Intel’s third quarter read like a company trying to pull itself back into contention—with just enough progress to keep the turnaround story alive, and just enough caveats to remind investors how much work remains.

Revenue rose 3% year over year to $13.7 billion, ahead of guidance, as the PC cycle finally lent a hand and data-center demand stabilized. On the bottom line, Intel reported GAAP EPS of $0.90 and non-GAAP EPS of $0.23, as gross margin recovered to 38.2% on a GAAP basis and 40.0% non-GAAP—still far from Intel’s historic comfort zone, but a marked improvement from last year’s trough. “Our Q3 results reflect improved execution and steady progress,” said CEO Lip-Bu Tan, arguing that AI is expanding—not cannibalizing—the company’s core compute markets. CFO David Zinsner added that “current demand is outpacing supply,” a dynamic Intel thinks could persist into 2026.

Under the hood, the quarter showed a company increasingly defined by two engines: a steadier PC franchise and an ambitious, still loss-making manufacturing push. Client Computing Group revenue rose 5% to $8.5 billion, helped by enterprise Windows 11 refresh and a healthier notebook mix. Data Center & AI slipped 1% to $4.1 billion—hardly a boom, but a far cry from the declines of the last downcycle. At the segment level, Intel Products delivered $12.7 billion in revenue while Intel Foundry, the stand-alone manufacturing unit, booked $4.2 billion of revenue and a sizeable operating loss—evidence that the foundry build-out is still in its investment phase even as it begins to contribute top-line.

If Q3’s financials were about stabilization, the strategy updates were about scale. Intel leaned hard into “balance-sheet as strategy,” closing or announcing a series of cash-raising and partnership moves that would have seemed unthinkable a few years ago. The company said it received U.S. government funding during the quarter as part of a broader $8.9 billion package tied to domestic manufacturing; it also unveiled a collaboration with Nvidia to co-develop custom data-center and PC products using NVLink—and disclosed that Nvidia agreed to invest $5 billion in Intel stock. SoftBank chipped in with a $2 billion equity investment. Whatever one’s view of the financial engineering, the message is unmistakable: Intel aims to be a primary Western supplier of leading-edge logic, and it’s recruiting deep-pocketed partners to get there.

The product roadmap—so often Intel’s stumbling block—showed signs of momentum. The company previewed its first client SoCs on the 18A node, the Core Ultra 3 “Panther Lake,” and offered a first look at next-gen server silicon, “Xeon 6+” (Clearwater Forest), also on 18A. It also teased an inference-focused data-center GPU, “Crescent Island,” signaling an intent to play more deliberately in the fast-growing AI inference arena, where power efficiency and memory bandwidth, not just raw training horsepower, will decide winners. Intel said Fab 52 in Chandler, Arizona—dedicated to high-volume 18A production—is now fully operational, a milestone for the domestic manufacturing narrative that underpins much of the company’s capital plan.

There were important portfolio moves as well. Intel closed the sale of a 51% stake in its Altera programmable-logic unit, deconsolidating the business from results going forward; guidance for the fourth quarter excludes Altera. The company also monetized a portion of its Mobileye stake, adding to the cash cushion it says it needs while it invests through the cycle. These transactions, together with the public-sector and strategic-investor funds, leave Intel with more operating flexibility—but also tie the turnaround to execution milestones that partners and policymakers will be watching closely.

Guidance reflected a company managing the calendar as much as the cycle. For the December quarter, Intel guided revenue to $12.8 billion to $13.8 billion, with non-GAAP EPS of $0.08 and GAAP EPS of negative $0.14 at the midpoint—an outlook that bakes in continued margin pressure from early 18A ramps and the absence of Altera. Intel also flagged unusual accounting complexity tied to its government transactions, noting it is consulting with the SEC and that results could be revised depending on the staff’s view. That’s not the sort of footnote investors love to see, but it is part of the cost of rebuilding a strategic manufacturing footprint with public support.

The larger question after Q3 is whether Intel’s two-track story—cash-generating PC and server CPUs on one side, capital-hungry foundry and advanced packaging on the other—can converge fast enough to deliver durable, self-funded growth. The quarter offers a plausible “yes”: margins moving up, PCs healing, foundry capacity coming online, and marquee partners leaning in. It also offers a candid “not yet”: data-center CPU share still contested, foundry still deep in the red, and a new product cycle that must hit dates and yields to change the narrative. For now, the turnaround remains a game of inches. But compared with where Intel was a year ago, the company has a clearer lane—and, after Q3, a bit more speed in it.