The AI Power Boom Is Being Built on Gas, Not the Clean Baseload the Market Is Buying

The AI Power Boom Is Being Built on Gas, Not the Clean Baseload the Market Is Buying Capital is rotating into storage and nuclear as the answer to data-center demand, but 72% of MISO's expedited interconnection queue is gas-fired — the buildout the equity market misreads. The market has decided the AI power crunch gets solved with clean baseload. Capital is rotating into the companies positioned to supply data centers, with nuclear and renewable generation pitched as the cleanest fix for the constraint, the narrative that forced Fluence Energy's stock up 98% in a single week on hyperscaler deals and a record backlog.¹ That is the story equity desks are paying for. The interconnection queue tells a different one, and the gap between the two is the signal nobody is pricing. Look at what is actually lined up to connect. Entergy's power projects, all of them gas-fired, make up nearly a third of the roughly 28 GW in MISO's fast-track interconnection queue, according to a Utility Dive analysis of the updated list.³ Across the whole of MISO's expedited resource addition process, gas-fired network interconnection capacity totals 20.3 GW, about 72% of the total, while battery, solar and wind together account for 4.3, 2.3 and 1.2 GW respectively.³ The buildout that the market is pricing as a clean-energy event is, in the queue that determines what gets built, almost three-quarters gas. The driver makes the mismatch sharper, because this is data-center load specifically. About 70% of Entergy's proposed capacity additions, spread across Louisiana, Mississippi and Texas, are designed to serve planned data-center complexes, and the company expects retail sales to grow 8.5% a year through 2030 partly on that industrial load.³ So the exact demand the Fluence rally is built on is being met, by the utility closest to it, with gas turbines rather than batteries. If the market is wrong that clean baseload wins this buildout, the re-rating of storage and renewable names is running ahead of the physical reality of what grid operators are approving. The capital allocation confirms where the conviction sits. Entergy expects to spend about $27 billion on new generation through 2029, against roughly $7 billion on renewables and storage.³ That is close to a four-to-one tilt toward conventional generation from the company building hardest for data-center demand. A market pricing the AI buildout as a clean-energy capex wave is looking at the wrong line of the budget. The clean alternatives, meanwhile, keep hitting friction the gas projects do not. The European Commission has opened an investigation into France's plan to subsidise six new reactors totalling 10 GW, a project estimated at around €73 billion.² Even state-backed nuclear, the firmest clean baseload there is, faces a regulatory probe before a shovel moves, while gas sits 20.3 GW deep in an expedited queue. The path of least resistance for near-term load is the carbon-intensive one, and that is precisely what the interconnection data shows utilities choosing. This is why the bearish lean in power markets may be missing the gas demand underneath. A market focused on the equity story of clean baseload is underpricing the structural pull on natural gas that 20.3 GW of new gas-fired capacity represents, much of it baseload-style data-center demand that runs around the clock.³ If even half of that queue energises on schedule, it is a durable new call on gas that the consensus, watching battery backlogs and reactor announcements, is not modelling. The signal to watch is MISO's next interconnection cycle, where the operator is reviewing about 3.7 GW more of gas-fired, solar and storage capacity, because the fuel mix of each new tranche tells you whether the gas tilt is accelerating or moderating.³ The equity market is trading the clean-baseload headline. The grid operators are approving gas. When the queue and the narrative diverge this far, the queue is the one pouring concrete, and the trade is in the fuel it actually burns.^3,1