Is There an A.I. Bubble? And What if It Pops?

What It Covers

Silicon Valley companies plan to spend nearly $3 trillion on AI infrastructure despite Wall Street concerns about a bubble, betting on artificial general intelligence while taking on unprecedented debt levels that could pose systemic economic risks.

Key Questions Answered

• •Infrastructure spending scale: OpenAI alone plans $500 billion for US data centers—equivalent to 15 Manhattan Projects or two Apollo programs—with global AI infrastructure spending projected at $3 trillion, far exceeding dot-com era investments.
• •Debt exposure risk: Companies building AI data centers take on massive leverage, with firms like Coreweave borrowing $3 billion for every $5 billion in infrastructure built. Morgan Stanley estimates $1 trillion of the total spending will be debt-financed.
• •Systemic risk opacity: Much AI infrastructure debt flows through private credit institutions and asset-backed securities similar to the housing crisis, making it impossible to track who holds the debt and creating potential for widespread economic contagion if revenues disappoint.
• •Dot-com parallel lesson: The 1990s internet bubble burst destroyed many companies but the fiber optic infrastructure they built eventually powered today's digital economy—Silicon Valley executives use this precedent to justify current spending despite acknowledging some will lose.