The Big Three gas turbine OEMs are sold out through 2029-2030, with GE Vernova alone sitting on an 80 GW backlog. Reciprocating gas engines — Jenbacher, Wärtsilä, Caterpillar — offer 12-18 month deployment timelines vs. 5-7 years for large-frame turbines. The companies getting power online fastest are pairing recip engines with behind-the-meter structures and integrated gas supply.
Every hyperscaler, GPU cloud provider, and data center developer in the United States is trying to buy the same thing right now: gas-fired generation capacity. And every one of them is running into the same problem. The equipment is not available.
GE Vernova ended 2025 with an 80 GW gas turbine backlog stretching into 2029. CEO Scott Strazik has said the company expects to be sold out through 2030 by the end of 2026. Siemens Energy nearly doubled its turbine sales from 100 units in 2024 to 194 in 2025 — and still cannot keep pace with demand. Mitsubishi Heavy Industries is quoting delivery dates of 2028-2030 for new orders placed today.
The lead time for a new combined-cycle gas plant has jumped from three and a half years in 2023 to five years in 2025. For large-frame turbines, the wait can stretch to seven years. In a market where GPU pods deploy in months and data center campuses are breaking ground quarterly, seven years is not a delivery timeline. It is an obituary for the project.
The question facing every power-hungry operator is not whether gas-fired generation is the right answer. It is. Gas is the only dispatchable, scalable, fuel-secure generation source that can serve base-load AI compute at the densities required. The question is how to get it built before the opportunity window closes.
The Bottleneck Is Not Gas. It Is Iron.
The natural gas supply chain is not the constraint. The Appalachian Basin alone produces over 36 Bcf/d. The Permian Basin adds another 25 Bcf/d. Pipeline infrastructure, while not unlimited, has meaningful available capacity — especially for behind-the-meter configurations that connect directly to existing transmission laterals rather than requiring new long-haul pipeline builds.
The constraint is the equipment itself. The large-frame gas turbine market — the GE 7HA, Siemens SGT6-8000H, MHI M501JAC class machines that power utility-scale combined-cycle plants — is effectively sold out for the next four to five years. These are 300-400 MW machines with 18-24 month manufacturing lead times, and the order books are full.
This matters because the traditional path to gas-fired power — develop a site, order a turbine, build a combined-cycle plant, interconnect to the grid — now takes the better part of a decade from decision to first electron. That timeline was already strained. With the current backlog, it is broken.
Reciprocating Engines: The Faster Path
The answer that is emerging across the industry is not to wait for the turbine backlog to clear. It is to use a different machine entirely.
Reciprocating gas engines — large-bore, high-efficiency engines manufactured by INNIO (Jenbacher), Wärtsilä, and Caterpillar — operate on fundamentally different manufacturing and deployment timelines than large-frame turbines. A single reciprocating engine unit produces 5-10 MW. A multi-engine power plant configured from modular containerized units can deliver 25-100 MW or more. And critically, these units can be manufactured, delivered, and commissioned in 12-18 months — not five to seven years.
The scale of demand for recip engines is already making headlines. In October 2025, INNIO announced the largest order in company history: a 2.3 GW project with VoltaGrid comprising 92 modular power packs, each delivering 25 MW using Jenbacher engine technology. Deliveries begin in 2026 with 2 GW targeted online by 2027. That is more generation capacity than many utility-scale projects will deliver in the same period — and it will be operational years sooner.
Reciprocating engines also bring operational characteristics that are well-suited to data center loads:
Fast ramping. Recip engines reach full load in minutes, not the 30-60 minutes required for a combined-cycle turbine to ramp through its heat recovery steam generator. For data centers with variable AI training workloads, this responsiveness matters.
High part-load efficiency. A combined-cycle plant loses significant efficiency below 60-70% load. Reciprocating engines maintain high efficiency across a wide load range because individual engines in a multi-unit plant can be started and stopped independently. Run six of ten engines at full load rather than ten engines at 60% — the fuel savings are material.
Modular scalability. Adding capacity means adding engines, not replacing a turbine with a larger turbine. A site that starts with 25 MW can grow to 100 MW by adding identical units. The civil works, fuel supply, and electrical infrastructure are designed for expansion from day one.
Maintenance flexibility. In a multi-engine plant, one engine can be taken offline for scheduled maintenance while the remaining units continue to serve the load. A single large turbine requires the entire plant to go offline for major inspections. For a data center operator running continuous AI workloads, this redundancy is a hard requirement.
Why Vertical Integration Wins the Timeline
Having access to reciprocating engines is necessary but not sufficient. The companies getting power online fastest are the ones that control the full stack: gas supply, land, generation, and the customer relationship.
Here is why the timeline compresses when those pieces are integrated under a single operator:
Gas supply certainty. If you own mineral rights or have firm transportation contracts on existing pipelines, you do not need to negotiate a new gas supply agreement — a process that can add 6-12 months to a project timeline. The fuel is already committed.
Site control. If you control land with existing pipeline access and favorable permitting characteristics, site selection does not require a multi-month search and option negotiation. The site is already in the portfolio.
Permitting speed. Behind-the-meter generation on privately controlled land, connected to existing gas infrastructure, typically faces a lighter permitting burden than a utility-scale power plant requiring grid interconnection, transmission upgrades, and public utility commission approval. Months come off the timeline at every regulatory step.
Single counterparty. A data center operator negotiating with separate landowners, gas suppliers, generation developers, and EPC contractors is coordinating a multi-party development process. A single counterparty who controls land, gas, and generation collapses that coordination overhead into a single term sheet.
The difference between a fragmented development process and a vertically integrated one is not marginal. It is often 18-24 months of timeline compression — which, in the current market, is the difference between having power when the GPUs arrive and having power two years after the GPUs were supposed to arrive.
The Window Is Now
The gas engine backlog will eventually ease. GE Vernova is expanding production capacity. Siemens is investing in new manufacturing lines. INNIO and Wärtsilä are scaling to meet data center demand. But “eventually” is 2028-2029 at the earliest for large-frame turbines, and recip engine lead times are already extending as demand accelerates.
The operators who secure generation equipment, gas supply, and site positions now will have power to sell when the next wave of GPU deployments arrives. The operators who wait for the backlog to clear will find themselves in the same line, competing for the same equipment, with the same multi-year delivery timelines.
Gas-fired generation is not going away as the backbone of data center power. The fuel economics, dispatchability, and energy density are unmatched by any alternative at the scale AI compute requires. The variable is not whether gas wins — it is who gets gas-fired power built first.
The answer, in 2026, is the companies that are not waiting in the turbine queue.
Read next: GPU Pods and Distributed Power · Behind the Meter: The Data Center Power Race · Our BTM Power Program
Have generation capacity or need it built? Contact Ryan