What's happening
Google has placed an order with Intel to manufacture more than 3 million Tensor Processing Units (TPUs) slated for 2028, according to a report from The Information dated around June 8, 2026, which cited four people familiar with the matter. The order follows months of Google testing Intel's chip packaging technology and represents a meaningful expansion of Intel's role in the AI hardware supply chain through its foundry business. Alphabet shares dropped 1.27% intraday on the news.
In a parallel development, Google is in active discussions with Samsung Electronics to manufacture the memory I/O die component of its 10th-generation TPU, internally codenamed Icefish, using 2nm process technology and also targeting a 2028 production window. Under this arrangement, TSMC would continue to handle the main compute engine of the Icefish chip, indicating a deliberate split-sourcing strategy rather than a wholesale departure from TSMC. Nvidia, meanwhile, is separately evaluating Intel's 18A process node and advanced packaging capabilities for chips based on its Feynman architecture, which is also due in 2028.
Why it matters for markets
The Google TPU order represents a concrete test of Intel's foundry ambitions at scale. Intel's foundry business has been a central pillar of the company's strategic repositioning, and an order of more than 3 million units from one of the world's largest cloud infrastructure operators — Alphabet, which reported $422.50 billion in revenue — would constitute a significant commercial validation. Intel's intraday move of between 9.2% and 13% on the news reflects the magnitude of market attention on the company's ability to attract hyperscaler customers, though Intel's 52-week range of $18.97 to $132.75 illustrates the degree of volatility the stock has experienced during its ongoing transformation.
For the broader AI chip supply chain, the emerging multi-foundry model carries structural implications. By splitting TPU production across Intel, Samsung, and TSMC — assigning different die components to different manufacturers — Google is distributing fabrication risk across multiple nodes and geographies. This approach also introduces Samsung as a direct participant in advanced AI chip production at the 2nm node, a segment that has been largely dominated by TSMC. Nvidia's parallel evaluation of Intel's 18A process for its Feynman architecture adds a second major AI hardware customer potentially diversifying away from exclusive TSMC dependence, compounding the strategic significance of Intel's foundry positioning heading into 2028.
For Alphabet, the capital and engineering commitment embedded in developing a 10th-generation custom TPU with a split-foundry architecture underscores the company's continued investment in proprietary AI silicon as an alternative to merchant GPU procurement. Alphabet's market capitalization stands at $4.36 trillion, and its Google Cloud segment is in direct competition with hyperscalers that rely heavily on Nvidia's GPU ecosystem — making custom silicon strategy a meaningful long-term cost and performance variable.
Sectors and assets to watch
Intel (INTC) is the most directly affected ticker. With a current market capitalization of $587.84 billion and annual revenue of $53.76 billion, Intel's foundry services unit has been seeking large-scale anchor customers to justify continued investment in advanced process nodes including 18A. A confirmed order of more than 3 million TPUs from Google, combined with Nvidia's evaluation of the 18A process for the Feynman architecture, would represent two of the most prominent names in AI hardware engaging Intel's manufacturing capabilities simultaneously. Investors and analysts will be monitoring whether these engagements translate into formal, multi-year production agreements.
Nvidia (NVDA), with a market capitalization of $4.96 trillion and revenue of $253.49 billion, sits in an adjacent position. Its evaluation of Intel's 18A process for the Feynman chip is exploratory at this stage, but any shift in Nvidia's foundry sourcing away from TSMC would carry significant implications for the competitive dynamics of advanced semiconductor manufacturing. Samsung Electronics, though not a primary ticker in this brief, is directly named as a potential foundry partner for Google's Icefish memory I/O die at the 2nm node, positioning it as a key company to monitor in the evolving AI chip fabrication landscape. TSMC, similarly not a primary ticker here, retains the compute engine role in Google's Icefish design, indicating its continued centrality even as customers pursue diversification at the component level.
What to watch next
Key developments to monitor include whether Google formally confirms the Intel TPU manufacturing agreement and provides any timeline or volume details beyond the initial report; whether Nvidia's evaluation of Intel's 18A process advances to a production commitment ahead of the Feynman architecture's 2028 target; and whether Samsung's discussions with Google regarding the Icefish memory I/O die result in a signed foundry agreement. Any formal announcements from Intel regarding hyperscaler customer wins for its foundry business would also serve as a material data point for assessing the commercial viability of its process node roadmap. Regulatory and geopolitical developments affecting advanced semiconductor manufacturing — particularly any changes to export controls or allied-nation chip production incentives — could further influence the pace and structure of supply chain diversification away from TSMC.