Google (Alphabet) has placed an order with Intel to manufacture over three million tensor processing units (TPUs) designed in-house for AI and neural network workloads. These chips are scheduled for production and delivery in 2028 [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11].
The news broke on June 8, 2026, triggering a surge in Intel's stock price by 9% to 13% during pre-market and early trading [1, 2, 3, 4, 7, 11]. DA Davidson analyst Gil Luria noted that beyond diversifying supply chains, Google and Nvidia are motivated to work with Intel because "supporting Intel effectively means supporting U.S. domestic manufacturing," which is crucial to maintaining good relations with the U.S. government [11].
Nvidia is currently evaluating Intel’s technology aimed at producing a processor that combines four graphics chips into a single unit but has not yet placed a formal order [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11].
Intel is gaining this major business opportunity amid severe capacity constraints at Taiwan Semiconductor Manufacturing Company (TSMC), which is struggling to meet hyperscale AI chip demand. TSMC’s chairman Wei Zhejia said, "global chip supply will not keep pace with AI demand growth in the coming years, even as TSMC expands capacity in the U.S., it will be hard to meet American customers' needs" [8].
AI's biggest players are actively diversifying a supply chain still heavily concentrated at TSMC, according to eMarketer analyst Jacob Bourne. Many major customers seek second-source suppliers like Intel to fill the supply gap [4, 7, 8, 9, 10, 11].
Intel's packaging technology (EMIB), offering a potentially lower-cost alternative to TSMC’s CoWoS, is under testing by partners like SK Hynix for memory products [8, 9, 10]. Intel is also collaborating with Tesla on chip manufacturing using its next-generation 14A process, unveiled as part of Tesla’s Terafab AI chip project announced in April 2026 [1, 4, 11].
Production of the Google TPU chips by Intel is targeted for 2028 delivery [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11].
