Investors are questioning the sky-high valuations of frontier AI firms like Anthropic and OpenAI. Skepticism is growing over 'circular deals'—interdependent financing loops where tech giants and startups invest in each other—and whether massive infrastructure spending can deliver actual profits.
The global artificial intelligence sector is facing a reality check as investors grow increasingly skeptical about the sustainability of multi-billion dollar valuations for top-tier AI labs. While companies like Anthropic and OpenAI command private valuations of $965 billion and $852 billion respectively, the financial structure supporting these numbers is drawing sharp scrutiny.
At the heart of the concern is the reliance on what analysts call circular deals. These are interdependent financial arrangements where massive amounts of capital flow in a closed loop. For instance, large tech giants invest heavily in AI labs. Those labs, in turn, spend a significant portion of that capital on chips and compute power from hardware manufacturers. In some cases, the chip manufacturers then reinvest in the ecosystem or secure long-term supply agreements. This cycle can artificially inflate both revenue and valuation, making it difficult for investors to determine if the AI business model is genuinely profitable or simply fueled by the capital being pumped into it.
The Capex Challenge
The scale of capital spending required to build this AI infrastructure is unprecedented. Major U.S. technology companies—including Alphabet, Amazon, Microsoft, and Meta Platforms—are projected to pour between $720 billion and $760 billion into capital projects in 2026. Investors are now asking a critical question: when will this massive spending translate into clear, measurable returns?
Currently, there is a disconnect between the optimistic growth projections and the actual productivity gains seen in the real world. While predictive AI is already delivering value in areas like logistics and fraud detection, the newer generative and agentic AI tools are facing complaints regarding high costs, errors, and the inability of these tools to replace human roles effectively without incurring high hidden costs. The European Central Bank has recently flagged the potential for a valuation reset if earnings fail to catch up with these aggressive growth expectations.
Risks of Interdependence
The deeply connected nature of these financial deals creates a systemic risk often referred to as a potential domino effect. Because the industry relies on a tight web of agreements between labs, cloud providers, and chip makers, the failure of a single major participant could trigger financial instability across the entire sector. If the supply of computing power or the funding for these labs were to dry up, it could rapidly erode the multi-trillion dollar valuation targets that many market players are currently assuming.
For investors, the key monitorable is no longer just the size of the funding round or the revenue run rate. Instead, the focus is shifting toward demonstrable business value. As the industry moves into a phase where the market demands tangible proof of efficiency, investors will likely track whether the cost of running these AI models—often referred to as 'token costs'—can drop enough to allow for sustainable, profitable operations without the constant need for fresh capital injections.