Intel shares surge on report that Google ordered more than 3 million TPUs for 2028 manufacturing

Intel shares jumped sharply on Monday after a report that Alphabet’s Google placed an order with Intel to manufacture more than three million tensor processing units, or TPUs, in 2028. The Information, citing people familiar with the discussions, said the order is tied to Google’s in-house AI chip program and that Nvidia is also looking at Intel technology for a future processor concept, though Nvidia has not placed an order. Reuters reported that Intel declined to comment, while Alphabet and Nvidia did not respond immediately to requests for comment, and Reuters said it could not independently verify the report.

TPUs are custom AI accelerators designed to run machine learning workloads, particularly the mathematical operations that drive neural networks. TechTarget describes a TPU as an application-specific integrated circuit (ASIC) built to accelerate tensor computations common in machine learning tasks. Google has spent years building a stack around TPUs for its own data centers, including using multiple generations optimized for training and inference. In a recent update, Google introduced Ironwood as its seventh-generation TPU designed specifically for inference workloads, underlining how central its custom silicon has become to serving AI models at scale.

For Intel, the reported Google order would be a meaningful validation of its contract manufacturing push, often referred to as Intel Foundry services. Reuters said the possible TPU order would bolster Intel’s contract chip manufacturing business at a time when the broader AI chip boom is straining supply from leading manufacturers. The report also pointed to Intel’s goal of regaining semiconductor manufacturing leadership it lost after years in which Taiwan Semiconductor Manufacturing Co. became the dominant supplier for many advanced chips.

Intel is positioning its manufacturing roadmap around future process nodes and advanced packaging approaches meant to improve performance and power efficiency for demanding compute workloads like AI. Intel’s materials on its Intel 18A platform describe it as a technology and packaging combination intended to deliver efficiency and density improvements versus earlier nodes, and it emphasizes the platform nature of the process plus packaging. If Google’s reported 2028 TPU volume routes through Intel’s foundry capabilities, it could also help Intel translate its technical roadmap into repeatable production work for a hyperscaler that designs its own chips.

The same Reuters report said Nvidia is evaluating whether Intel technology can be used to make a processor concept that combines four graphics chips into a single unit. Reuters noted that Nvidia has not placed an order yet, which suggests the effort is still exploratory and could depend on performance targets, yield, and packaging integration. Even without a confirmed purchase, Nvidia’s interest would announcement that multiple AI chip designers are searching for supply-chain diversification beyond a single dominant foundry.

Several details remain unclear, and neither Google nor Intel publicly confirmed the reported arrangement. Reuters said it could not independently verify the report and that the companies did not immediately respond to comment requests. The disclosures in the report, as summarized by Reuters, did not provide the contract terms, unit economics, whether the order is binding, what portion of production is tied to specific Intel process nodes, or how the chips would be packaged and integrated.

Investors and customers will likely watch for follow-through: whether Alphabet and Intel later acknowledge the order, whether Nvidia advances from evaluation into a purchase, and whether the arrangement expands into additional TPU generations or other internal chip families. Reuters also framed the news as part of a broader push by AI industry leaders to diversify a supply chain still concentrated around TSMC, meaning any confirmed hyperscaler-to-foundry relationship could have ripple effects across the competitive landscape for contract manufacturing.