Data center power supply: 2030 installation forecast revised upward to 67GW, fuel cells rank first.

The expansion of AI data centers is currently hampered not by funding, but by electricity. While chip and server room orders can be secured in advance, the time it takes to get electricity into the parks is increasing—in the US and Europe, waiting times for access to the public power grid will continue to grow over the next five years. As electricity availability replaces capital expenditure as the hard constraint for deployment, the "off-grid" solution—self-built and self-used, without waiting in line—is being pushed to the forefront.

According to TrendForce, Goldman Sachs' Michele Della Vigna team, in their September 23 report, "Carbonomics: Behind-the-meter power solutions for data centers: gas turbines, fuel cells, and reciprocating engines," presented a framework that scores gas turbines, fuel cells, and reciprocating engines on a level playing field, and revised their 2030 forecast for behind-the-meter power generation capacity upward from 40GW to 67GW. Supporting this upward revision is a set of faster electricity consumption readings.

Two conclusions were reached: all three technical routes will be used; and after reweighting based on delivery speed and availability, fuel cells ranked first in overall score.

Why did the power supply plate behind the meter suddenly get bigger?

The demand-side figures have changed first. Global data center capacity is projected to reach 217GW in 2030, up from 101GW in 2025; US data center demand is expected to reach 108GW during the same period. Based on this, global data center electricity demand in 2030 is projected to increase by 170% compared to 2025, with more than 60% of that growth coming from the US. This growth rate is higher than previously assumed and directly clashes with the grid's delivery capacity.

Delivery capacity on the grid side is deteriorating. The median time from grid interconnection application to commercial operation in the United States is approaching five years for projects completed in 2023-2024; the average annual construction of new high-voltage transmission lines has dropped from approximately 1,700 miles in 2010-2014 to an average of 350 miles in 2020-2023. Transmission and generation facilities typically take five to ten years to build, and the planning process is lagging behind this influx.

The combination of upward revisions in demand and a slowing grid has led to the following conclusion on the back end of the forecast: the installed capacity of back-end power generation has increased from 40GW to 67GW, and all three routes are indispensable.

Where are the three routes stuck?

The bottleneck for conventional gas turbines has shifted from technological maturity to production capacity and delivery time. Under normal market conditions, large-scale gas turbine projects take about two to three years from order placement to commissioning. Currently, the delivery cycle for combined cycle gas turbines is four to seven years, with a single cycle taking 18 to 36 months, and heavy-duty gas turbines generally taking about five to seven years. Manufacturers' capacity expansion is insufficient to immediately change the tight balance: GE Vernova's contract volume will exceed 125GW by the end of the year, with its 2030 capacity already sold out and more than half of its 2031 capacity pre-sold; Siemens Energy has a backlog of nearly 69GW, with another 27GW booked, implying a delivery time of about three to four years; Mitsubishi Heavy Industries' projects under negotiation are scheduled for delivery into the 2030s.

The strengths of fuel cells lie in their speed of deployment and availability, not in their power generation costs. Modular projects have a delivery window of approximately 12 to 24 months, do not require grid connection, have the longest full-load operating time, and require the least amount of additional backup capacity.

This rating system aggregates delivery speed (20%), availability (15%), levelized cost of electricity (LCOE) (15%), and initial capital expenditure (10%), with over 60% of the weighting focused on execution rather than economics. The results showed fuel cells leading with 76.6 points, followed by aero-derivative gas turbines at 68.2 points (the most balanced conventional option), reciprocating engines at 67.0 points, and heavy-duty gas turbines and combined cycle engines at 59.6 points. However, the rating system doesn't mask its weaknesses: the LCOE for fuel cells is approximately $117/MWh, about 45% higher than combined cycle and reciprocating engines.

What are the chances for fuel cells, and who will win them?

According to the article's calculations, fuel cells will correspond to approximately 12GW of installed capacity in the US and 18GW globally by 2030, translating to a cumulative equipment market of approximately US$35 billion (US) and US$55 billion (global). Combined with gas meters, fuel cells could cover approximately 28% of data center electricity consumption in the US and 25% globally by 2030, while this proportion will be close to zero by 2025.

The supply gap is even clearer. At full capacity and 85% utilization, Bloom Energy's 2GW expansion to 2030 would deliver approximately 7.7GW of solid oxide fuel cells, far below the global installation forecast of approximately 18GW. To realize this forecast, multiple manufacturers would need to expand capacity simultaneously, including Ceres Power's licensees Doosan, Delta Electronics, and Weichai Power.

The beneficiaries are separated by route: on the fuel cell side are Ceres Power, Weichai Power, and Delta Electronics, and on the gas meter side are GE Vernova, Siemens Energy, Mitsubishi Heavy Industries, and INNIO.

What's next?

Results metrics are disclosed quarterly by manufacturers: order backlog and year when production capacity is sold out for gas turbine and fuel cell manufacturers, including GE Vernova, Siemens Energy, Mitsubishi Heavy Industries, as well as Ceres Power and its licensees Doosan, Delta Electronics, and Weichai Power.

The mechanism indicators focus on three things: whether the procurement and grid connection announcements for data center self-owned power plants continue to increase, whether the delivery cycle of gas turbines continues to lengthen, and whether the unit cost and delivery cycle of fuel cells are declining as predicted.

Counter-evidence: If the expansion of the public power grid significantly shortens the access queue, or if the cost reduction of fuel cells falls short of expectations and the delivery cycle is lengthened instead, the forecasts for the proportion of electricity supplied after the meter and the share of fuel cells need to be revised downward.

Verification window: October 15, 2026 to November 15, 2026, i.e., the third quarter financial report window, and the disclosure date of subsequent self-provided power supply procurement announcements.

The power bottleneck is reshaping the selection process, which in turn is reshaping the allocation of this round of orders. The 40GW to 67GW figure is an upward revision, not actual capacity already realized. The next focus should be on whether delivery dates and orders continue to shift towards execution, and whether the cost curve for fuel cells is declining as predicted.

~~~~~~~~~~~~~~~~~~~~~~~

The above content is from Zhuifeng Trading Platform .

For more detailed analysis, including real-time updates and firsthand research, please join the [ Trading Channel Annual Membership ].

Risk warning and disclaimerInvesting involves risk; please exercise caution. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific circumstances. Any investment decisions made based on this information are at your own risk.