Intel just printed its fastest revenue growth in more than 15 years. The quarter proved the product recovery has real substance. Yet the same report raised the cost of finishing the job. Investors now have to judge a comeback that keeps getting more expensive to own.
The market caught that tension in real time. Shares fell 2.3% in regular trading on July 23, ahead of the release. They then surged after hours. Reported gains ranged from about 3% to 13%, depending on the timestamp. By Thursday morning, the premarket gain had settled near 3% to 4%. The starting point explains the churn. Intel's 2026 gain stood above 170% before the print, with some tallies closer to 178%. A strong product recovery was already in the price. The report had to support something bigger: proof that surging demand can become acceptable returns.
A beat built on operations
Revenue of $16.1 billion rose 25% from a year earlier. That cleared the top of Intel's own guidance by about $1.3 billion. It also beat the $14.42 billion LSEG consensus. Other providers carried different estimates, so the beat's size depends on the source. Adjusted earnings of 42 cents doubled the 21-cent LSEG figure.
One accounting note belongs up front. Intel reported a GAAP loss of $2.16 per share. A $12.5 billion mark-to-market charge drove the loss. The charge relates to shares held in escrow for the U.S. government. Consensus was an adjusted number. Setting it against the GAAP loss would mix accounting bases.
Beneath the headline, the quality held up. GAAP operating income reached $1.8 billion. Adjusted gross margin hit 41.8%, about 280 basis points above guidance. Management credited higher factory yields, faster cycle times, richer mix and pricing. CFO David Zinsner said margins can stay “comfortably in the 40s.” Treat that as a claim, not a record. Early 18A wafer costs and rising depreciation will test it.
Data center revenue rose 59% to $6.3 billion. The segment earned $2.5 billion in operating income. That is the strongest evidence the turnaround has substance. Client revenue rose 13% to $8.9 billion. But Zinsner told Bernstein's Stacy Rasgon the client upside came from pricing and mix. He called third-quarter client revenue “flattish” and second-half PC demand sub-seasonal. Rising memory prices are squeezing PC makers. The client unit is a profit engine, then, and nothing more than that.
Too much demand, not enough supply
CEO Lip-Bu Tan said demand for Intel products continues to outpace supply. He called the shortage one of the most severe in company history. The chokepoints run through leading-edge wafers, memory and substrates. Zinsner told UBS's Timothy Arcuri that supply will not catch up to demand in the fourth quarter.
The constraint reads two ways at once. Customers do not queue for chips without competitive value, so the backlog validates the products. At the same time, Intel cannot ship the demand it has won. Unfilled orders can drift to AMD, Arm designs or custom silicon. The shortage also explains the quarter's biggest strategic shift: more spending, sooner.
A foundry with customers still pending
Intel Foundry posted $5.8 billion of segment revenue. Nearly all of it came from inside Intel. External customers paid just $293 million. The unit lost $2.1 billion at the operating level. Those two numbers frame the entire investment case.
Execution improved on every disclosed measure. Zinsner said 18A output ran about 25% above target and rose more than 50% from the first quarter. Management says yields are ahead of plan. The lead Panther Lake chip's cost is down about 50% this year. Intel also committed to high-volume 14A production in 2028.
That 2028 commitment is where the risk concentrates. No new external 14A customer appeared in Intel's earnings materials. Tesla has reportedly signed on for a 14A project. Intel's own documents do not confirm it. On the call, Bank of America's Vivek Arya asked when a customer announcement would come. Tan expressed confidence but gave no date. Melius's Ben Reitzes asked whether hard orders support the higher spending. Zinsner cited discipline and expected returns rather than committed volumes. He was candid about the limits of long-term deals: “You can't completely hang your hat on (long-term agreements) because when things change, a lot of times things get renegotiated.”
Watch that gap closely. Management is building toward 2028 capacity while the revenue meant to fill it remains largely unsigned.
Guidance extended the momentum. Intel sees third-quarter revenue of $15.8 billion to $16.8 billion. The midpoint tops the $15.10 billion LSEG estimate. Adjusted earnings guidance of 38 cents beat the 27-cent LSEG figure. Management gave little detail on the fourth quarter. Arcuri pressed on it. Zinsner declined to guide. The durability of the current run rate stays open as a result.
Paying for the comeback
Intel raised 2026 capital spending to more than $20 billion, up from $18 billion. Zinsner said 2027 spending will run significantly higher, mostly in U.S. plants. He gave no 2027 number. Analysts asked and did not get one.
The cash statement needs careful reading. Operating cash flow was a healthy $7 billion. Adjusted free cash flow was negative $8.4 billion. A $12.2 billion net outflow tied to partner contributions drove that swing, not routine capex. Strip out the noise and one conclusion stands. Intel has yet to show that improving earnings convert into steady free cash.
The balance sheet holds $12.9 billion of cash and $16.9 billion of short-term investments. Debt stands at $50.5 billion. Zinsner said on the call that Intel could tap capital markets if expansion goes very well. Credit him for honesty. The comment still marks the risk. Spending into real demand can dilute returns if cash conversion lags behind.
Where the burden of proof now sits
Both camps leave this quarter carrying heavier obligations. Bears must explain away seven straight quarters above guidance, by management's count, plus accelerating server demand. Bulls must underwrite a spending plan that keeps growing without a signed external anchor customer.
Four markers would move the debate. A named foundry customer with committed production, not evaluations. A foundry loss shrinking through utilization rather than easier comparisons. Margins holding in the low 40s as pricing tailwinds fade. And a 2027 capex number with a funding plan attached.
Intel spent years asking investors to believe the turnaround could happen. After this quarter, belief is the cheap part. The stock, up at least 170% this year, already carries it. The expensive part is the tens of billions now committed before the customers arrive. The next real catalyst is a signature on a wafer contract, not another beat.