Newsletter
Google will supply up to 1 million of its custom AI chips (TPUs) to Anthropic, in a deal reportedly worth tens of billions of dollars.
FYI, The agreement stands out as one of the biggest bets in the ongoing AI hardware race, highlighting how expensive it’s becoming to build advanced AI systems that need massive computing power to train and operate large language models.
The deets: the tensor processing units (TPUs), Google’s custom-built chips designed to accelerate machine learning tasks, will go live in 2026. They’re expected to add over 1 gigawatt of capacity, roughly enough to power an entire city.
The deal cements Google’s dual role as both a major investor, having pumped in over $3 billion so far and a core infrastructure partner for Anthropic. Meanwhile, Amazon, another Anthropic backer, has pledged up to $8 billion and provides its own AI chips through AWS.
Why it matters: as AI costs skyrocket, powerful chips have become the new oil everyone’s fighting for. Anthropic’s deal lets it reduce dependence on Nvidia’s GPUs, which dominate the market but are expensive and supply-constrained.
By locking in Google’s advanced TPU network, Anthropic gains scalability, while Google strengthens its cloud ecosystem and challenges Microsoft and Amazon in the AI infrastructure game.
Zoom out: Anthropic recently announced plans to open an office here, signalling how the country is fast emerging as a key talent and innovation hub in the global AI race.
In India, the AI chip market is still young but growing rapidly, valued at around $1.2 billion in 2024 and expected to touch $10–12 billion by 2033.
The government’s ₹76,000 crore semiconductor mission is pushing to build homegrown chip manufacturing capacity from the ground up.
