Columbia report: 182.7 GW of US data centers must yield $3.7T annual revenue by 2032
rohanpaul_ai · x · 2026-09-27
A Columbia Business School report, "Financing the AI Buildout," puts numbers on the $800B–$1T AI capex wave:
- Its central scenario is 182.7 GW of additional US data-center capacity between 2025 and 2032
- Using GB300 NVL72 configurations, that equals 1.07 million racks / 77 million GPUs
- The buildout requires $3.725 trillion in mature annual revenue — roughly $5.5 per installed GPU-hour at full utilization
- That sits below current $6–$10+/GPU-hour pricing for high-end NVIDIA capacity, making the central scenario economically plausible
The report frames the wave as hyperscalers contributing credit quality, outside investors providing capital, and frontier AI firms gaining compute they couldn't finance alone.
More from Venture
- Anthropic IPO multiple times oversubscribed, seeking up to $90.4B at ~$1.4T valuation — kimmonismus · 2026-11-11
- Not Boring: the three arguments that AI isn't a bubble are exactly how bubbles form — Not Boring (Packy McCormick) · 2026-11-11
- US data brokers selling $500M in RL training environments to Chinese AI labs — The Decoder · 2026-11-11
- CoreWeave: 2027 GPU renewals could add ~$4B annualized revenue, analyst skeptical — zephyr_z9 · 2026-11-11
- What do you check first when your agent's API bill doubles? Reddit debates cost tracing — rcleffke · 2026-11-11
- Polymarket gives 48% odds Anthropic's IPO closes above $1.5T market cap — Polymarket · 2026-11-11