AI Bubble Anxiety
AI euphoria is shifting into anxiety. The industry that once looked unstoppable is now getting hit with hard questions about debt, sustainability, and the real economics of building artificial intelligence at scale. Short sellers like Michael Burry and Jim Chanos are warning that the current spending cycle is not a one-time surge but a recurring and expensive commitment. Behind the scenes, debt is becoming the quiet engine powering the boom, with loans, bonds, and securitized financing now standard tools to keep the infrastructure race moving. CNBC’s Deirdre Bosa has been tracking these signals, and leaders inside the sector are beginning to weigh in on what they’re seeing from the ground level.
Debt Is Becoming the Foundation of the AI Buildout
The most important shift in the AI economy is how the buildout is being financed. Instead of relying primarily on cash flow, many companies are tapping credit markets at an aggressive pace. Corporate bonds linked to AI infrastructure have surged into the hundreds of billions. Data centers are increasingly financed through asset-backed securities and commercial mortgage-backed structures, and private credit funds are stepping in with mezzanine loans when traditional lenders hesitate. This debt-driven expansion allows companies to scale rapidly, but it also creates a leverage cycle that depends on strong and sustained demand for AI products. If energy costs rise, if GPU prices stay high, or if enterprise adoption slows, companies could find themselves with shrinking margins and rising debt service costs. Credit ratings agencies have already flagged the risk that some firms are signing massive AI contracts without fully addressing their longer-term leverage.
Short Sellers Are Highlighting Understated Costs
Michael Burry has publicly argued that large tech companies are dramatically underestimating depreciation on their AI hardware. He estimates that hyperscalers may understate depreciation by tens of billions of dollars over the next several years, which would make earnings appear stronger than they really are. Jim Chanos has similarly warned that the AI spending boom mirrors the pattern of past bubbles, where infrastructure spending surged long before the economic payoff materialized. The warning is simple and blunt. This buildout is expensive, these assets age fast, and if companies miscalculate the pace of technological obsolescence, they will end up with enormous write-downs. Investors who focus only on revenue growth without examining the debt behind it may miss the larger risk.
What Amjad Masad Sees Inside the Compute Economy
Replit CEO Amjad Masad has been direct about the pressures facing AI builders. He has repeatedly warned that compute is not getting cheaper in any meaningful way. GPU supply remains tight, energy costs remain high, and the operational overhead of running large AI systems continues to rise. Replit’s own margins have fluctuated between negative and positive territory as model costs surge and stabilize unpredictably. Masad says the industry needs to stop assuming that Moore’s Law-style cost declines will save them. In his view, companies must design products and business models that reflect the real economics of compute, rather than the idealized version used in investor presentations. He has also argued that the role of coding itself is changing, emphasizing that creativity and problem-solving will matter more than traditional programming as AI systems generate a larger share of code.
Navrina Singh Warns of a Different Kind of Debt
Credo AI CEO Navrina Singh has introduced a separate concept into the debate: governance debt. She argues that as companies sprint to deploy AI systems, many are doing so without adequate oversight, safeguards, or transparency. That lack of structure becomes a form of hidden liability. Singh has testified before Congress and spoken repeatedly about the financial risks of scaling AI without proper governance. She highlights the threat of regulatory backlash, reputational damage, and systemic risk that can arise when companies treat responsible AI as optional. In her view, the sustainability of the AI boom will depend on better alignment between innovation teams and compliance frameworks. Without it, the industry risks building models faster than it can secure them.
The Bottom Line for Investors and Founders
The AI boom is not slowing, but the foundation beneath it is changing. What once looked like an unstoppable rise now carries clear financial and operational warning signs. Debt is fueling growth. Depreciation is growing harder to ignore. Costs are proving stubborn. And governance gaps are turning into real liabilities. AI founders are not blind to these pressures. Some are sounding the alarm early, not because the industry is collapsing, but because the next phase will require more discipline, more transparency, and more realism about what it truly costs to build and run artificial intelligence at scale.
