News
Michael Burry warns of declining returns on AI infrastructure investment amid chip spending boom
Thursday, January 22, 2026 at 01:51 PM
A discussion between Michael Burry and Anthropic co-founder Jack Clark highlights the tension between AI capital cycles and capability curves. Burry argues that the ratio of infrastructure spending to application revenue is imbalanced, with Nvidia selling $400 billion in chips against less than $100 billion in revenue, suggesting a collapse in return on invested capital. In contrast, proponents focus on expanding total addressable markets and the potential for recursive self-improvement in AI research.
Context
Michael Burry warns of a broken infrastructure-to-application ratio, noting Nvidia is selling $400 billion in chips while end-user revenue remains under $100 billion. He argues that Google and other hyperscalers are trapped in a competitive "escalator" dynamic, spending heavily on hardware to maintain parity without increasing profit margins. Burry specifically flags a duration mismatch in private credit, where assets are securitized for 20 years despite hardware replacement cycles of just 4 to 5 years, suggesting a looming correction as software companies become hardware-heavy.
Conversely, Anthropic co-founder Jack Clark contends that rapid capability gains will soon justify these costs. While Burry focuses on declining Return on Invested Capital, optimists see the addressable market expanding from the $400 billion ad industry to the multi-trillion dollar labor market. This creates a stark divide for 2026 investors: either the current buildout is a FOMO-driven bubble with no durable advantage, or it is an essential investment in recursive AI before the next capability leap.
Related Companies
Nvidia
NVDA
Google
GOOGL