Key Takeaways:
Google's commitment to manufacture more than 3 million tensor processing units on Intel's process and packaging technology marks the foundry's largest external AI chip order to date.
Intel secured a commitment from Alphabet's Google to manufacture more than 3 million tensor processing units, the foundry's largest external AI chip order, with production targeted for 2028. The deal covers Intel's advanced EMIB-T packaging technology and potentially its foundry and design services, according to a report from The Information.
"This is a storm in a teacup — these chips are still being fabbed at TSMC," JPMorgan analysts wrote in a note, arguing Intel's role is limited to packaging rather than full chip fabrication. Citi's Taiwan semiconductor analyst Laura Chen said the deal could extend beyond packaging to include Intel's foundry and design services, though she noted most buyside investors believe the report covers packaging only.
Nvidia is also running early trials on Intel's 18A process for potential future GPUs, according to people familiar with the matter. The 18A node (the company's most advanced process, equivalent to roughly 1.8nm in industry terms) is already in volume production for Intel's Core Ultra Series 3 PC processors, with yields running ahead of internal expectations, management said on the Q1 earnings call.
For Intel, the Google order provides multiyear demand visibility as its foundry unit — still loss-making and capital-intensive — races to become a credible second-source manufacturing partner to Taiwan Semiconductor Manufacturing Co. Intel shares, up 174% year to date, closed at $107.92, though they have fallen 13.6% over the past month as investors weigh the gap between headline customer wins and profitable execution.
The Foundry Credibility Gap
Intel's foundry business reported a non-GAAP net loss in Q1 fiscal 2026, even as overall company revenue rose 7% year over year to $13.6 billion and non-GAAP gross margin improved 1.8 percentage points to 41%. The segment requires billions in capital expenditure to build out capacity for external customers, and converting large orders into attractive margins is not guaranteed.
The Google deal sits alongside fresh alliances with Hitachi, Foxconn and Cadence, all aimed at tightening Intel's grip across design tools, manufacturing equipment and AI infrastructure. Together, these partnerships suggest Intel is being treated less as a legacy CPU supplier and more as a potential second-source manufacturing partner for AI hardware.
But competition from TSMC on manufacturing and from Nvidia and Advanced Micro Devices on AI chips remains intense. If Google or Nvidia limit volumes or keep Intel as a backup supplier, the revenue impact could be smaller than headlines suggest. TSMC's CoWoS (chip-on-wafer-on-substrate) packaging technology remains the industry standard for high-bandwidth AI accelerators, and Intel's EMIB-T alternative is still gaining adoption.
What to Watch
Investors should focus on whether more concrete contracts follow the Nvidia trials, how Intel discloses pricing and profitability for external foundry deals, and whether 18A and advanced packaging ramp without major delays. Any updates on capacity expansions, government co-funding and customer mix between internal products and external clients will help determine if Intel is becoming a durable alternative to TSMC or mainly capturing one-off AI projects.
Intel shares trade at roughly 99 times forward earnings, reflecting optimism that the foundry turnaround will eventually deliver margin expansion. The Google order gives the market a tangible demand signal to anchor those expectations — but the gap between a headline win and profitable volume production remains wide.
This article is for informational purposes only and does not constitute investment advice.
