Thesis. The hardest bottleneck in AI power — taking gas turbine slot reservations for 2031 delivery, with first-half equipment orders priced over 20% above Q4 2025 and a Power segment book-to-bill of 3.05x. Scarcity showing up in price, not narrative.
Reviewed 19 Sep 2026, after Q2 2026 results. Next review after Q3 2026 earnings on 21 Oct 2026.
What it does
GE Vernova makes heavy-duty gas turbines, grid equipment and wind turbines. A gas turbine is effectively a jet engine bolted to a generator — the workhorse of new dispatchable power, meaning power you can turn on when you need it rather than when the weather allows. Spun out of General Electric in 2024.
Two order concepts matter here. Backlog is firm orders. Slot reservation agreements are customers paying to hold a manufacturing slot years ahead — not yet firm, but a strong signal, and the reason this company can talk about 2031 deliveries at all.
The clearest evidence in this entire file that the AI power shortage is real is not a backlog number. It is that first-half equipment orders were priced more than 20% above where they were nine months earlier, and customers paid it.
Bull case
- Gas turbine backlog plus slot reservations went from 100 GW to 116 GW in a single quarter, targeting at least 125 GW under contract by year-end 2026, with backlog extending through 2031
- First-half equipment orders priced over 20% above Q4 2025 levels — the cleanest pricing-power evidence anywhere in this research
- Q2 2026 orders $24.2B, +88% organic, book-to-bill above 2x. Power segment orders $16.7B, +134% organic, book-to-bill 3.05x
- Revenue $11.1B (+22%); adjusted EBITDA $1.2B (+61%) with margin up 340bps
- Free cash flow $5.1B in the quarter, ~$10B year to date — more than 2.5x all of 2025
- Record backlog $176B, +$13B sequentially, expected to reach ~$200B in FY27. Equipment backlog +77% to $88B
- Signed 52 heavy-duty units (15 HA-class) and 61 aeroderivative units
- Datacentre-driven demand exceeded $5B year to date, with ~20% of new gas capacity destined for datacentres. Supplying xAI's Colossus 1 and Microsoft datacentres in Texas
- Solid-state transformers and medium-voltage UPS approaching commercialisation, with potential revenue per GW rising 2–3x; SST prototypes shipping to a hyperscaler in late 2026
- FY26 revenue guidance raised to $45.5–46.5B
Bear case
- Expanding capacity destroys the scarcity it currently monetises. Manufacturing is scaling to 20 GW annualised in Q3 2026, 24 GW by 2028 and toward 30 GW by 2030 — which is precisely what ends 20% pricing power
- SpaceX is building a turbine-blade casting foundry in Bastrop, Texas, targeting up to 18 months faster turbine availability. Genuinely ambiguous: more blade supply could unlock GEV's own capped output, but a new captive source changes allocation and pricing
- Wind orders fell sharply with a deeper segment loss, on US onshore softness, tariffs and permitting
- Beat revenue by $330M but missed EPS by $0.71
- Slot reservations are not firm orders until converted — 116 GW is a softer number than it reads
- Megacap size dampens the move relative to smaller power names
Major customers
- Hyperscalers and datacentre developers — xAI (Colossus 1) and Microsoft (Texas) named. ~20% of new gas capacity is destined for datacentres
- Electric utilities and independent power producers — the bulk of turbine demand, including VST, CEG and NRG
- Grid operators and transmission owners — the electrification segment
- Wind developers — the currently weak segment
Read-through. GE Vernova is the deepest bottleneck in AI power, and its 20% price increases are the strongest single margin-test pass in this project. Its turbine availability gates how fast VST, CEG and the whole datacentre buildout can add dispatchable capacity — every power-producer announcement in this file ultimately queues here. It sells into the same utilities that PWR and MTZ build for, and it sits opposite BE and CAT in the same customer decision — which is why GEV's capacity expansion is a bear point for Bloom as much as for itself. Watch the SpaceX foundry story: the photo in the coverage showed Musk with a GE Vernova turbine part, which suggests SpaceX may be casting blades that feed GEV turbines rather than replacing them. That would turn the single biggest threat on this page into an accelerant.
What would change the view
- The year-end target of 125 GW under contract — the single metric for this name
- Equipment pricing, currently +20% YoY. Deceleration is the thesis warning, and it will show up before volumes do
- Slot reservations converting to firm orders
- Wind segment losses, and whether they stabilise
- SST prototypes shipping to a hyperscaler in late 2026
- SpaceX Bastrop foundry progress — and specifically whether it feeds or competes
Update log
19 Sep 2026 — Converted to the standard template. Current through Q2; no new disclosures since. Made the backlog-versus-slot-reservation distinction explicit, since the 116 GW figure mixes the two.
Q2 2026 — Orders $24.2B (+88% organic), book-to-bill above 2x; Power segment orders $16.7B (+134% organic), book-to-bill 3.05x. Revenue $11.1B (+22%), adjusted EBITDA $1.2B (+61%) with margin +340bps. Free cash flow $5.1B in the quarter. Record backlog $176B (+$13B QoQ); equipment backlog +77% to $88B. Gas turbine backlog plus slot reservations 100 GW → 116 GW. First-half equipment orders priced >20% above Q4 2025. FY26 revenue guidance raised to $45.5–46.5B. Revenue beat by $330M; EPS missed by $0.71.
Research and education only — not investment advice.
Made with Bullet