Anthropic Taps Google’s AI Chips for Claude Expansion

Anthropic has expanded its partnership with Alphabet’s Google (GOOGL). The company will gain access to up to one million AI chips and more than a gigawatt of computing power. This multi-year deal, starting in 2026, gives Anthropic massive capacity to train its Claude chatbot models on Google’s custom tensor processing units (TPUs).

Google opens its AI hardware vault to outsiders

The agreement lets Anthropic use tens of billions of dollars’ worth of Google hardware—once reserved for internal projects. Google’s TPUs, available through Google Cloud, are built to deliver strong price-to-performance gains compared with Nvidia’s (NVDA) GPUs, which remain the industry standard. Anthropic said it chose TPUs because of their efficiency, reliability, and its positive results using them in earlier Claude models.

This deal underscores the growing battle for AI infrastructure, now seen as a scarce and strategic resource. Anthropic’s allocation of more than a gigawatt—about the size of a mid-sized power plant—puts it among an elite few. Only major AI players like OpenAI have that level of access.

Compute becomes the new oil

We view this as a major step in the hardware race driving AI development. Every big model upgrade—from OpenAI’s GPT-5 to Anthropic’s Claude 4—demands exponentially more compute. Specialized chips now determine how fast companies can innovate. Google’s move to rent out its prized TPUs shows that AI hardware is shifting from a private asset to a key revenue stream.

For investors, the message is clear: the AI market is clustering around a few compute-rich players. Google gains a strong, long-term customer and proof of its cloud strength. Anthropic solidifies its role as OpenAI’s main rival. The broader shift is that the main constraint in AI is no longer algorithms—it’s energy and hardware. In short, control over compute means control over AI’s future.

Here’s what this means for Big Tech stocks

This deal looks bullish for Alphabet. Renting TPUs through Google Cloud turns idle hardware into steady revenue while boosting Google’s standing as an AI cloud provider, not just a model builder. It also helps narrow the gap with Amazon (AMZN) and Microsoft (MSFT), both competing for AI startups through their cloud offerings.

Nvidia still benefits from surging AI demand, but its dominance in GPUs exposes a risk—dependence on one supplier. Anthropic’s shift to Google chips adds variety to the market and could pressure Nvidia’s profit margins if TPUs prove cheaper. AMD (AMD) also faces challenges as large cloud providers build their own custom chips.

We see this as the start of a new AI hardware supply chain focused on energy, chips, and cloud capacity instead of just software quality. The winners will be those who own or rent enough compute power.

In the end, this agreement underscores that AI chips are becoming the true currency of AI innovation.

Google (GOOG), Nvidia (NVDA), and Microsoft (MSFT) have a Disruption Score of 5 and are part of the Disruption Aristocrats.

Amazon (AMZN) and AMD have a Disruption Score of 3 and 4, respectively. Click here to learn more about our scoring system.