Goldman Sachs sees US data center power demand surging on AI buildout
Goldman Sachs projected US data center power demand to hit 66 GW by 2027, but warned grid bottlenecks and rising electricity prices could slow AI deals.
.png&w=1920&q=75)
Goldman Sachs said the AI buildout is creating a new power market fast enough to reshape deal flow, but it is also creating execution risk for clients trying to secure electricity on time and at a predictable price. The bank projected US data center power demand would rise from 31 GW in 2025 to 41 GW in 2026 and 66 GW in 2027, enough to make data centers 8.5% of total US power demand by then.
That forecast rests on a sharp expansion in US data center capacity, which Goldman said would reach roughly 95 GW by the end of 2027, more than double the end-2025 level. The bank also assumed a 70% capacity utilization rate. In practical terms, that means the market still has to clear a large amount of power, equipment and interconnection work before the demand Goldman expects can actually be served.

For Goldman’s own franchise, the surge points to more business for power, utilities, infrastructure, financing and M&A bankers as hyperscalers, developers and governments look for ways to fund and secure generation. Goldman has said financing solutions in public and private markets are becoming increasingly important, and that hyperscalers and governments are taking a diversified approach to procuring energy from multiple sources. But the same scramble creates headaches for clients and employees alike: grid capacity is tight, electricity prices are rising and project timing is becoming harder to control.

Goldman has already warned that electricity prices are likely to keep climbing because data-center demand is growing faster than supply. In February 2026, CNBC reported that Goldman analysts told clients electricity prices had jumped 6.9% in 2025 from a year earlier, more than double headline inflation. The bank said investors should hedge both upside and downside power price risk, a sign that the energy trade around AI is already forcing harder conversations in trading, structuring and risk management desks.

The broader backdrop shows how quickly the load is compounding. The US Department of Energy said data centers used about 4.4% of US electricity in 2023, while a DOE and Lawrence Berkeley National Laboratory outlook said they could consume 6.7% to 12% of total US electricity by 2028. Goldman had previously projected global data-center power demand would rise 50% by 2027 and as much as 165% by the end of the decade versus 2023, with AI making up 28% of the market in its baseline case. For Goldman staff, that means the AI story is increasingly about power contracts, capital formation and project execution, not just chip spending.
Know something we missed? Have a correction or additional information?
Submit a Tip

