AI's Power Race Is Shifting Leverage from Chipmakers to the Grid

The primary bottleneck for artificial intelligence is no longer chips or talent, but electricity. As demand for power outpaces grid capacity, the advantage is moving away from hardware manufacturers and toward the utilities, grid operators, and power producers that control the supply.

The shift is already manifesting in regulatory and structural changes. On June 2, the Electric Reliability Council of Texas voted to overhaul how it admits large power users to the grid, addressing a backlog of data centers, crypto mines, and industrial sites competing for the same megawatts. Simultaneously, lawmakers in Albany, New York, moved to pass a one-year moratorium on new large-scale data centers, a move that could make the state the first in the country to pause the buildout.

The companies training frontier models are hitting a wall made of copper, concrete, and regulatory patience. While the previous decade focused on the supply of advanced GPUs, the current constraint is industrial economics. The limiting inputs are now land, generation capacity, water, high-voltage transformers, and local boards.

The scale of the projected demand is significant. Goldman Sachs expects US data center power demand to climb from 31 gigawatts in 2025 to 41 in 2026 and 66 in 2027. This surge is expected to lift the share of US peak summer demand from 4.1% to 8.5% over that same period.

However, the infrastructure may not be ready to meet these projections. Goldman Sachs noted that only about 50% to 60% of the capacity scheduled over the next year or two is likely to arrive on time, due to delays and cancellations. The grid is being asked to absorb in two years what it usually takes a decade to add.

The beneficiary of this demand is the entity at the end of the wire: the utility and the power producer. These companies now decide who gets electricity, when they get it, and at what price. As power becomes the scarcest asset for AI, the power companies—not the chipmakers—hold the strongest hand in the technology's expansion.

The central question for the next phase of AI development is whether the grid can be expanded fast enough to prevent a permanent plateau in compute capacity.

Source