Intel secures Google AI chip contract while Micron's memory shortage fuels US policy debate

Intel announced a contract to manufacture three million Google Tensor Processing Units (TPUs) on its advanced 3‑nanometre node, with production partly in the United States. The deal is seen as a step toward reducing reliance on Taiwan's TSMC and aligns with the US CHIPS Act’s push for domestic chip sovereignty. Nvidia is also evaluating Intel's 18A process for future multi‑die GPUs, further expanding Intel's foundry prospects.

Micron, now valued at over $1 trillion, is at the centre of a high‑bandwidth memory (HBM) shortage that powers AI accelerators. The company has sold its entire HBM production for fiscal year 2026 and is expanding capacity with a new fab in New York and a wafer line in Idaho, an investment exceeding $200 billion. The scarcity has drawn political attention in the United States, with trade groups warning regulators about allocating AI data‑center memory across sectors. The combined developments highlight a shifting AI‑chip supply chain and growing geopolitical stakes in semiconductor manufacturing.