The AI buildout has grown so interconnected that a single point of failure could ripple across the multitrillion-dollar sector, according to a new analysis from London-based Sona Asset Management. With AI investment powering the U.S. economy, any stumble would likely carry broad market consequences - one reason the Trump administration has pursued an anti-regulatory approach to avoid crashing markets, as the New York Times reports.
Echoes of the financial crisis
The current AI ecosystem mirrors the pre-2008 mortgage market in troubling ways, the Sona authors argue. They point to "opaque and concentrated exposures, counterparties linked in complex ways that few have mapped, and demand part-underwritten by the same balance sheets that depend on it." The warning arrives as S&P cautions that hyperscaler credit quality is weakening and borrowing costs rise across the sector, which carries nearly $6 trillion in debt.
The analysis maps a fast-growing supply chain of 255 public companies: hyperscalers like Microsoft and Meta, chipmaker Nvidia, plus smaller data center operators and neoclouds that rent computing power. Two giant AI labs - OpenAI and Anthropic - sit in the mix. Together these firms hold a combined market cap of $50 trillion, more than double what it was five years ago.
A closed-loop market
Money circulates among the same group of companies in what the authors describe as a closed loop. They finance one another, buy from one another, and invest in each other. "Capital, product and demand chase each other around the same handful of names," the authors wrote.
Some smaller businesses earn the bulk of their revenue from just one or two large players. CoreWeave draws about 67% of its revenue from Microsoft alone. Applied Digital, a data center infrastructure company, gets 56% from Oracle and 30% from CoreWeave - which is itself heavily dependent on Microsoft. That circularity is not inherently problematic, the authors note, comparing it to financing a car through a dealer.
Where the risk concentrates
The core of the ecosystem - the big hyperscalers, chipmakers, and memory companies - remains in strong financial shape. They generate significant cash and hold solid credit ratings. Unlike the mortgage crisis, individuals and their homes are not on the line. Many of these companies build and sell real, physical assets rather than creating synthetic leverage.
The financial risks sit "one ring out from the core," with neoclouds and data center platforms that carry the highest leverage, thinnest margins, and weakest cash flows. A single investment decision by a larger company may be "existential" for these firms, the authors write. They face particular danger if tech advances cause the price of compute to fall.
Why this matters for finance and strategy executives
The concentration risk mapped in this analysis should sharpen credit assessments across the AI supply chain. Finance leaders with exposure to data center operators or neoclouds need to stress-test scenarios where a major hyperscaler shifts procurement strategy or compute pricing drops. The circular revenue patterns - where companies are simultaneously customers, suppliers, and investors in one another - create balance-sheet interdependencies that standard risk models often miss. For CFOs and strategy heads evaluating AI infrastructure investments, AI for CFO Training offers frameworks for assessing these layered counterparty risks. The Sona paper underscores that strong top-line growth in the sector can mask dangerously thin margins and concentrated customer bases further down the chain.
Your membership also unlocks:
