GE Vernova's gas equipment orders quadrupled last quarter as AI power demand outruns supply
GE Vernova reported second-quarter gas equipment orders up fourfold year over year and a backlog of $176 billion, the clearest single read yet on how fast AI data center demand is converting into orders for generation capacity. Anthropic's Build AI in America policy paper puts the requirement at roughly 50 gigawatts of new US capacity by 2028. Brent crude sitting in the low-to-mid $90s is adding a second, unrelated bid under energy equities.
GEV reported second-quarter gas equipment orders up fourfold from a year earlier and a total backlog of $176 billion, the most concrete evidence available that AI data center demand is now converting into firm orders for generation capacity rather than announcements. Gas turbines are the swing supply for this build-out because they can be sited and permitted faster than nuclear and dispatched more reliably than renewables, which is precisely what a data center operator with a fixed opening date is buying.
The scale of the requirement is contested, but one published figure anchors it. Anthropic's Build AI in America policy paper estimates roughly 50 gigawatts of new US power capacity will be needed by 2028 to support planned AI infrastructure. That figure comes from an AI developer making a policy argument rather than from an independent forecaster, so treat it as an interested party's estimate: useful for order of magnitude, not as a neutral projection.
Turbine order books are effectively full through the end of the decade, which turns a demand story into a queue-position story. When capacity is allocated years ahead, the value shifts to whoever already holds slots, and GE Vernova's backlog is the measure of that position. It also means incremental demand shows up as pricing and longer delivery dates rather than as volume in the near term.
A second, separate force is acting on energy equities at the same time. Brent crude traded in the low-to-mid $90s on September 1 and 2 on US-Iran tensions, which lifts the whole sector regardless of any AI connection. Conflating the two is the easy mistake here: an oil-driven rally in energy names tells you nothing about data center power demand, and the two can reverse independently.
What to watch is order conversion rather than order intake. A backlog is a promise to deliver, and the constraint on delivery is turbine manufacturing capacity, grid interconnection and skilled labor, none of which scale on the same timeline as the orders. The gap between the $176 billion backlog and revenue recognition is where the next disappointment or the next re-rating comes from.
Related Stocks
Powered by SentiSense - Intelligent Market Analysis