Intel beat the quarter cleanly, then got punished for the one thing investors are suddenly tired of funding: another big AI infrastructure bill with the payoff pushed into next year.
The quarter couldn't have looked much better on paper. Intel reported $16.1 billion in Q2 2026 revenue, up 25% year-over-year, and non-GAAP EPS of $0.42. MarketBeat's earnings data showed Wall Street had expected about $14.43 billion in revenue and $0.21 in EPS. That is a real beat. It still wasn't enough.
Intel shares closed Friday, July 24, at $92.32, down 7.9% from the previous close, according to MarketBeat and StatMuse pricing data. The stock had jumped after the earnings release. Then investors read the spending plan. Intel raised its 2026 capital expenditure outlook to more than $20 billion from $18 billion, with spending expected to move higher again in 2027. You don't need a complicated market theory for that reaction. Wall Street liked the revenue. It hated the bill.
The operating numbers deserve to be taken seriously. Intel's own earnings release called it the company's strongest revenue growth in more than 15 years, and Business Insider reported that the Data Center and AI business rose 59% year-over-year to $6.3 billion. That segment is doing the work. Server CPUs are being pulled into AI systems alongside GPUs, and Intel is benefiting from that demand after years of looking late to the most important computing shift in the market.
The ugly number was the GAAP loss. Intel reported a loss per share of $2.16, while the non-GAAP number was positive. The difference came down to a large mark-to-market charge on escrowed shares under the US government's CHIPS Act investment. That's an accounting detail, not an operating collapse. Still, an $11 billion reported net loss lands badly when the company is also asking investors to finance heavier spending.
The foundry proof is still thin #
The harder question is still Intel Foundry. The segment brought in $5.8 billion in revenue and lost about $2.1 billion in the quarter, according to recent earnings summaries and company filings. That is better than the mess Intel carried a year ago, but better isn't fixed. A foundry turnaround has to end with outside customers paying real money for real volume. Intel doesn't have enough of that yet.
Fortinet gave Intel something useful last week. Tom's Hardware reported that Fortinet will use Intel to co-develop and manufacture its next-generation Security Processor 6 on the Intel 4 process. The Motley Fool described Fortinet as the first named outside foundry customer since Lip-Bu Tan became Intel CEO in March 2025. A named customer matters. It gives Intel something checkable after months of broad language about interest and engagement.
Don't overstate it. Intel 4 is a 2023-era process used for Intel's own Meteor Lake chips, not the leading-edge 18A process the turnaround story depends on. Tom's Hardware also noted that the companies didn't give a production timeline. The deal is a signal, not a verdict. If you're buying the foundry thesis, you still need to see a major customer commit volume to 18A or 14A. Until then, the most important part of Intel's story remains unproven.
Tan has been blunt about the timing. Reports from his Stanford appearance earlier this year said he framed 2026 as an execution year and put the growth inflection in 2027. That is probably the honest version of the story. It is also a difficult pitch in a market that is no longer rewarding AI spending just because it has the letters A and I attached to it.
AI spending has stopped getting a free pass #
Intel walked into a wider investor mood shift. Alphabet raised its 2026 capex outlook to $195 billion to $205 billion, and Reuters reported that the company burned $5.9 billion in free cash flow in the second quarter as AI spending climbed. The stock sold off. Tesla was hit after heavy spending on AI, autonomy and chip facilities. The market isn't rejecting AI demand. It is asking when the cash comes back.
That is the trap Intel now sits in. Its Data Center and AI revenue proves customers are buying more Intel products for the AI buildout. Its foundry spending asks shareholders to believe outside manufacturing customers will arrive later. Those are different claims. One is in the results. The other is still mostly in the plan.
There is a reasonable long-term case here. The core product business is recovering, foundry losses are narrowing, and the CHIPS Act gives Intel a kind of US government backing TSMC and Samsung don't have in the same way. But investors aren't obliged to pay today for a 2027 payoff. Frankly, after the run Intel's stock has had this year, they were right to demand more proof.
A beat-and-raise quarter usually gives management room to breathe. Intel got the opposite. The next test isn't whether Tan can produce another strong headline quarter. It is whether Intel can put a serious outside foundry customer on 18A and show that more than $20 billion of annual capex is building a business, not just keeping a turnaround story alive.
Also read: China starts mass-producing its own DUV chip tools and ASML loses 6% in a single day • JPMorgan eliminates its standalone AI chief role and the move says everything about where enterprise AI is headed • Nvidia's $750 billion AI financing web is drawing scrutiny over circular money flows