Gartner: AI data centers need $8.5 trillion to break even

The debate over whether AI is a bubble has been running since ChatGPT launched in late 2022, with warnings repeatedly triggered by milestones like Nvidia's market cap crossing $2 trillion, $3 trillion, $4 trillion, and $5 trillion, or by events like the DeepSeek shock in early 2025. Each time, earnings kept soaring and stocks rebounded to record highs, which is why Solus strategist Dan Greenhaus dismissed CoreWeave's recent stock tumble and surge in credit default swaps as just another false alarm. But the current anxiety is different in focus: it centers on whether the trillions being spent on AI infrastructure will generate enough revenue to justify the debt taken on by companies like CoreWeave. While Nvidia CEO Jensen Huang and CoreWeave CEO Mike Intrator expressed confidence, Gartner's estimate that data centers will need $8.5 trillion through 2030 against at most $3.6 trillion in revenue points to a significant gap. Bank of America's comparison to past capital cycles, where early high margins led to overinvestment and eventual disappointment, adds weight to concerns that the spending may not pay off as hoped.