Made with Midjourney
We’re witnessing one of the biggest capital expenditures in modern history. AI ‘hyperscalers’ are projected to spend $2.9T on data centers through to 2029, according to Morgan Stanley analysts. But Big Tech is only expected to foot about $1.4T of the bill, leaving a $1.5T funding gap.
Why the spending spree? AI models are compute-hungry. Training and running them requires exponentially more processing power than traditional cloud services. With the promise of superintelligent AI at stake, falling behind isn’t an option for the big tech players.
The scale of the spend is staggering: Meta’s “Prometheus,” xAI’s “Colossus,” and OpenAI’s “Stargate” each represent $100B+ investments in next-gen supercomputing power. Google, Amazon, Microsoft, and Meta are gearing up to spend over $400B on data centers in 2026 alone.
Debt financing is emerging as the preferred solution. About $60B of loans are going into roughly $440B of data center projects this year — twice as much debt as in 2024, according to law firm Norton Rose Fulbright. Private capital firms like Blackstone, Apollo, and KKR are competing aggressively to drum up cash for AI companies, with Meta recently landing $29B ($26B in debt) to fund data centers in Ohio and Louisiana.
But some industry watchers are sounding the alarm. As billions flood into AI, concerns about overcapacity, long-term profitability, and energy demands are growing. One risk is that data centers may become obsolete far quicker than we think, requiring new investment that decreases returns for owners or forces them to sell at a discount.