The rapid expansion of artificial intelligence infrastructure has intensified concerns over electrical grid capacity. A single Nvidia GB300 rack reportedly draws over 180 kilowatts—roughly the electricity load of 60 average American homes contained within a single server cabinet. When scaled up to a full AI cluster exceeding 100 megawatts, the electrical demand rivals that of a small city.
Stark Projections and Supply Shortfalls
US data centers consumed approximately 176 terawatt-hours (TWh) in 2025, accounting for about 4.4% of total national power generation. Projections indicate this figure could rise to between 466 and 580 TWh by 2030, meaning data centers would consume 9% to 12% of all US electricity generation.
According to Goldman Sachs forecasts, US data center power demand is expected to climb from 31 gigawatts in 2025 to 66 gigawatts by 2027, more than doubling in two years. Meanwhile, utilities can realistically deliver an estimated 93 GW of additional practical power supply, creating a substantial shortfall relative to projected AI data center demand.
Physical Infrastructure Bottlenecks
The primary constraint facing the industry is physical infrastructure rather than chip availability. Critical grid equipment such as transformers and switchgear are in short supply, and manufacturing lead times for large power transformers have stretched significantly. This creates a cascading delay effect across the data center pipeline, where delayed transformers can stall projects regardless of whether servers, land, and contracts are secured.
These dynamics present distinct advantages for utilities with spare generation capacity, independent power producers near major data center corridors, and original equipment manufacturers building grid infrastructure. Conversely, energy constraints that slow data center buildouts could impact Nvidia, which relies on the grid keeping pace with its product roadmap.
Impact on Power Markets and Crypto Mining
In certain regions, data center power requests have been delayed or denied to prevent risks to grid reliability for existing customers. This situation mirrors challenges previously navigated by proof-of-work cryptocurrency miners, who have spent years managing power procurement and relocating for cheap electricity.
As a result, some crypto miners now hold valuable power purchase agreements and grid interconnections coveted by AI companies. This overlap has led to deals where mining operations pivot or lease their capacity to accommodate AI workloads.
