Who makes LDES for AI data centers

Long-duration storage is the category that would close the gap solar and wind leave. Lithium batteries are economic for a few hours; holding a flat load through a still night or a cloudy week needs something that discharges for much longer, and cheaply enough per unit of energy stored that sizing it for many hours is affordable. That is a different engineering problem from a battery, and the approaches on this part are correspondingly varied — flow batteries, iron-air chemistry, thermal storage in hot solid media, compressed and liquefied gas. What they share is the same trade in different forms: they give up round-trip efficiency and power density in exchange for cheap energy capacity, on the reasoning that if storing energy is cheap enough, losing some of it matters less than the capital saved. Whether that trade wins depends on how often the system cycles, which depends on the shape of the grid it sits on. For AI operators this is a technology to watch rather than one to buy: it would make an hourly-matched renewable supply achievable without firm generation, which is why it sits alongside geothermal and nuclear in this layer. What to look for is demonstrated projects at grid scale rather than pilots, cost per unit of stored energy rather than per unit of power, and whether a company has a manufacturing base or a first plant under construction. Most of these are pre-commercial, and the catalog's confidence tags should be read that way.