Google Unveils Two New TPUs to Cut Nvidia Dependence

Google unveiled two new TPUs, TPU 8tx and TPU 8i Max, on Monday, signaling a push to cut its reliance on external chip makers for cloud AI. The move comes as hyperscalers like Microsoft, Amazon, OpenAI and Anthropic race to build their own AI‑specific semiconductors. Nvidia, the current market leader, is under scrutiny after its CEO’s recent scandal.

Google’s new chips could tilt the balance in the hardware war. TPU 8tx, built with AMD, embeds security directly into the hardware layer, keeping model weights and data isolated even while in use. The chip also runs a custom inference network that lets it reason while training, enabling on‑the‑fly distillation of small models.

This means a company can train on proprietary data without moving it between devices, a feature that could protect sensitive information during AI development. TPU 8i Max stitches together two of the earlier TPU 8i chips into a single package, doubling memory capacity, processing power and bandwidth. The result is a chip that lets AI models handle a larger range of information during queries.

By packaging the two chips together, Google simplifies deployment while boosting performance for large workloads.

The announcement underscores Google’s strategy to internalize AI hardware, but it also highlights the intense competition in the space.

While Nvidia remains dominant, its recent leadership controversy could weaken its position.

Our read: Google’s new chips may reduce its dependence on Nvidia, but the company still faces the challenge of scaling production and convincing customers to switch.

As the AI race heats up, the real test will be whether these chips can deliver the promised performance gains in real‑world workloads.

Watch how Google’s TPU 8tx and TPU 8i Max perform in cloud deployments and whether they prompt other hyperscalers to accelerate their own silicon projects.

The outcome could reshape the balance of power in AI hardware.