Goldman Sachs is forecasting a 38% increase in US data-center power demand, a projection that lands as utilities and grid operators already face questions about whether they can keep up. The bank's estimate puts hard numbers behind a trend that's been building for years: the electricity appetite of the buildings that run cloud computing and, increasingly, artificial intelligence workloads.
The forecast doesn't come with a timeline attached, but the direction is clear. More data centers, more power. And that power has to come from somewhere.
Why the number matters
A 38% growth figure isn't abstract. It translates into real load on a grid that, in many parts of the country, is already stretched. Data centers are among the most power-hungry facilities in commercial real estate, drawing continuous electricity for servers, cooling systems, and backup infrastructure. When demand climbs that sharply, the effects ripple outward.
Utilities have to decide whether to build new generation, upgrade transmission lines, or both. Those decisions take years and billions of dollars. Some of that cost typically ends up in rate cases — the proceedings where regulators decide what customers pay.
The strain on energy resources
Goldman's forecast points to a potential squeeze on energy resources. It's not just about generating enough electrons. It's about getting them to the right places at the right times. Data centers tend to cluster in specific regions, and those regions don't always have surplus power sitting around.
That mismatch creates pressure. Grid operators may need to rely more heavily on peaking plants, which are expensive to run. In some cases, new natural gas capacity gets proposed. In others, developers look at nuclear or large-scale renewables, though those come with their own lead times.
The forecast doesn't specify which resources will fill the gap, but it makes the gap harder to ignore.
What it could mean for utility bills
One of the more direct consequences of rising data-center demand is its potential influence on utility costs. When a large new load shows up on a system, the utility has to serve it. The infrastructure upgrades required to do that get paid for through rates, and rate increases apply to everyone on the system, not just the data center.
That dynamic has started to draw attention from state regulators and consumer advocates. The question they're asking is straightforward: should residential customers shoulder a share of the cost for infrastructure built primarily to serve a handful of hyperscale facilities? There's no uniform answer yet, and the rules vary by state.
Goldman's forecast doesn't resolve that debate. It just makes it more urgent.
Local pushback is part of the picture
Data centers also face political resistance in some communities. Local opposition can slow or block projects, often over concerns about noise, water use, tax incentives, and visual impact. Power demand is another item on that list. Residents who see new transmission lines or substations proposed near their homes don't always welcome the change, even when the data center itself promises jobs and tax revenue.
That resistance matters because it can delay the very infrastructure needed to meet the demand Goldman is projecting. A forecast of 38% growth assumes the power can actually be delivered. If local approvals stall, the number becomes harder to hit.
What happens next
Goldman's projection is a forecast, not a mandate. It doesn't say when the 38% growth will materialize or which markets will see the sharpest increases. What it does do is add a credible estimate to a conversation that's been long on anecdotes and short on numbers.
Utilities, regulators, and community groups are already reacting to the broader trend, and the specifics will play out in rate cases, permitting decisions, and generation plans over the next several years. Whether the grid can absorb a 38% jump without major cost shifts or delays is the open question — and it's one that won't be answered by a single forecast.
