Alphabet Taps Intel to Manufacture Three Million Chips

The Information reported that Alphabet's Google has placed an order with Intel to manufacture more than three million tensor processing units for delivery in 2028, citing people with direct knowledge of the discussions, per Reuters and Bloomberg. Reuters and The Economic Times also reported that Nvidia has been evaluating whether Intel's technology can be used to build a processor that combines four graphics chips into a single package, but has not placed an order, according to those reports. Reuters noted that it could not independently verify the Information's reporting, and that Intel shares climbed on the news. Industry context: Companies facing constrained foundry capacity have increasingly explored alternate contract manufacturers and in-house production options.
What happened
The Information reported, citing people with direct knowledge of the discussions, that Alphabet's Google placed an order with Intel to manufacture more than three million tensor processing units for 2028. Reuters and Bloomberg covered The Information's report and said they could not independently verify the claim. Per Reuters, Nvidia is evaluating whether Intel's technology could be used to make a processor that combines four graphics chips into a single unit, but has not placed an order, according to The Information. Reuters reported that Intel shares rose sharply on the news, and The Economic Times published a story based on Reuters coverage.
Technical details (reported)
Bloomberg and The Information said Google would use Intel to make some of its TPUs after months of testing Intel's manufacturing technology, per those reports. Reuters reported that the potential order would boost Intel's contract chip manufacturing business and cited broader industry capacity constraints at Taiwan Semiconductor Manufacturing Co. (TSMC) as a factor pushing some designers to consider Intel. Reuters also noted prior reporting that Tesla had signalled interest in Intel's 14A process for its Terafab project, and that Intel has drawn investment attention from U.S. government initiatives, Nvidia and others since leadership changes at Intel.
Editorial analysis - technical context
Built-in constraints at leading foundries during the AI demand surge have produced a pattern where large cloud and chip-design firms pursue diversified manufacturing routes, including external foundries, in-house builds, and alternative contract manufacturers. For practitioners, that pattern increases the likelihood of more heterogeneous supply chains, which raises integration, validation, and packaging complexity for AI accelerator deployment.
Context and significance
A reported multi-million-unit order from a hyperscaler to a non-TSMC foundry would be notable because it reflects both demand for AI-specific silicon and willingness by major designers to spread production across suppliers. Observers following the sector have highlighted that securing large-volume capacity is a primary limiter for new accelerator designs, and that shifting some production to Intel could relieve pressure on TSMC capacity, according to Reuters and Bloomberg coverage of The Information's reporting.
What to watch
Observers should track whether The Information's report is corroborated by additional reporting or company disclosures, shipping timetables tied to the 2028 delivery window, and whether Nvidia or other major designers place formal orders with Intel. Also watch for technical disclosures about packaging approaches if multi-GPU combination chips are pursued, and for any statements from Intel, Google or Nvidia responding to the reports. Reuters explicitly said it could not independently verify The Information's account, and both Reuters and Bloomberg noted the Information as the primary source.
Scoring Rationale
A reported multi-million-unit order from a hyperscaler to Intel matters for AI infrastructure and supply chains, indicating material shifts in manufacturing sourcing and capacity dynamics. The item is notable but not paradigm-changing, and primary reporting rests on The Information with independent verification pending.
Practice with real Ad Tech data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Ad Tech problems

