So heres the thing. The market keeps calling this an AI trade but the better pressure point may be land-to-power-to-energized capacity. A chip order doesnt matter much if the site isnt built, the electrical work isnt staffed, the turbine slot isnt there, the grid gear isnt available, or the interconnection queue is still sitting in mud.
Thats where STRL & GEV fit the board. STRL is the site-to-electrical execution leg, not just dirt. Sterling’s Q1 showed $825.7M revenue, $3.8B signed backlog, $5.15B combined backlog & E-Infrastructure at 72% of revenue. Mission-critical work was over 90% of E-Infrastructure backlog. Thats data centers, manufacturing, semis & all the physical stuff the market pretends just appears after the AI headline.
GEV is the power-to-interconnection leg. Q1 orders were $18.3B, backlog hit $163B & Electrification booked $2.4B of data-center equipment orders, more than all last year. Thats the cleaner proof that AI demand is hitting grid hardware & power equipment.
BE fits too but its the dangerous one. Onsite power bypass is real. Oracle has an initial 1.2 GW contracted under a master agreement that supports up to 2.8 GW. Bloom’s Q1 revenue was up 130.4%, with product revenue up 208.4%. But the stock is priced like the setup runs clean, scales clean & never trips over warranty, dilution, customer concentration, fuel, permits, or execution.
So my map is simple. STRL helps build the physical capacity. GEV helps power & connect it. BE helps bypass the grid when time-to-power is the problem.
The failure line isnt whether AI stays popular. The failure line is backlog conversion, margins, delivery timing & whether the market already paid too much for the bottleneck solvers. The room keeps chasing compute. The pressure may be in concrete, electricians, turbines, transformers, switchgear, gas, permits & interconnection.