Anthropic Turns to Google Chips, Signaling Shift in AI Hardware Race

Anthropic's shift to Google TPUs signals rising competition to Nvidia and a broader move toward diversified AI infrastructure strategies.

In a move that could reshape the competitive landscape of artificial intelligence infrastructure, Anthropic has expanded its collaboration with Google and Broadcom to power its Claude AI models using Google's custom-built Tensor Processing Units (TPUs). The partnership marks a strategic shift away from an overwhelming dependence on Nvidia's GPUs, which currently dominate the AI hardware market.

For years, Nvidia has remained at the center of the AI boom, supplying chips that power everything from model training to deployment across leading tech companies. Organizations such as OpenAI, Meta, and Anthropic have relied heavily on Nvidia's high-performance GPUs to run large-scale AI systems. However, Anthropic's latest decision suggests that the industry may be entering a phase of diversification in its hardware choices.

Under the new agreement, Anthropic will gain access to multiple gigawatts of next-generation TPU capacity, with the infrastructure expected to go live starting in 2027. These chips will be used to run parts of the Claude AI ecosystem, enabling the company to scale its operations while reducing reliance on a single supplier.

Google, which has long used its in-house TPUs alongside Nvidia hardware, is now opening up its infrastructure to external partners. This shift not only strengthens its cloud and AI offerings but also positions it as a viable alternative in a market largely controlled by Nvidia. Broadcom's involvement further highlights the growing collaboration between chipmakers and AI firms seeking to build more resilient supply chains.

Despite this move, Anthropic is not cutting ties with Nvidia. Instead, it is adopting a multi-platform strategy that includes Nvidia GPUs, AWS Trainium chips, and Google TPUs. This diversified approach reflects a broader trend among AI companies aiming to mitigate risks associated with supply shortages, pricing pressures, and vendor dependency.

The timing of this partnership is significant. Demand for AI hardware continues to surge, driven by rapid advancements in generative AI and increasing enterprise adoption. Anthropic itself has witnessed remarkable growth, with its annualised revenue reportedly surpassing $30 billion in 2026, a sharp rise from approximately $9 billion at the end of 2025. This growth underscores the need for scalable and cost-effective infrastructure solutions.

The broader implications of this deal extend beyond Anthropic and Google. Nvidia, while still a dominant force, is beginning to face credible competition. Emerging initiatives, including large-scale chip manufacturing projects like TeraFab -- a vertically integrated facility backed by Elon Musk's ventures such as Tesla, SpaceX, and xAI -- signal that the industry is actively exploring alternatives.

While Nvidia is unlikely to lose its leadership position overnight, developments like these indicate a gradual shift in the balance of power. As AI companies continue to expand and innovate, the demand for diverse, high-performance computing solutions will only intensify.

Ultimately, Anthropic's partnership with Google and Broadcom reflects a calculated move to future-proof its infrastructure, while also contributing to a more competitive and dynamic AI hardware ecosystem.