Building an AI data center already costs billions. Now operators are discovering that running one might cost even more than they planned, because the facilities are literally destroying themselves.
Power fluctuations at AI data centers are exceeding design capacity by up to 50%, according to a Bloomberg report. That kind of volatility is burning through batteries, generators, and cooling systems far faster than expected, turning what were supposed to be long-lived capital assets into expensive consumables.
The grid can’t keep up, and neither can the hardware
A single AI data center can demand power equivalent to what 100,000 households consume. When those workloads spike, the electrical infrastructure inside these facilities takes a beating that no one fully modeled for.
The result is premature equipment failure across the board. Backup batteries degrade faster. Generators cycle more frequently than their maintenance schedules anticipated. Cooling systems, already pushed to their limits by the thermal density of GPU clusters, face additional strain from the unpredictable power draw patterns.
These aren’t minor annoyances. They represent unforeseen operational expenditures layered on top of facilities that already require multibillion-dollar capital outlays to construct. For hyperscalers like Microsoft, Google, and Amazon, all of which have announced massive AI infrastructure buildouts, the calculus on return timelines just got more complicated.
US data center power demand is forecast to nearly double, climbing from 80 gigawatts in 2025 to 150 GW by 2028. For context, 150 GW is roughly equivalent to the entire electricity consumption of Spain.
Bitcoin miners smell opportunity in the wreckage
Bitcoin miners are pivoting aggressively into AI data center hosting, leveraging existing large-scale power contracts. Bitcoin mining already requires massive electrical capacity, and miners who locked in favorable power agreements now find that those same contracts are worth far more when repurposed for AI workloads.
Core Scientific illustrates this pivot better than anyone. The company has signed a $3.5 billion agreement to provide AI hosting services. Projections suggest that by late 2026, AI revenues could account for 70% of total revenue for some publicly listed miners, up from roughly 30%.
What this means for crypto investors
The power volatility problem creates a two-sided dynamic for anyone with exposure to crypto markets.
On the risk side, the escalating costs of operating AI data centers could squeeze margins for companies that rushed into the space without fully understanding the operational demands. If equipment replacement cycles are significantly shorter than projected, the economics of AI hosting contracts may not look as rosy as the headline numbers suggest.
On the opportunity side, the migration of Bitcoin miners into AI hosting could reshape valuations across the sector. Mining stocks have historically traded at steep discounts to their net asset value during bear markets. If these companies successfully transition into AI infrastructure providers, they may command valuations more in line with traditional data center REITs, which tend to trade at significant premiums.
For those watching the competitive landscape, the key metric to monitor is the ratio of AI revenue to mining revenue among publicly listed miners. As that number climbs toward 70% and beyond, the correlation between these stocks and Bitcoin’s price should weaken, potentially making them more attractive to traditional investors who want AI exposure without direct crypto volatility.