Goldman Sachs Warns of Tightening Texas Grid as Data Center Demand May Cause Supply-Demand Imbalance
Alina Collins
Goldman Sachs warns that surging data-center and crypto-mining demand is pushing the Texas grid toward its limits — without faster supply growth, severe power shortages loom by decade's end, when Texas alone could account for nearly 40% of U.S. peak summer demand.
How fast is Texas power demand actually growing?
ERCOT's latest forecast puts baseline summer peak demand growth at 5.2% annually for 2026–2030, up sharply from the 3.4% average realized in prior years.
Include data centers and other large loads, and the growth rate jumps to 31% annually. This means → data centers are not adding marginal load — they are reshaping the entire demand curve.
Goldman projects that by 2030, Texas alone will account for 39% of U.S. summer peak demand, up from just 11% in 2024. In plain terms = one state consuming nearly two-fifths of the nation's peak power.
Can supply keep up?
ERCOT's large-load interconnection queue already totals 200 gigawatts. Even if only 10% materialize, demand growth would exceed 9%.
Effective generation capacity is forecast to grow at just 6% per year over the same period. This means → the supply-demand gap is arithmetically unavoidable — power plants cannot be built as fast as loads are plugging in.
The generation mix is uneven: solar keeps growing and gas responds flexibly, but nuclear is weakened by maintenance outages, hydro is dragged down by drought across more than half the country, and coal remains under pressure.
What do the voltage-test failures signal?
ERCOT disclosed that multiple batches of hyperscale loads and crypto facilities failed ride-through voltage tests — tests that simulate whether equipment can stay online during normal grid disturbances.
Just four clusters could shed over 5,000 MW during a routine transmission event — roughly equal to the entire city of Boston's electricity consumption.
Goldman compared this risk to the cascading failure in Spain's blackout: mass disconnection plus insufficient reactive power — the type of grid support that keeps voltage stable — can turn a manageable event into a systemic collapse. This reflects a risk beyond "not enough power": the grid also faces "a huge chunk drops off at once" stability threats.
Is this a Texas-only problem?
PJM has already raised its 10-year summer peak demand growth forecast this year from 3.1% to 3.6%.
MISO in April lifted its 20-year forecast from 1.6% to 2%, more than double the 0.8% average it realized in recent years.
U.S. commercial-sector power demand grew +1.8% year-on-year in January–February, the strongest of any sector. This means → AI-driven commercial electricity growth has moved from forecast to hard data.
Construction is accelerating — but is power actually reaching the grid?
Goldman notes that employment in data-center and power-infrastructure construction is climbing rapidly.
But job growth does not equal megawatts connected or transmission lines reinforced. In plain terms = sites are breaking ground, but that does not mean electrons are flowing — the gap between construction start and actual power delivery remains long.
Whether the supply side can accelerate sharply over the next several years is the pivotal variable for Texas — and the broader U.S. grid — in absorbing this wave of AI infrastructure expansion.
Content is for reference only, not financial advice.