THE APEX TIMES
Google’s reported backup TPU order targets 2028 manufacturing at Intel as TSMC capacity tightens
A Reuters report says Google has placed an order with Intel to produce more than 3 million tensor processing units in 2028, reflecting how AI-chip demand is straining the industry’s leading foundry supplier.
Google has placed an order with Intel for the manufacture of its tensor processing units, with deliveries planned for 2028, according to a Reuters report that cites people familiar with the discussions. The move would add capacity options for Google’s custom AI chips as chip demand continues to outstrip supply at Taiwan Semiconductor Manufacturing Co.
A TPU, or tensor processing unit, is an application-specific AI accelerator designed to speed up training and inference for machine-learning workloads. Google markets TPUs through Google Cloud, and says they power Gemini and other Google AI services.
Reuters said the reported Intel order is for “more than three million” TPUs in 2028, and that the claim was reported by The Information. Reuters added it could not independently verify the report, and said Intel declined to comment while Alphabet and Nvidia did not immediately respond to requests for comment.
The Reuters report also said Nvidia is evaluating whether Intel’s technology could be used to make a processor that combines four graphics chips into a single unit, though Reuters did not say Nvidia has placed an order. The potential Google deal would, if it materializes as described, strengthen Intel’s efforts to expand foundry services, an area it has tried to rebuild after losing advanced chipmaking share to TSMC.
For Google, the reported shift points to a broader trend in the AI hardware market: hyperscalers and chip designers are increasingly looking for second sources of leading-edge manufacturing capacity. Reuters linked that behavior to a supply crunch at TSMC, saying soaring demand has left the company struggling to bring in enough supply, prompting multiple AI-chip customers to consider Intel.
Intel’s reported upside is straightforward but not guaranteed. A major AI-chip customer contract would boost confidence in Intel’s contract-manufacturing roadmap, but semiconductor production timelines depend on complex process technology readiness, packaging capacity, yield performance, and customer qualification. Reuters framed the story as evidence of supply-chain diversification, not as an announced, finalized order.
The biggest unknown is the commercial detail. Neither Google nor Intel confirmed the reported order publicly, and Reuters did not disclose the contract’s financial terms, the exact TPU generation involved, or how the manufacturing agreement would interface with Google’s broader TPU system supply chain. Any future confirmation would likely arrive through company commentary, investor materials, or supplier reporting closer to production.
Why It Matters
- If Google diversifies TPU production beyond TSMC, it could shift leverage in AI-chip supply negotiations and increase pressure on foundry suppliers to add capacity faster.
- A large TPU manufacturing commitment would be a test case for Intel’s manufacturing turnaround, with potential knock-on effects for how other AI hardware buyers plan second sourcing.
- For the AI sector, more multi-manufacturer strategies can reduce delivery risk, but they also raise qualification and integration costs across complex hardware stacks.
- Investors and analysts will watch whether the reported volume, timeline, and TPU generation details are confirmed through later disclosures and whether Intel’s production execution matches the scale implied by the report.
Sources
Key Facts
- Reports, citing people familiar with discussions, say Google has placed an order with Intel to manufacture more than 3 million TPUs in 2028.
- Reuters said it could not independently verify the report, and that Intel declined to comment while Alphabet and Nvidia did not immediately respond to requests for comment.
- TPUs are custom-designed AI accelerators optimized for training and inference workloads, and Google says they power Gemini and Google’s AI applications.
- The reported order is framed as part of a broader industry response to capacity constraints at TSMC amid AI-boom demand.
- The Reuters report also said Nvidia is evaluating whether Intel’s technology could be used to build a processor that would combine four graphics chips into one unit, without indicating an order was placed.
- If confirmed, the deal would support Intel’s foundry ambitions as it seeks to regain share in leading-edge chip manufacturing.
Technology Related
Elon Musk’s chip preference spotlights Nvidia’s edge over AMD, but investors still watch execution
A Yahoo Finance analysis highlighted Nvidia’s faster growth relative to AMD, drawing attention to how high-profile tech users, including Elon Musk, frame the semiconductor race.
Ming-Chi Kuo says Nvidia has revived Rubin CPX after it seemingly vanished from the AI roadmap
The analyst Ming-Chi Kuo says Nvidia’s Rubin CPX accelerator is back, with what he characterizes as a substantial redesign after the chip appeared to be shelved earlier this year.
Apple’s next CEO arrives with a different kind of power: money, and an AI test
A new leadership chapter at Apple, as reported by Yahoo Finance, raises a central question for investors and customers alike: will Apple use its unusual financial profile to change its AI direction, or simply defend its status quo?
ZonPrep buys inbound-inventory software and services, betting on Amazon logistics automation
The Amazon-focused supply chain and FBA prep company says it acquired Wizard-Industries and FNSKU Studio, tools aimed at helping sellers get inventory into Amazon faster and with fewer process steps.
Nvidia pauses part of its AI customer financing after a strong quarter, raising questions about timing
After delivering another heavy AI-related quarter, Nvidia indicated it is stepping back from a portion of its financing approach for customers. Market coverage framed the move as potentially awkward, given investor expectations tied to continued momentum in AI infrastructure spending.
Apple CEO transition hands AI test to John Ternus as AAPL slips
John Ternus takes over as Apple’s chief executive role as Phil Schiller steps back, with market attention focused on how leadership changes could affect ongoing work on artificial intelligence initiatives. Apple shares slid in early trading following the transition reports.
Anthropic reportedly signs $35 billion cloud deal involving Nvidia-backed Lambda and a Texas data-center lease
A Yahoo Finance report says Anthropic has agreed to a long-term cloud-computing arrangement worth $35 billion, with the infrastructure and data-center lease tied to Lambda, an Nvidia-backed provider.
FTC and 22 states sue Amazon, alleging it overcharged advertisers using its retail platform
The U.S. Federal Trade Commission and a coalition of state attorneys general accused Amazon of misleading businesses about pricing tied to advertising on its shopping marketplace, alleging the conduct resulted in billions in gains for the company.
Intel’s push toward on-prem, privacy-focused AI gets a partnership spotlight as Xeon 6 platform work expands
A new extension to Kasm Technologies’ deal work with Intel highlights a market trend toward running large language model workloads locally on enterprise hardware, aiming to reduce data exposure and reliance on GPUs.
Broadcom (AVGO) set to report earnings Wednesday after the bell, with investors focused on guidance and demand outlines
The fabless chip and software maker Broadcom will release its next quarterly results this Wednesday after market close, according to a preview posted by Yahoo Finance.