Is the Recent Microsoft, Nvidia and Anthropic Deal Evidence of an AI Bubble?

The artificial intelligence sector has become a magnet for global investment, drawing vast sums from technology giants and financial institutions. The latest deal to fuel this activity is a multi-billion-pound agreement between Microsoft, Nvidia and Anthropic, which has reignited speculation about whether the AI industry is entering bubble territory. Whilst the scale of the investment is unprecedented, its structure is equally striking, raising questions about whether capital is circulating within the AI ecosystem in a self-reinforcing loop.

The deal?

In November, Anthropic, the company behind the Claude models, announced an agreement involving Microsoft and Nvidia. Microsoft will invest up to five billion dollars, while Nvidia will contribute ten billion. Anthropic has committed to spend thirty billion on computing capacity from Microsoft’s cloud, which relies almost entirely on Nvidia chips, and may secure an additional gigawatt of compute, estimated to cost fifty billion dollars. This deal is part of a fundraising round valuing Anthropic at over three hundred billion dollars, remarkable for a company only four years old.

Why the Deal is Occurring

For Microsoft, the deal with Anthropic acts as a hedge against overreliance on a single AI partner. While OpenAI remains central to Microsoft’s AI strategy, powering tools such as Copilot in Microsoft 365 and Azure’s AI offerings depending solely on one provider carries risks. OpenAI’s independent governance, pricing decisions or potential partnerships with other cloud providers could limit Microsoft’s flexibility in the rapidly evolving AI market. By securing early and deep access to Anthropic’s Claude models, Microsoft diversifies its sources of cutting-edge AI technology, strengthens Azure’s position as the leading platform for frontier AI, and ensures it can continue delivering advanced AI capabilities even if OpenAI’s direction shifts.

For Nvidia, the incentive is similarly clear. As the dominant supplier of AI chips, it benefits from any expansion in global compute demand. Investing in Anthropic allows Nvidia to lock in one of the fastest-growing customers in the sector and influence the direction of future model development, ensuring sustained demand for its GPUs.

For Anthropic, the value of the deal is immediate and strategic: it guarantees stable, long-term access to the compute resources essential for frontier AI development. Training models like Claude require enormous computational power, and in today’s market, access to high-end GPUs and cloud infrastructure is both scarce and expensive. By securing dedicated resources from Microsoft’s data centres and Nvidia’s chips, Anthropic can scale its models efficiently, innovate without interruption, and remain competitive with other leading AI developers. The agreement effectively removes one of the most significant constraints on AI progress, allowing the company to focus on advancing its technology rather than sourcing the hardware needed to run it.

Individually, each party’s motives make sense. Collectively, however, they raise structural questions. 

The AI Bubble Debate

Concerns about an AI bubble have grown in 2025, not due to doubts about AI’s usefulness, but because of the scale of money and expectations. Technology companies are valued in the hundreds of billions, whilst compute purchases reach tens of billions, with total chip and infrastructure commitments potentially approaching trillion-dollar levels. Much of this investment is driven by the belief that whoever controls the most compute will dominate the next technological era, rather than proven commercial demand.

This echoes earlier speculative cycles, such as the telecom fibre build-out in the early 2000s or the shale boom in the 2010s. Companies invested heavily in infrastructure based on expected demand that did not fully materialise. Telecom firms laid fibre anticipating internet growth, only to find much of it unused. Energy companies invested in shale, expecting high oil prices, and many projects proved unprofitable.

Why This Deal Looks Like Evidence of a Bubble

What makes the Microsoft-Nvidia-Anthropic pact particularly striking is the circularity of the arrangement. The investors are also the suppliers. The suppliers are also the customers. And the customer, Anthropic, uses its funding to buy infrastructure from the very companies that invested in it. 

Microsoft invests in Anthropic → Anthropic uses that investment to buy compute from Microsoft → Microsoft buys chips from Nvidia → Nvidia profits and invests back into Anthropic → increasing Anthropic’s need for even more compute.

Satya Nadella stated plainly, “We are increasingly going to be customers of each other.” For critics, that sentence captures the essence of the bubble thesis. Value is being created within the ecosystem rather than from external market demand. If expectations falter, or if the pace of AI improvement slows, these self reinforcing cycles could unwind quickly.

A Necessary Counterpoint: Nvidia’s Strong Earnings

Yet Nvidia’s latest earnings suggest nuance. The company posted record results, with broad-based revenue growth across multiple customer segments and high demand for top-end chips exceeding supply. This indicates the AI infrastructure boom is not solely driven by circular investments, but by genuine, diversified market adoption. If this is a bubble, it is underpinned by substantial real economic activity.

Whether the Microsoft, Nvidia and Anthropic pact becomes a case study in strategic foresight or bubble psychology will depend on how quickly AI adoption spreads beyond early-stage hype. What is clear is that the deal is more than a funding round: it is a revealing snapshot of how interdependent the AI economy has become, and how that interdependence shapes both growth and risk.