Goldman Sachs: Even if projects are delayed or canceled, the electricity demand for U.S. data centers will double within two years

Ask AI · What are the two-way risks in Goldman Sachs’s forecast data center growth?

The AI wave is reshaping the U.S. power landscape — data centers may become the core driver of U.S. electricity demand growth over the next two years.

According to WindTrader, on May 5th, Goldman Sachs’s commodities research team released the latest electricity market research report, which, based on detailed project development progress data from the data tracking platform Aterio, makes a systematic forecast of U.S. data center electricity demand.

The firm believes that, even if many projects are delayed or canceled, U.S. data center electricity demand will still double within two years.

Data center expansion accelerates, far surpassing historical levels

First, look at the basic data.

According to Aterio data, the new capacity added to U.S. data centers in 2024 and 2025 will be 6.4GW and 8.5GW respectively. Based on current development plans, the annual new additions in 2026 and 2027 will jump to 19GW and 69GW — a significant leap in magnitude.

Goldman Sachs points out that growth is not just a story at the national level. In the three major regional power markets — PJM (Mid-Atlantic), ERCOT (Texas), MISO (Midcontinent) — each’s planned new additions in 2027 will exceed the total new additions across the U.S. in 2025.

Delays and cancellations are real, but cannot stop the overall trend

Plans are plans; whether they come to fruition is another matter.

Goldman Sachs compared the development plans from December 2024 with actual commissioning data afterward and found: Only 72% of data centers scheduled to be commissioned in the next four quarters actually came online on time.

The main reasons for delays are threefold:

1. “Site selection battles”: Developers often submit applications in multiple regions to hedge risks, ultimately only advancing the most favorable one;

2. Supply chain and labor issues: Waiting for electrical equipment during construction can cause months of work stoppages;

3. Construction cycle itself: It typically takes 1.5 to 2 years from approval to commissioning for a data center, and the later the plan, the lower the probability of realization.

Based on historical patterns, Goldman Sachs estimates that: Currently, about 60% of the new capacity planned will be realized on time within the next year, decreasing to about 50% over the next two years.

Even with discounts, demand growth remains astonishing

After applying these discount factors into their model, Goldman Sachs’s forecast numbers remain quite robust.

Specifically:

• 2025: 8.5GW of new capacity added throughout the year; 2.2GW already achieved in Q1 2026;

• Q2 to Q4 2026: an additional 11.5GW expected;

• 2027: annual new additions will reach 36GW;

• By the end of 2027, U.S. data center total installed capacity will reach 95GW, more than double the level at the end of 2025.

Corresponding to electricity demand (calculated at 70% capacity utilization, consistent with the five-year historical average):

• 2025: 31GW

• 2026: 41GW

• 2027: 66GW

It is worth noting that predictions from different Goldman Sachs teams vary. The equity research team, based on S&P 451 Research data, forecasts 39GW and 50GW for 2026/2027 respectively; the commodities research team, based on Aterio data, forecasts 41GW and 66GW. Goldman Sachs admits in the report that differences in data sources and methodologies are the main reasons for these discrepancies, and the overall market forecast range is quite broad — from the U.S. Department of Energy/Lawrence Berkeley National Laboratory’s approximately 30 to 55GW, to McKinsey’s 42GW, and Boston Consulting Group’s 67GW, estimates vary widely.

By 2027, data centers will account for 8.5% of the U.S. summer peak electricity demand, up from just 4.1% in 2025.

Power market regional divergence: some “thirst for electricity,” others relatively relaxed

The explosive growth of data centers will not be evenly distributed, and the impact on regional power markets will vary significantly.

The report categorizes regions into three groups:

1. Increased reliability risk (most tense): PJM (Mid-Atlantic, covering Virginia, Ohio, etc.), MISO (Midcontinent), NW (Northwest) — these markets have limited new generation capacity but are experiencing surging data center demand. Goldman Sachs believes these markets may have to reject some data center connection requests in the future and recommends hedging against rising local electricity prices.

2. Marginal tightening (relatively relaxed): ERCOT (Texas), SPP (Mid-South), SE (Georgia) — these regions have faster-growing generation capacity, and data center demand will only cause moderate tightening.

3. Saturated and limited new additions: TVA (Tennessee), ISONE (New England), FL (Florida) — these markets are already under critical tension, with very limited capacity for new data centers.

Goldman Sachs notes that this differentiation aligns closely with data center site selection logic: electricity availability is the primary consideration. This explains why Texas and Georgia attract a large number of new projects; despite already tight power supply, markets like PJM continue to attract investment due to their “close to customers” geographic advantages (especially Virginia and Ohio).

Forecasts carry two-way risks; Goldman Sachs recommends hedging both

Analysts explicitly list sources of forecast uncertainty:

• Upside risks: As time progresses, new projects will continue to enter development plans, and actual new additions may exceed current forecasts; high capital expenditure could also compress construction cycles to within a year;

• Downside risks: Supply chain and labor issues could prolong construction; the relatively short historical data (only since December 2024) and rapid market environment changes increase forecast uncertainty; additionally, many data centers do not publicly disclose information, leading to data blind spots.

Based on these two-way risks, Goldman Sachs’s strategic advice is: Hedge against both upward and downward electricity price risks — guard against soaring prices in tight markets like PJM, and protect against downward pressure in markets like ERCOT with faster supply growth.