Intel hits record on AI inference CPU demand

TL;DR:

• Intel shares jumped more than 24% to $83 in early Friday trading, surpassing the company’s 2000 dot-com peak and pushing market value above $416 billion.
• AMD and Arm rose more than 11% each on the view that AI inference is reviving CPU demand after years of GPU dominance.
• For UK enterprise IT teams, the more interesting signal is supply: Intel says first-quarter demand was so tight it sold legacy and de-spec chips it had previously written off.

Intel’s first-quarter results triggered a more than 24% share-price jump on Friday, briefly carrying the company’s market value above $416 billion and surpassing its dot-com era peak from 2000, according to Reuters. Demand for Xeon server CPUs from firms providing AI services drove the upbeat forecast — including, unusually, the company selling chips it had originally written off as unsellable.

The stock ended a long stretch in which graphics processing units made by Nvidia and AMD dominated the AI investment narrative. AI inference — the process by which a trained model answers user queries at deployment — runs frequently on CPUs once a model is in production, which is the dynamic Intel chief financial officer David Zinsner pointed to. Rivals AMD and Arm Holdings climbed more than 11% each on Friday in sympathy. Even Nvidia, which last month unveiled a new central processor of its own — a rare move into territory it had ceded — was up more than 1%.

Tight supply, written-off inventory selling out

Zinsner said the forecast was partly driven by higher prices and a first-quarter supply crunch that forced Intel to dig into finished-goods inventory and sell chips it had not expected to move. “It was either de-spec product or legacy product we had shelved and then we worked with customers,” he told analysts. He cautioned the same benefit was unlikely to repeat in the second quarter. At least 23 brokerages raised their price targets after the results; HSBC explicitly cited rising demand for Xeon server CPUs in AI data centres.

Looking forward

Intel’s run combines two signals UK businesses should watch separately. The first is investment: the chip sector’s premium pricing is now spreading from training-side accelerators to inference-side CPUs, which feeds directly into the cost of running AI in production rather than just building it. The second is procurement reality: legacy and de-specified Xeon parts being sold out is an indication of how tight the inference supply chain is becoming. Companies architecting their AI rollouts on assumptions of cheap, abundant CPU compute should sanity-check those assumptions against current lead times. Intel’s Tesla foundry win this week and the planned 2027 contribution from its 14A process are the longer-term return-to-form story; the inference-cycle story is the one that affects 2026 budgets.