Behind-the-meter generation is not a way around the gas turbine shortage. It is the mechanism routing hyperscaler capital into it.
On-site generation can shorten grid interconnection delays, but it still requires turbines, castings and forgings, generator step-up transformers, switchgear, and gas delivery infrastructure. The same limited group of manufacturers supplies the equipment whether the buyer is a utility or a hyperscaler.
The Backlog Is Already Here
The three dominant heavy-frame turbine manufacturers report order books extending into the next decade.
| OEM | Gas Turbine Backlog | Trajectory | Delivery Horizon |
|---|---|---|---|
| GE Vernova | 116 GW at end of Q2 2026 | Up from 100 GW in Q1 and 83 GW at end of 2025 | Sold out through 2030; taking 2031 reservations |
| Siemens Energy | 69 GW firm in fiscal Q3 2026 | Added 15 GW in one quarter | Booked through 2028; lead times exceed three years |
| Mitsubishi Power | 35 GW of large-frame capacity | Up from 23 GW one year earlier | Orders booked for 2028 to 2030 delivery |
GE Vernova attributes roughly 20% of its backlog to data-center load. Siemens Energy reported that data centers generated 65% of its U.S. orders in one recent quarter. The three OEMs disclosed at least 157.6 GW of firm orders or backlog worldwide by mid-2026.
Large-turbine prices are projected to reach $600 per kilowatt by the end of 2027, nearly three times their 2019 level. Average waits have risen to about five years, with some orders stretching to seven. Doosan Enerbility and Ansaldo are re-entering the U.S. market to absorb demand the dominant manufacturers cannot serve.
Behind-the-Meter Routes Demand Into the Constraint
More than 130 GW of energy resources have been proposed for planned U.S. data centers. Developers have announced roughly 101 GW of on-site natural-gas generation as they confront grid interconnection timelines of 36 to 84 months and data-center construction cycles of 12 to 24 months.
| Project | Capacity | Status |
|---|---|---|
| xAI Colossus 1 and 2, Memphis | Approximately 1,498 MW | Operating through dozens of temporary mobile turbines |
| Meta Hyperion, Louisiana | 7.46 GW | Expanded after Entergy agreed to build seven additional plants |
| Microsoft, West Virginia | 1.4 GW | Letter of intent signed |
| Microsoft, West Texas | 2.5 GW proposed | Exclusive talks with Chevron and Engine No. 1 |
| Meta Socrates North, Ohio | 200 MW | Approved |
These projects draw from the same heavy-frame and aeroderivative supply chains already booked into 2030 and 2031. Hyperscalers can also secure scarce production slots by paying deposits that utilities may be unable or unwilling to match. GE Vernova now requires deposits of 20% to 25% to hold a reservation.
Behind-the-meter generation bypasses one queue: grid interconnection. It does not bypass the equipment or fuel infrastructure needed to produce and deliver electricity.
The Ranked Constraint Structure
The supply chain contains several bottlenecks with different expansion timelines.
| Rank | Constraint | Lead Time or Difficulty | Structural Cause |
|---|---|---|---|
| 1 | Castings and forgings | Multi-year and oligopolistic | Few qualified plants; single-crystal casting and heat treatment are difficult to scale |
| 2 | Heavy-frame turbines | Three to seven years | Assembly is constrained, but upstream components impose the harder ceiling |
| 3 | High-voltage transformers and switchgear | Three to five years | Production is limited by grain-oriented electrical steel |
| 4 | Pipeline and gas delivery | Region-specific | Sites require pipeline capacity, firm transport, and manageable basis exposure |
| 5 | Reciprocating engines and aeroderivatives | Shorter production cycles | Faster to deploy but smaller and generally less efficient |
The tightest constraint is not turbine assembly. It is the specialized casting and forging capacity required to produce blades and rotors.
The Bottleneck Beneath the Turbine OEMs
Siemens Energy identifies large castings and forgings as its tightest supply constraint. EPRI researchers point to rotor forgings and hot-section blades as the primary bottlenecks. Some large-frame turbines are being shipped without rotors or blades, with final assembly deferred until those components arrive on site.
Major suppliers include Howmet Aerospace, Doncasters, Berkshire Hathaway-owned PCC Airfoils, Chromalloy, and Consolidated Precision Products.
Modern turbine blades operate above the melting temperature of their constituent metals. They survive by using nickel-based superalloys grown as single crystals without grain boundaries. Their hollow cooling geometries require investment casting, directional solidification, and single-crystal vacuum furnaces.
This process has high scrap and rejection rates. A new supplier requires specialized furnaces, regulatory approvals, substantial capital, and years of qualification by engine manufacturers. Expanding an assembly plant does not increase output if qualified blades and rotors remain unavailable.
Constraints That On-Site Generation Cannot Avoid
| Constraint | Why It Remains Binding |
|---|---|
| Transformers | Islanded plants still need generator step-up transformers. Demand for these units rose 274% from 2019 to 2025, while high-capacity lead times reached four years. |
| Pipeline and gas delivery | Data-center gas demand may reach 6.1 Bcf per day by 2030. Sites selected for land and fiber access may lack sufficient pipeline capacity or firm transport. |
| Electrical steel | Cleveland-Cliffs’ Butler Works is the only domestic source of grain-oriented electrical steel. Its $195 million expansion is not scheduled to operate until July 2028. |
A roughly 20% increase in national gas demand in January 2026 drove Henry Hub prices up 400%. The problem was insufficient pipeline capacity, not a shortage of gas in the ground. Behind-the-meter generation relocates the infrastructure constraint rather than eliminating it.
Reciprocating Engines Provide Partial Relief
Large reciprocating engines can be installed faster than heavy-frame turbines and are increasingly used for primary generation rather than backup power.
Caterpillar’s backlog reached a record $72.1 billion in Q2 2026, up 92% year over year. Sales of large generator sets and turbines increased 72%. The company plans to expand large-engine capacity to nearly three times its 2024 level, with most related capital spending scheduled for 2027 through 2029.
Reciprocating engines extend available capacity but do not replace heavy-frame combined-cycle plants at gigawatt scale. They operate in smaller units and generally at lower efficiency.
The Material Risk Is Backlog Conversion
The equipment shortage is already constraining development. The investment risk is whether reserved capacity converts into completed, revenue-generating projects on the expected schedule.
Three factors can interrupt that conversion:
-
Reservations exceed committed demand. Large deposits make turbine orders more credible, but developers may reserve equipment for multiple sites before financing and final approvals are secured.
-
Announcements exceed construction. Of roughly 12 GW of U.S. AI data-center capacity announced for 2026 in one sample, only about 5 GW was under active construction. Equipment shortages and permitting delays held back the remainder.
-
Labor limits installation. An August 2026 estimate indicated that U.S. buildout plans would require roughly 500,000 additional electricians, 300,000 welders, and 550,000 plumbers. Equipment cannot generate revenue until qualified workers install it.
If hyperscaler capital spending slows, behind-the-meter projects are more exposed than utility plants serving diversified, rate-based demand. Turbine and transformer valuations could weaken before the physical supply constraint clears.
Bottom Line
Behind-the-meter generation accelerates demand for scarce power equipment. It shortens the grid interconnection queue while colliding with a ranked chain of other constraints:
Castings and forgings → heavy-frame turbines → transformers and switchgear → pipeline capacity → reciprocating engines and aeroderivatives
The hardest layer to expand is single-crystal castings and large forgings, not turbine final assembly. The commercially important suppliers are those controlling the components and infrastructure required to deliver power on hyperscaler construction timelines.
The physical bottleneck is established. The unresolved variable is how much of the backlog becomes completed, operating capacity.