Google sees a chance to compete with Nvidia in the AI chip arena, but it is not an easy task

New players in the AI accelerator segment

Until recently, Nvidia’s main competitors in the accelerator space were considered to be AMD and, at best, Intel. However, Google, which is developing its own TPU (Tensor Processing Unit) processors, now occupies a significant position in this market.

1. Google as an Nvidia consumer

- Over the past three years, Google CEO Sundar Pichai has mentioned using Nvidia chips in 10 out of 12 public appearances at quarterly earnings conferences.

- This confirms that Google still relies on Nvidia solutions for its computing infrastructure.

2. Turning to its own TPUs

- The release of Gemini 3 changed the picture: the company announced that it trained this AI model exclusively on its TPU processors.

- The success of Gemini 3 became a kind of “advertisement” for TPUs, and now Google is negotiating supplies of these chips to third‑party AI system developers.

3. Expanding interest in TPUs

- Until recently, TPUs were used only within Google, but reports have emerged that Meta Platforms is showing interest in them.

- At the same time, Meta signed a contract for Nvidia CPU supply, demonstrating flexibility in technology choices.

4. Financial support for startups

- Google plans to invest up to $100 million in startups ready to use TPUs to build server systems (example – Fluidstack).

- Similar collaboration schemes have already been applied by Nvidia: investments in CoreWeave, Nebius and Lambda.

5. Production capacity issue

- In the AI‑chip market it is difficult to “push” Nvidia due to a shortage of contract fabs.

- TSMC, the largest chip supplier, already serves Nvidia as its main client; even Apple struggles to obtain the required volume.

- In such a situation Google will not be able to simultaneously meet its own needs and sell to customers through TSMC.

Conclusion

Google is moving from being just an Nvidia accelerator consumer to playing an active role as a TPU supplier for external clients. This intensifies competition in the AI‑accelerator segment and creates new challenges for existing leaders due to limited manufacturing resources.