Google Orders 3 Million AI Chips from Intel

The Information reported, and Reuters and Bloomberg corroborated, that Alphabet's Google has ordered Intel to manufacture more than 3 million tensor processing units (TPUs) for production in 2028. Reuters said Intel shares rose more than 9% on the news, with some outlets reporting premarket gains above 13%. Reporting frames the order against a capacity crunch at TSMC, which builds most leading-edge AI chips. The Information also reported that Nvidia is evaluating whether Intel's advanced packaging could fuse four GPU dies into a single processor for its next-generation Feynman architecture, though Nvidia has not placed an order. Intel declined to comment, and Reuters said it could not independently verify The Information's account. Morgan Stanley has forecast Google's TPU output at roughly 5 million units in 2027 and 7 million in 2028, making the reported Intel order a sizable slice of that buildout.
What happened
According to The Information, corroborated by Reuters and Bloomberg, Google has placed an order with Intel to manufacture more than 3 million tensor processing units (TPUs) for production in 2028. Reuters reported that Intel shares rose more than 9% in early trading, while some outlets reported premarket gains above 13%. Reuters said Intel declined to comment, that Alphabet and Nvidia did not immediately respond, and that it could not independently verify The Information's account.
The deal
The order is for Google-designed TPUs, with Intel acting as contract manufacturer rather than selling its own branded chips. The Information and follow-on coverage report that Google tested Intel's advanced packaging for months; that technology stitches multiple chip dies and high-bandwidth memory into a single module. Reporting also says Nvidia is evaluating whether Intel's packaging could build a processor that fuses four GPU dies into one unit for its next-generation Feynman architecture, though Nvidia has not placed an order, per Reuters and Quartz.
Why TSMC matters here
Outlets frame the story against a capacity crunch at Taiwan Semiconductor Manufacturing Company (TSMC), which fabricates most leading-edge AI silicon. Reuters quoted eMarketer analyst Jacob Bourne calling the development "evidence that AI's biggest players are racing to diversify a supply chain still heavily concentrated in TSMC." Morgan Stanley has separately forecast Google's TPU output at roughly 5 million units in 2027 and 7 million in 2028, which would make the reported 3-million-unit Intel order a sizable slice of that buildout.
Why it matters
A multi-million-unit TPU run would be a significant external win for Intel Foundry, which has been positioned as a TSMC alternative amid investments from the US government, SoftBank, and Nvidia. It would also give Google a second leading-edge production line for hardware central to training and serving its models.
What to watch
- •confirmations or denials from Intel, Alphabet, or Nvidia in filings or official statements
- •yield and packaging milestones tied to the 2028 delivery window
- •how TSMC capacity guidance and customer allocations shift in earnings commentary
- •additional third-party orders that would signal broader adoption of Intel packaging for high-end AI chips
Editorial analysis: Teams reliant on large-scale accelerators will watch supply diversification closely, since multiple contract manufacturers change procurement lead times, risk profiles, and capacity planning. Initial reports should be treated as an early signal rather than a finalized supply contract until companies confirm.
Scoring Rationale
A reported multi-million-unit TPU order placed with Intel Foundry, corroborated by Reuters and Bloomberg, is a material AI-infrastructure supply-chain development that moved Intel shares 9 to 13%. It signals leading AI buyers diversifying leading-edge manufacturing away from a strained TSMC, with implications for foundry allocation and accelerator procurement. Held just above 7 because the deal is single-originated (The Information) and unconfirmed by the named companies.
Practice with real Ad Tech data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Ad Tech problems


