The Gigawatt Era: How AI’s Insatiable Appetite for Power Is Reshaping the Turbomachinery Industry

There is a number worth sitting with before anything else: approximately 25% of Siemens Energy's current gas turbine order book is tied directly to data center projects. A year ago, that number was a fraction of what it is today.

That single data point—shared exclusively with Turbomachinery International by Richard Reisig, Managing Director of Global Investments & Asset Management at Siemens Energy—is as good a measure as any of how completely and how quickly artificial intelligence has redrawn the landscape of the turbomachinery industry. The machines at the center of this story are the same ones that have powered civilization for decades: heavy-duty gas turbines, aeroderivatives, combined-cycle plants, and the entire rotating equipment ecosystem that supports them. What has changed is who is buying them, why, and with what urgency.

Baker Hughes Chairman and CEO Lorenzo Simonelli put the stakes plainly at the company's 2026 Annual Meeting in Florence earlier this year. Global power demand is on a trajectory to double by 2040—driven by electrification, EV adoption, and above all, AI data centers.¹ Simonelli called this dynamic "The Energy Equation": the interdependence between the industrial outcomes society demands and the energy sources required to meet them.²

The scale of investment underway is staggering. Total investment in AI data center infrastructure is estimated to reach approximately $7 trillion through 2030.³ According to the International Energy Agency, global data centers consumed around 415 TWh of electricity in 2024, with consumption expected to more than double to 945 TWh by 2030—an incremental load comparable to Germany's entire national power demand.³ In the U.S., data centers are projected to be the single largest driver of new electricity demand, contributing 65 to 90 GW of incremental peak load, potentially up to 12% of total U.S. peak demand within the next several years.³

For the professionals who build, operate, and service the machines that generate that power, this is the moment that defines the next decade of their careers.

Why Are Gas Turbines the Technology of Choice for AI Data Center Power?

The answer is both technical and practical. Modern hyperscale data centers require massive amounts of dispatchable, always-on baseload power. Unlike intermittent renewable generation, gas turbines can be called up on demand, operate at consistent output regardless of weather conditions, and provide the grid inertia that large compute campuses require to maintain stable voltage and frequency.

At the heavy-duty end of the market, H-class and J-class machines deliver power outputs exceeding 600 MW per unit, with thermal efficiencies above 65% in combined-cycle configurations — the right profile for the largest hyperscaler campuses.³ But the data center market is also pulling demand toward industrial gas turbines and aeroderivatives, which offer smaller footprints and faster deployment timelines, even at some cost to thermal efficiency.

"Energy infrastructure becomes part of the data center's core design — often requiring new partnership models and financing approaches." -Richard Reisig, Managing Director, Global Investments & Asset Management, Siemens Energy

Natural gas-fired generation is expected to capture roughly 50% of new data center power supply through 2030, particularly for facilities requiring high reliability or located in regions where renewable-plus-storage configurations cannot meet baseload requirements.³ Baker Hughes estimates gas-fired power solutions will grow more than 80% between 2025 and 2030, with behind-the-meter and off-grid solutions representing a $25 billion-plus total addressable market by decade's end.¹

"Data centers have introduced a specific customer dynamic—one that prioritizes speed, flexibility, and integrated solutions,” Reisig told Turbomachinery International. “Unlike traditional utilities, hyperscalers often require tailored, modular systems that can be deployed quickly and scaled over time, with a strong emphasis on uptime and resilience. This is driving a shift toward more end-to-end offerings, spanning generation, electrical infrastructure, and services, often delivered in tighter collaboration with partners and customers."

GE Vernova frames the technology choice as a matter of matching machine type to mission. "For data centers facing grid constraints, aeroderivative units provide the reliability, fast-start and ramping flexibility needed for bridging power, while heavy-duty turbines in combined cycle application deliver the high efficiency required for continuous baseload operation," said Ihab Chaaban, P.Eng., MBA, Data Centers & Hydrogen Commercialization Director at GE Vernova. "By combining these technologies, operators can effectively manage the transition from peaking support to reliable base-load generation within behind-the-meter configurations." Because the HA and AERO fleets have already proven their durability in traditional utility environments, Chaaban added, GE Vernova can offer hyperscalers a level of risk mitigation that newer or less mature energy solutions cannot always provide — a shift he describes as moving not just in hardware, but toward an AI-enabled service paradigm that keeps those machines online to meet the high-utilization demands of AI infrastructure.

What Is Driving the Behind-the-Meter Revolution?

One of the most consequential structural shifts reshaping the turbomachinery market is the rapid move toward behind-the-meter power generation, where data centers build and own their own dedicated generation assets rather than drawing from the grid.

The reason is straightforward: the grid cannot support the AI power surge quickly enough. Utility interconnection queues in the U.S. are backlogged by years, and the grid infrastructure—much of which was built in the 1950s through 1970s—was not designed for the magnitude of concentrated load that a modern hyperscale campus represents. For developers who need power in 2026 or 2027, waiting for a utility interconnection is simply not an option.

"Behind-the-meter models are reshaping both technical requirements and commercial structures," Reisig says. "Technically, they demand highly flexible, modular systems capable of operating independently or in hybrid configurations, often with fast-start capability and integration with storage or other generation. Commercially, they shift projects toward integrated, vertically coordinated solutions where energy infrastructure becomes part of the data center's core design—often requiring new partnership models and financing approaches."

The result has been a proliferation of projects where gas turbines, generators, and balance-of-plant equipment are being deployed directly on data center campuses. Industry estimates suggest that data centers could rely on 35 GW of behind-the-meter power by 2030, with natural gas generation as the dominant technology.

Baker Hughes has moved aggressively into this space. At the Bernstein 42nd Annual Strategic Decisions Conference in May 2026, Simonelli identified a "sweet spot" for off-grid immediate power in the 150 MW to 300 MW range—spanning the NovaLT16, Frame 5, and BRUSH generator portfolio.⁴ The company booked $1 billion in data center-related power systems orders in Q1 2026 alone, and Simonelli indicated it would likely raise its $3 billion data center order target for 2025 to 2027.⁴ "The demand for data center power is not a one-year event," he said, "but a multi-year opportunity as grid infrastructure and alternative power solutions take time to develop."⁴

Natural gas reciprocating engines—and steam turbines in combined heat and power configurations—are also viable at the smaller end of the campus power scale, offering faster permitting timelines and modularity advantages compared to large frame turbines.⁵

What Has the AI Boom Done to Gas Turbine Supply Chains and Construction Costs?

This is where the story turns from opportunity to constraint—and where the implications for the industry are most acute.

The three OEMs that collectively supply more than 75% of large gas turbines globally—GE Vernova, Siemens Energy, and Mitsubishi Power—are each reporting order backlogs that push new turbine availability to late 2028 at the earliest, with many delivery slots already committed into 2030 and beyond.⁶ Siemens Energy reported its own all-time high in quarterly orders for Q2 FY2026, reaching a record order backlog of €154 billion on the strength of demand at its Gas Services business—prompting the company to raise its full-year 2026 outlook for revenue growth, profit margin, and free cash flow.⁷ Planned gas capacity has grown nearly sixfold over the period of the demand surge, according to BloombergNEF analysis of U.S. utility filings, while manufacturing capacity has not kept pace.⁸

"Scaling remains constrained by skilled labor availability, material supply, and the complexity of highly engineered components." - Richard Reisig, Managing Director, Global Investments & Asset Management, Siemens Energy

The cost consequences have been severe. According to BloombergNEF's "US Gas Power Plants: Higher Costs, Longer Waits" report, the average project expense for a combined-cycle gas turbine plant rose to $2,157 per kilowatt in 2025, up from less than $1,500/kW in 2023—a 66% increase in two years.⁸ The time required to bring a new gas-fired plant online has also stretched by 23% over the same period, reflecting the depth of supply chain bottlenecks across turbines, heat recovery steam generators, transformers, and skilled labor.⁸

All three major OEMs are moving to expand manufacturing capacity, but the expansions are measured and the ramp-up timelines are long. "We have always taken a disciplined approach to capacity expansion—phasing investments, focusing on productivity gains, and leveraging flexible manufacturing strategies and existing production facilities," Reisig says. "We're taking a cautious but optimistic approach."

GE Vernova's own $300 million expansion in South Carolina and New York illustrates what that discipline looks like in practice. "GE Vernova is investing $300 million to scale turbine manufacturing in South Carolina and New York by improving factory floors through intensive Lean 'Kaizen' events and bringing critical component production in-house to bypass supply chain bottlenecks," Chaaban said. The company has already reached a production capability of roughly 280 new machines, with the investments aimed at an annualized output of 20 GW by the third quarter of 2026 and further scaling toward 24 GW by 2028. Even so, Chaaban was candid about the limits of that expansion: "While this expansion significantly increases GE Vernova's throughput and revenue conversion, it will not immediately shorten customer lead times; due to the scale of our current backlog and the continued influx of orders through 2031, this increased capacity serves to fulfill existing demand rather than relieve wait times."

The caution is intentional. With demand cycles of this magnitude, the risk of overcapacity is real if AI infrastructure investment moderates before new manufacturing lines come fully online. Asked about the strength of the demand outlook, Reisig is direct: "We see strong gas turbine demand, not just from data centers but across markets and applications, well into 2035."

"Utilities and grid customers are both essential stakeholders in solving the global power challenge, and our focus is not on choosing between them but on helping improve the critical 'time to power.'" - Ihab Chaaban, P.Eng, MBA, Data Centers & Hydrogen Commercialization Director, GE Vernova

GE Vernova is pressure-testing its own strategy against the same risk. "Unlike the dot-com era, it is expected to see a transition from generative to inference AI, a shift that may introduce an entropy in demand as the technology matures," Chaaban said. "We are closely monitoring this evolution and intend to pace our operations to align with these emerging requirements." He pointed to three levers: a modular, Lean-driven manufacturing footprint that can adjust output without incurring the stranded-cost structures of past cycles; a "string of pearls" strategy bundling turbines with grid software and electrification solutions so that, as he put it, "an HA turbine, after all, is a fundamental infrastructure asset that serves both hyperscalers and traditional utility load growth" even if data center demand specifically fluctuates; and a reservation-agreement backlog that lets the company "pace our capital deployment in lockstep with confirmed, long-term project commitments rather than speculative peaks in demand."

Baker Hughes has positioned itself in the mid-market gap the big OEM backlogs have opened—the 150 MW to 300 MW behind-the-meter range where developers need turbines on timelines the major order books cannot accommodate. NovaLT capacity is effectively sold out through 2028, and on the company's Q1 2026 earnings call, management said the installed base for NovaLT units "is also set to expand dramatically in the coming years" and would benefit the aftermarket services business "into 2030 and beyond."⁹

For developers locked out of primary-market slots entirely, the secondary market for unfired OEM turbines has emerged as a viable path. Hallador Energy's June 2026 acquisition of 460 MW of unfired Siemens turbines from Energy World Corporation—routed through Siemens USA for factory refurbishment before commissioning—is an early indicator of a procurement model that could scale.

How Are OEM Backlogs Affecting Project Prioritization?

Heavy-duty turbines, such as GE Vernova's HA series, Siemens Energy's SGT-8000H, and Mitsubishi's JAC platform, dominate at the gigawatt-scale campus end, where thermal efficiency and long-run operating economics justify the lead time and capital cost. GE Vernova reported a 77% increase in its Q4 2025 power order book; Siemens Energy recorded its highest-ever quarterly order intake in Q1 FY2026; and Mitsubishi Heavy Industries saw Energy Systems bookings increase 45% year-over-year.³ At the mid-market end, Solar Turbines (Caterpillar) reported Q4 FY2025 sales growth of 23% and has announced plans to expand capacity 2.5x by 2030, while new entrants including FTAI Aviation and Boom Supersonic are converting retired jet engines and novel turbine platforms into data center power units.³ Baker Hughes has partnered with Boom to supply 1.21 GW of BRUSH generator capacity for Boom's Superpower turbine.¹⁰ Baker Hughes' own NovaLT16 spans the industrial mid-market and is 100% hydrogen-ready, capable of burning natural gas, hydrogen blends, syngas, ammonia, and green diesel.¹

For the turbomachinery industry's major OEMs, the demand surge has introduced an unfamiliar allocation problem: how to prioritize scarce capacity across legacy utility customers and new hyperscaler buyers simultaneously.

"While traditional utilities remain the bedrock of our business, the data center 'customer type' is growing rapidly in both volume and scope," Chaaban said. "We are successfully moving up the value chain by transitioning from a supplier of isolated hardware—gas turbines or transformers—to a strategic collaborator providing integrated electrical and software ecosystems for data center energy management." On how GE Vernova balances long-standing utility relationships against urgent hyperscaler timelines, Chaaban described a deliberately non-zero-sum approach: "Utilities and grid customers are both essential stakeholders in solving the global power challenge, and our focus is not on choosing between them but on helping improve the critical 'time to power.' We employ a balanced, integrated approach that focuses on long-term project stability rather than short-term prioritization." Because large-scale grid power remains the preferred long-term solution for data centers but often faces lengthy interconnection and permitting timelines, he added, GE Vernova increasingly uses fast-deploying aeroderivative and simple-cycle turbines as a strategic bridge while longer-term grid infrastructure is developed.

"With a record order backlog, capacity allocation becomes a question of strategic fit, execution certainty, and long-term partnership value," Reisig explains. "Siemens Energy is engaging customers early and transparently on timelines, while prioritizing projects that align with readiness, financing, and system impact. In parallel, the company is expanding capacity—but cautiously—recognizing the need to balance near-term opportunity with long-term stability."

The manufacturing challenge is equally real. "Operationally, that level of growth reflects a ramp-up across manufacturing, workforce, and supply chain," Reisig says of Siemens Energy's near-doubling in gas turbines sold in 2025. "Siemens Energy is expanding critical component production—such as blades and vanes—and thoughtfully investing in advanced manufacturing at its existing sites to increase output. Scaling remains constrained by skilled labor availability, material supply, and the complexity of highly engineered components."

What Is the Decarbonization Tension and How Are OEMs Navigating It?

No treatment of this moment would be complete without addressing the uncomfortable reality sitting at the center of it: the hyperscalers driving this unprecedented wave of gas turbine demand have also made some of the most aggressive net-zero commitments of any sector in the global economy. Microsoft, Google, Amazon, and Meta have all pledged to be carbon-negative or carbon-neutral by 2030 or shortly thereafter. They are simultaneously the buyers funding the most significant expansion of gas-fired generation infrastructure the U.S. has seen in a generation.

The resolution, to the extent one exists, runs through two parallel pathways. The first is efficiency: the gas turbines being deployed today are dramatically cleaner per megawatt-hour than the coal capacity they are displacing or the simple-cycle peakers they are replacing. H-class and J-class combined-cycle plants operating above 60% thermal efficiency represent the current state of the art in fossil fuel power generation.

The second is fuel flexibility. Siemens Energy, GE Vernova, Mitsubishi Power, and Baker Hughes are all designing turbines today for hydrogen co-firing and, ultimately, hydrogen-only operation, creating an upgrade pathway from natural gas to zero-carbon fuels as hydrogen supply and economics mature.

A parallel pathway is carbon capture. GE Vernova is supplying a 9HA.02 turbine for the NZT Power project in the UK, expected to be the world's first commercial-scale gas-fired plant with carbon capture. "Decarbonization is a foundational element of our procurement discussions with hyperscalers," said Matthew DavidSaver, Carbon Capture Product Champion at GE Vernova. "While our work on the NZT Power project in the UK—utilizing our 9HA.02 turbine—puts us at the forefront of commercial-scale carbon capture, we find that AI customers are currently highly pragmatic." Because CO2 pipeline and sequestration infrastructure is often unavailable on the timelines hyperscalers require, DavidSaver said, most are instead prioritizing "Carbon Capture Capable" provisions—allowing them to deploy assets now while embedding the flexibility for lower-risk, cost-effective retrofitting later. "This approach is aimed to ensure that our customers can meet their immediate compute demands today without risking the future viability of their energy assets as the regulatory and technical landscape evolves," he said.

"Siemens Energy is designing turbines today for long-term decarbonization pathways, including hydrogen-ready capabilities and integration with hybrid energy systems," Reisig says. "This allows operators to meet immediate power needs while maintaining flexibility to transition to lower-carbon fuels over time. In practice, however, many data center customers today are prioritizing reliability, speed to power, and uptime above all else, making decarbonization a critical, but often secondary, decision factor in the near term."

At Baker Hughes' Annual Meeting in Florence, Simonelli called for a holistic approach to the energy transition.² He argued that the classic energy trilemma of sustainable, affordable, and secure energy must be resolved alongside the demands of industrial growth, requiring pragmatic deployment of every available technology pathway. Baker Hughes has partnered with Google Cloud to develop AI-enabled power optimization solutions for data centers, exploring how turbomachinery performance data can be used to improve efficiency across the assets powering AI infrastructure.¹¹

The rapid buildout of lower-efficiency generation technologies to meet data center demand is expected to result in a substantial increase in U.S. power sector carbon emissions in the near term. It’s been advised that companies should therefore consider future upgrade pathways—including engines capable of running on alternative low-emission fuels—particularly as geopolitical conditions and environmental policies continue to evolve.³

What Does the Frontier of Gas Turbine Technology Look Like and What Is the Efficiency Ceiling?

Even as the industry races to build more of what it already knows, a quiet but consequential research front is advancing at the U.S. Department of Energy's National Energy Technology Laboratory (NETL) that could redefine what efficiency means for gas turbines in the decade ahead.

NETL researchers have recently announced a breakthrough in the design of a key component of rotating detonation engines (RDEs) — an emerging class of combustion technology that could enable what engineers call "pressure gain combustion." Unlike conventional gas turbines, which rely on steady, subsonic deflagration that inherently loses pressure across the combustor, RDEs create a controlled, continuous detonation wave that rotates around the inside of a modified combustion chamber. Because detonation generates its own pressure rather than dissipating it, the thermodynamic cycle is more efficient.¹²

"Detonations release energy more rapidly, forming a high-pressure pulse and shock wave," said NETL's Justin Weber, who leads the project. "But transitioning this concept into reliable hardware has long posed scientific and engineering challenges — particularly in ensuring stable, repeatable detonation under varying operating conditions."¹²

NETL's team addressed that challenge using high-fidelity computational fluid dynamics simulations to redesign the injector—the component that feeds fuel and air into the detonation chamber—solving the startup instability problems that had previously hindered reliable operation.¹² Studies suggest that integrating pressure gain combustion into a combined-cycle configuration could yield a 2 to 4 percentage-point improvement in overall cycle efficiency—a gain that would be transformational in an industry where tenths of a percentage point represent years of engineering effort.¹³

For context: if the gas turbine fleet serving AI data centers through 2030 operated at even 2 percentage points higher thermal efficiency, the reduction in fuel consumption and associated carbon emissions would be measured in the tens of millions of tons annually. RDE technology remains at the applied research stage, but NETL's injector breakthrough marks a meaningful step toward hardware that could realistically be integrated into next-generation commercial gas turbines.

What Does This Moment Mean for the Broader Rotating Equipment and Services Ecosystem?

Gas turbines are the headline story, but the turbomachinery industry is far larger than the OEMs who build them, and the implications of the AI data center boom extend through the entire rotating equipment and flow control ecosystem.

According to the JFMA Consulting white paper, centrifugal, reciprocating, and screw compressors are experiencing indirect growth through the natural gas pipeline infrastructure required to supply fuel to new generation capacity. Industrial pumps, valves, and actuators serving liquid handling—including the cooling systems that maintain reliable data center operation—represent a growing niche demand. And nuclear is emerging as a long-term growth vector, with life extension programs for existing plants and future small modular reactor deployments providing significant opportunities for companies capable of meeting the industry's stringent certification requirements.³

The aftermarket and services opportunity may ultimately be the most durable dimension of the cycle. Each new turbine installation creates decades of lifecycle revenue: from maintenance and upgrades to digital optimization and performance services. "As a result, the surge in installations is expected to translate into sustained growth in aftermarket services, reinforcing the strategic importance of fleet-wide support capabilities across the turbomachinery ecosystem," Reisig says.

GE Vernova describes a similar evolution underway in its own services business. "As our HA and AERO fleets expand to support data centers, our services business is evolving from traditional maintenance into an AI-enabled reliability collaboration," Chaaban said. "Unlike many traditional utility customers who often manage their own operational scheduling, hyperscalers demand a 'guaranteed uptime' model that treats our turbines as high-availability compute infrastructure." GE Vernova is addressing that expectation, he said, by integrating long-term service agreements with predictive maintenance software that preemptively manages asset health and reduces unscheduled outages — what Chaaban called "an end-to-end service ecosystem" aimed at the consistent power quality and operational continuity the next decade of high-density compute will demand.

Baker Hughes generated more than $2 billion in new energy revenue in 2025 and expects $2.4 billion to $2.6 billion in 2026, reflecting the rapid expansion of its power systems installed base and the long-cycle aftermarket revenue that follows.⁴ The JFMA white paper notes that aftermarket services already account for more than 50% of revenues at leading OEMs—at margins nearly double those of new equipment—and that 81% of OEMs plan to increase aftermarket investment over the next five years.³ As OEMs concentrate capacity on new equipment, qualified independent service providers stand to gain meaningful market share across growing fleets of recently commissioned data center turbines.

Notably, M&A valuations for rotating equipment and flow control companies with power generation exposure have risen sharply, with both strategic buyers and private equity competing aggressively for attractive assets.³

What Should Turbomachinery Professionals Be Watching?

The Gigawatt Era is already underway. Order books are full, plants are being permitted, and behind-the-meter turbines are being installed. The engineering workforce that designs, commissions, and maintains these machines is in demand at levels not seen in a generation.

The near-term priorities are clear: understanding how the customer base is shifting from utilities to hyperscalers with different expectations around uptime, services, and contract structures; positioning for the wave of new installations that will require decades of lifecycle support; and tracking the technology frontier, where hydrogen co-firing, digital monitoring, and pressure gain combustion are laying the groundwork for the next step-change in efficiency.

The longer arc runs through the decarbonization challenge that cannot be deferred indefinitely. The turbines being sold today will still be operating in 2045 and beyond. The decisions being made now—about fuel flexibility, hydrogen readiness, and emissions performance—will define what the industry looks like when those machines reach the end of their first major service interval.

As Reisig puts it: "Each new installation creates decades of lifecycle opportunity." The question for every company in the turbomachinery ecosystem is whether they are positioned to capture it.

REFERENCES
1. Baker Hughes. "Power Generation Technology Overview." Baker Hughes Annual Meeting 2026, January 28, 2026, Florence, Italy. 2. Baker Hughes. "Key Takeaways from the 2026 Annual Meeting." bakerhughes.com, February 2026. 3. Frank Ma, JFMA Consulting. "The AI Power Surge: How Data Center Growth is Impacting the U.S. Rotating Equipment and Flow Control Markets." Turbomachinery International / World Pumps, April–June 2026. 4. Lorenzo Simonelli, Baker Hughes. Remarks at Bernstein 42nd Annual Strategic Decisions Conference, May 2026. Reported by MarketBeat, May 27, 2026. 5. Data Center Knowledge. "Replacing Diesel in AI-Scale Data Centers: Gas Engines, Turbines, and Steam." 2026. 6. TechTimes. "US Solar Energy Boom Hits 28 Months Running: Gas Turbine Backlog and AI Demand Outweigh Policy Cuts." June 7, 2026. 7. Siemens Energy. "Earnings Release Q2 FY 2026: Following a strong second quarter, Siemens Energy raises its outlook." May 12, 2026. 8. BloombergNEF. "US Gas Power Plants: Higher Costs, Longer Waits." Reported by Bloomberg, April 23, 2026. 9. Baker Hughes Company (BKR) Q1 2026 Earnings Call Transcript. Insider Monkey, April 25, 2026. 10. Baker Hughes. "Baker Hughes Secures 1.21 Gigawatt Generator Order to Power Boom Supersonic's AI Data Center Solution." 2026. 11. Baker Hughes / Google Cloud. "Baker Hughes Develops AI-Enabled Power Optimization and Sustainability Solutions for Data Centers with Google Cloud Technology." GlobeNewswire, March 24, 2026. 12. U.S. Department of Energy / NETL. "NETL Advances Next-Generation Gas Turbine Technology With Breakthrough Rotating Detonation Engine Injector." June 2026. 13. U.S. Department of Energy / NETL. "Pressure Gain Combustion." netl.doe.gov.