Intel surges on reports that Google orders TPUs from Intel Foundry, while Nvidia tests Intel technology

Intel shares jumped in early trading Monday after a report said Alphabet’s Google placed an order with Intel to manufacture more than three million tensor processing units, or TPUs, for 2028. The same report also said Nvidia is evaluating whether Intel’s technology could be used to combine four graphics chips into a single unit. The news reinforced the idea that Google and Nvidia are looking for additional chip-making capacity outside of Taiwan Semiconductor Manufacturing Co., where much of the industry’s leading AI hardware is produced.

According to the Reuters report citing people familiar with the discussions, Google’s order is for more than three million TPUs to be manufactured in 2028. Nvidia, meanwhile, is said to be weighing whether Intel’s technology can support a processor concept that would integrate four graphics chips. The report added that Nvidia had not placed an order at the time. Reuters also said Intel declined to comment, while Alphabet and Nvidia did not immediately respond to requests for comment, and that Reuters could not independently verify the claim.

Intel’s rally reflected investor expectations that its contract manufacturing business, commonly known as a “foundry,” could finally gain meaningful, external customers. A foundry sells manufacturing capacity and process expertise to chip designers that want their chips produced without building their own factories. The Reuters report suggested the AI boom’s demand surge has left TSMC struggling to bring in enough supply, pushing some major chip designers to diversify where their custom silicon is made.

For Intel, the potential relevance goes beyond volume. Nvidia’s interest, as described in the report, points toward Intel’s most ambitious process technology and its approach to power delivery. Intel’s 18A process node, the company’s next major manufacturing step, is built around RibbonFET, a gate-all-around transistor design, and PowerVia, a backside power delivery approach. In plain terms, these are structural changes intended to improve performance and efficiency at smaller transistor geometries, while making power distribution more practical for advanced chips.

Intel’s foundry push has also had support from Nvidia in a separate, already-public relationship. In September, Reuters reported that Nvidia planned to invest $5 billion in Intel and that the companies would jointly develop future PC and data center chips. Crucially, Reuters reported at the time that the collaboration would not involve Intel’s foundry business manufacturing computing chips for Nvidia, though Intel’s foundry and advanced packaging capabilities would still be used for central processors and packaged components within the joint products. That distinction matters because it suggests Nvidia’s evaluation of Intel technology described Monday may be aimed at future designs that could extend Intel’s role, even if no contract was confirmed in the report.

Still, the market-read is likely to remain cautious until details are clearer. The Reuters report did not provide financial terms, the specific Intel process node(s) or packaging methods that would be used for Google’s TPU production, or a start date for any manufacturing ramp. It also did not say whether Nvidia’s evaluation is limited to tests, prototyping, or a potential manufacturing qualification process. With Intel, Google, and Nvidia not confirming, and Reuters noting it could not independently verify the information, there is still meaningful uncertainty about whether these plans will translate into durable foundry revenue.

Investors will likely watch for confirmations through company statements, customer announcements, and future filings that could connect specific chip programs to Intel Foundry capacity and timelines. For Google, the key question will be how quickly its in-house TPU program can expand beyond its existing supply chain. For Nvidia, the watch item is whether its evaluation evolves into an explicit manufacturing order tied to a named architecture, and whether the integration concept involving multiple graphics chips moves from testing into volume production.