Anthropic expands AI chip partnerships with Broadcom and Google

Anthropic is expanding its AI infrastructure through deeper partnerships with Broadcom and Google, as demand for compute continues to scale with the development of advanced models.

The collaboration centers on securing access to next-generation custom AI chips, including Google’s Tensor Processing Units (TPUs), with Broadcom playing a key role in chip design and supply. As part of the agreement, Anthropic is expected to gain access to several gigawatts of compute capacity starting in 2027.

This marks one of the largest infrastructure commitments by the company to date, reflecting the rapid growth in usage of its Claude AI models and increasing enterprise demand for large-scale AI systems.

The partnership also highlights a broader industry shift toward custom silicon. Rather than relying solely on third-party GPUs, AI companies are increasingly working with chipmakers and cloud providers to build specialized hardware tailored to their workloads.

For Google, the deal reinforces its position as both a cloud provider and a hardware ecosystem player, using its TPU architecture to attract AI developers. For Broadcom, the collaboration strengthens its role in designing and manufacturing custom AI chips for hyperscalers.

From an AI and martech perspective, access to large-scale compute is becoming a defining competitive factor. As models grow more complex and enterprise adoption accelerates, partnerships across chip design, cloud infrastructure, and AI development are becoming increasingly interconnected.

Anthropic’s move signals how frontier AI companies are shifting from experimentation to long-term infrastructure planning, where securing reliable and scalable compute capacity is critical to sustaining growth.