Intel Foundry Secures 3 Million TPU Order from Google for 2028

Intel Foundry secured an order for over three million TPUs from Google by 2028. This deal follows validation of Intel's advanced packaging technology and signals a shift in the AI accelerator supply chain. The agreement positions Intel as a key manufacturing partner for major cloud providers.

Google orders over three million AI chips while NVIDIA tests advanced packaging

Google plans to produce these chips using Intel's 14A process node. This agreement encompasses several iterations of Tensor Processing Unit architectures, specifically highlighting the forthcoming v8e variant. The scale of this order positions Intel Foundry as a major participant in the mass production of artificial intelligence processors.

NVIDIA is currently evaluating Intel's 18A process for next-generation multi-die GPU designs. The evaluation focuses on packaging four GPUs into a single package using advanced interconnect technology. NVIDIA has begun initial testing through a Multi-Project Wafer model to assess the manufacturing capabilities.

Intel reported that its EMIB technology achieved approximately 90 percent yield during recent production runs. This metric supports potential adoption by Google and Meta for future designs. Tesla serves as the first major customer for Intel's 14A process, with chips destined for a planned AI computing hub in Austin.

The NVIDIA evaluation represents an early stage of testing rather than a finalized partnership. Market analysts suggest that reducing dependence on a single supplier may significantly alter the competitive landscape of semiconductor manufacturing. The outcomes will determine whether major GPU makers integrate Intel's 18A process into their roadmaps.