Investment thesis · Live
Demand for firm power is becoming the bottleneck for the AI sector, and three turbine makers hold the keys to it
Data centers running artificial intelligence need gigawatts of firm capacity faster than the transmission grid can deliver it. The catch: barely a handful of companies in the world build large gas turbines, and their order books are bursting.
Published July 5, 2026 · 180-day horizon · supply and demand
Basket return
-6.4%
equal weight, since publication
The market over the same window
+2.6%
benchmark for this basket
Edge over the market
-8.9%
in percentage points
Causal chain: Build-out of compute infrastructure for AI → Bottlenecks in power grid interconnection → Gas turbines as a fast, firm source of power → Oligopoly of 3 manufacturers and sold-out capacity → Pricing power and decades of high-margin service revenue → Heavy gas turbine manufacturers
What this thesis rests on
Each one is a statement that has to be true. When a filing says otherwise, the thesis is in trouble, and this is where we say so.
-
○
No manufacturer outside GE Vernova, Siemens Energy and Mitsubishi Power books an order for a heavy-duty gas turbine above 300 MW (H/J/JAC class) during the 180-day horizon.
competitive · Unverified
-
○
GE Vernova reports a Power segment book-to-bill ratio above 1.0 in each quarterly report filed within the 180-day horizon.
financial · Unverified
-
○
GE Vernova's reported Power services backlog stays above 60 billion USD and remains at least twice its Power equipment backlog in every 10-Q filed in the horizon.
financial · Unverified
-
○
Siemens Energy reports a Gas Services profit margin before special items above 10% in both quarterly reports published within the horizon.
financial · Unverified
-
○
GE Vernova discloses heavy-duty gas turbine slot reservations filled through at least 2028 and announces no capacity expansion above 20 GW of annual heavy-duty output during the horizon.
operational · Unverified
Most of the capital in artificial intelligence flows toward chipmakers. Meanwhile, the hardest constraint, the one nobody can step around, sits a layer below: it is electricity, and more precisely immediate access to very firm power. A modern data center can draw as much power as a small town. These sites need uninterrupted supply located close to the load and switched on within a short window of two to four years. That is a physical limit on growth that no amount of source code will overcome. And that is exactly where the real supply problem in the market sits today.
Why gas turbines specifically, rather than green energy or grid expansion? Interconnection queues in the US and Europe run for years. Transmission system operators cannot build high-voltage lines at the pace the tech giants demand. Renewables, for their part, get bogged down in permitting and are intermittent by nature. Batteries can only shift energy in time, they do not create it out of nothing. Next-generation nuclear (SMRs included) is still a decade away from commercial scale. Heavy gas turbines remain the only technology that lets operators build large, fully dispatchable capacity right next to the server halls inside the timeframe they are working with. The leading cloud providers and the utilities tied to them understand this perfectly well, and are ordering gas units in bulk.
This thesis, over time
What has happened since publication. Every entry is computed as of its own date, from price history.
August 19, 2026
Day 45 of 180: basket -3.9%
The first quarter of the horizon is behind this thesis. The equal-weight basket stands at -3.9% since publication day. The benchmark (S&P 500) returned +3.1% over the same window, so the basket trails its market by 7.0 percentage points. The strongest contributor is 7011 (+8.0%), the weakest is GEV (-11.5%). 135 days remain until the verdict.
Open the interactive thesis
Interactive causal map, the basket charted against its market, and the full timeline.
Open in Invesaro →
This is analysis, not investment advice and not a recommendation to buy or sell anything. We publish it and track it in public, mistakes included. Any decision is yours and yours alone.