is approaching a $2 billion annual run rate. Tan has described the addressable custom-silicon opportunity as a market worth more than $100 billion. Critically, both Apple and Microsoft are now confirmed as design partners on Intel's 18A process node -- a meaningfully different signal than the vague "customer interest" Intel has cited in past quarters, since HSBC analyst Frank Lee expects those commitments to convert into recognized revenue as soon as the second half of 2026.
“Tan has described the addressable custom-silicon opportunity as a market worth more than $100 billion.”
The numbers land in sharp contrast to Intel's own recent history and its foundry rival's trajectory. Intel spent much of the past three years ceding leading-edge manufacturing ground to TSMC, whose foundry customers -- Apple, Nvidia, AMD -- effectively built the modern AI compute stack somewhere else. Every dollar of the new capex guidance is Intel's attempt to reclaim a slice of that business by proving 18A can compete on both cost and yield. Samsung Foundry, the other major alternative, has faced its own yield and customer-concentration struggles, meaning Intel's window to re-establish itself as a credible third option is real but narrowing.
Data Center and AI segment strength -- with server revenue growth reported around 59%, among Intel's best in years -- is the part of the print that should reassure the bulls. It shows Intel's core silicon business is still winning share in an AI infrastructure buildout dominated by Nvidia GPUs and custom accelerators from Google, Amazon and Microsoft. But a chip company's server revenue growing alongside a foundry business still bleeding cash is a genuinely different investment case than a pure AI-demand story, and Friday's selloff shows the market is still deciding which one it's pricing.
For venture investors and founders in the semiconductor and AI-infrastructure space, Intel's quarter is a useful data point on capital intensity. Even a legacy chipmaker with decades of manufacturing scale needs to raise its spending guidance twice in two quarters and openly discuss tapping capital markets to fund an AI-era transition. Startups pitching custom silicon, foundry-adjacent tooling, or AI chip design should expect the same investor scrutiny on unit economics and path-to-breakeven that just hit Intel's stock, only with far less balance-sheet cushion to absorb it.
The bear case is straightforward: capex guidance has moved twice this year already, and "significantly above $20 billion" for 2027 is a soft commitment, not a number. If 18A design wins from Apple and Microsoft don't convert into real shipped-unit revenue on the timeline HSBC and others expect, Intel will have spent tens of billions chasing a foundry turnaround that still isn't profitable. Analysts, per CNBC's own framing of the print, remain genuinely "on the fence."
Watch whether Intel's external foundry revenue actually climbs toward the $400-500 million range analysts are watching for over the next couple of quarters, whether a third marquee customer commits to 18A or the next-gen 14A node in real volume, and whether Intel's Q3 report shows the Foundry operating loss continuing to narrow at the same pace management is promising.
