Anthropic Secures Google AI Chips to Advance Claude Chatbot
- tech360.tv
- 25 minutes ago
- 2 min read
Anthropic is significantly expanding its collaboration with Google, securing access to up to one million of the tech giant’s artificial intelligence chips. This deal, valued at tens of billions of USD, aims to accelerate the development of its Claude AI models in a competitive market.
Under the terms of the expanded deal, Anthropic will gain access to over one gigawatt of computing capacity. This capacity is scheduled to come online in 2026.
The computing power will be used to train future generations of its Claude AI model on Google’s in-house tensor processing units, or TPUs. These TPUs were traditionally reserved for Google’s internal use.
Anthropic selected TPUs due to their price-performance ratio, efficiency, and its existing experience in training and serving its Claude models with the processors.
This agreement highlights the intense demand for chips within the artificial intelligence industry. Companies are rapidly developing technology to potentially match or surpass human intelligence.
Alphabet-owned Google will also provide additional cloud computing services to Anthropic. Google Cloud offers TPUs for rent, providing an alternative to the supply-constrained Nvidia chips.
Rival OpenAI recently secured multiple deals that may cost over USD 1 trillion. These deals are for about 26 gigawatts of computing capacity, enough power for roughly 20 million US homes.
Industry executives have indicated that one gigawatt of compute can cost approximately USD 50 billion. ChatGPT-maker OpenAI actively utilises Nvidia’s graphics processing units and AMD’s AI chips to meet its increasing demand.
Reuters exclusively reported Anthropic is projecting a substantial increase in its annualized revenue run rate next year. This could involve more than doubling and potentially nearly tripling its current revenue.
The projected growth is fuelled by the rapid adoption of Anthropic’s enterprise products. The startup focuses on AI safety and creating models for various enterprise use cases.
Anthropic’s models have also played a role in powering the boom of vibe coding startups, including Cursor.
Anthropic is expanding its deal with Google to access up to one million AI chips.
The agreement, valued at tens of billions of USD, will support the training of future Claude AI models.
Anthropic will gain over one gigawatt of computing capacity, available from 2026, using Google’s TPUs.
Source: REUTERS