Science & Tech · AI · 4 hrs ago
Google adds specialised AI chip versions to challenge Nvidia
Google is expanding its range of AI chips as it seeks to compete with Nvidia in the market for processors used to develop and run artificial intelligence.
The new versions, called TPU 8tx and 8i Max, are reconfigured versions of existing Google chips rather than entirely new designs.
The 8tx is designed for confidential computing, including work by government agencies and companies in regulated industries, and works with chips from AMD.
The 8i Max is built for AI tasks that need long context windows, meaning the system can consider more information in a single request; possible uses include drug discovery, mathematics and coding.
Google’s chips compete with Nvidia’s widely used processors, and the company already sells them beyond its own cloud business to customers including Apple, Meta and Oracle.
Google says the specialised chips will be available in 2027, as demand for AI processors continues to give competing suppliers room to grow.
Google is introducing two specialised versions of its tensor processing units as it seeks to expand its challenge to Nvidia in the AI chip market.
The TPU 8tx is designed for secure, confidential computing, while the 8i Max is intended for AI applications that need to process long-context requests.
Google says the new versions respond to customer needs and reflect a shift towards specialised chips and AI agents.
The variations are due to be available in 2027.
Google chip chief Amin Vahdat said the company could deliver more value by building specialised versions rather than one chip for the whole market.
- Who
- Alphabet’s Google is introducing the chips. Google chip chief Amin Vahdat discussed the plans.
- What
- Google is adding the TPU 8tx and 8i Max, specialised versions of its AI chips.
- When
- The variations were due to be announced on Nov 9, 2026, and will be available in 2027.
- Where
- Not stated
- Why
- Google says the chips address customer needs for confidential computing and AI applications that require long-context processing.
This story does not have two clearly opposing sides.
Google began developing TPUs alongside Broadcom to handle its growing computing needs.
Google put out two AI processors at the same time.
Google’s two latest TPU variations were due to be announced.
The TPU variations are expected to become available.
Bloomberg Intelligence projected TPU sales, including related software and hardware, would exceed US$135 billion.
- New chip variations
- TPU 8tx and 8i Max
- TPU 8tx purpose
- Secure, confidential computing
- 8i Max purpose
- AI applications requiring long-context windows
- Availability
- 2027
- Estimated TPU sales
- US$12 billion in 2026
- Projected TPU sales
- More than US$135 billion by 2028, including related software and hardware
What people are saying
This really follows the thesis that we have here at Google, which is that the future is going to be about specialisation, and the upsides for specialisation continue to be significant.
We could build a single chip that tries to address the whole market or, by building specialised versions of the chips, we actually can deliver much more value.