HSBC: SOFC is poised to power AI data centers, with tight supply through 2030

Overview

This HSBC equity research report discusses the opportunity for solid oxide fuel cells (SOFC) in behind-the-meter (BTM) primary power under the backdrop of rising power demand from AI data centers. The report points out that longer grid connection timelines and constraints on power availability are driving data center operators to adopt on-site generation solutions; although SOFC has higher capex and requires periodic stack replacement, it is gaining traction in more AIDC application scenarios due to fast delivery, no NOx combustion emissions, high electrical efficiency, and native DC output.

Core views

The core view is that SOFC supply may remain tight through 2030, with demand coming from AI data centers, microgrids, clean hydrogen policies, and low-carbon shipping applications in regions such as the US, South Korea, Taiwan, China, and the EU. Bloom Energy is seen as the global SOFC leader, with relatively de-risked expansion plans; Weichai Power, as a gas/diesel engine company also advancing SOFC capacity, is likewise listed as a favored name. Other relevant companies in the value chain include Ceres, Doosan Fuel Cell, Delta, Three-Circle, and Kaori.

Analysis framework

The report uses a combination of thematic value-chain analysis, techno-economic comparison, regional demand breakdown, capacity ramp-up paths, and cost reduction paths. It compares SOFC with gas turbines, combined cycle, and reciprocating internal combustion engines in terms of lead time, efficiency, cost, maintenance, and emissions, and combines survey data from Ceres, Rystad, DOE, and Bloom Energy to assess demand potential and supply bottlenecks.

Methodology notes

• Techno-economic comparisonComparison of behind-the-meter primary power solutions

  Compare SOFC, gas turbines, combined cycle, and reciprocating internal combustion engines in terms of unit scale, delivery cycle, efficiency, capital cost, maintenance, and emissions.

  SOFC's core advantages are a 3-4 month delivery cycle, 55%-65% electrical efficiency, native DC output, and low NOx emissions; disadvantages are high initial CAPEX, stack replacement costs every 5-7 years, and supply bottlenecks for some key materials.

• Market sizingSOFC TAM and regional demand framework

  Break down the 2030 SOFC market opportunity by end use and region.

  Ceres estimates SOFC TAM will reach 22GW by 2030, with the US accounting for about 24% and data center applications about 50%; South Korea, Taiwan, China, the EU, Japan, Southeast Asia, Australia, and India are respectively driven by clean hydrogen policies, RE100, grid fragility, carbon pricing, shipping decarbonization, or off-grid demand.

• Capacity and cost pathExpansion and cost reduction framework

  Track the capacity targets of Bloom Energy and Ceres partners, as well as DOE and Rystad forecasts for system cost declines.

  Bloom targets reaching 2GW by the end of 2026 and potentially 5GW by 2030; Ceres partners target about 100MW by the end of 2026 and about 1GW by 2030. Rystad expects SOFC system costs to fall 20%-25% by 2030, while DOE targets total system CAPEX of USD900/kW by 2030.

• Valuation frameworkDCF and SOTP

  The report mentions Bloom Energy DCF summary and Weichai Power SOTP approach in the valuation and risk sections.

  The source text does not provide complete valuation tables or target price details, so it can only confirm that DCF and SOTP are the valuation frameworks used for the relevant names, without extracting specific valuation parameters.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Bloom Energy

  Global SOFC leader, a favored Buy-rated name in the report, serving the US commercial and industrial market and AIDC, with system scale of about 100-1,000kW.

  Strengths

  Short delivery cycle, leading commercialization, clear expansion targets, and SOFC can provide high efficiency and native DC output.

  Weaknesses

  High initial CAPEX, stack replacement required every 5-7 years, and still affected by key materials and BOP costs.

  Comparison

  Compared with Ceres' technology licensing model, Bloom is closer to a vertically integrated system supplier; compared with gas turbines and generator sets, SOFC has faster delivery and lower emissions, but higher costs.

  Risks

  AIDC construction slowdown, SOFC cost reduction below expectations, and gas turbine and generator set supply recovering faster than expected.

• Weichai Power

  A gas/diesel engine company that is also a Ceres partner and is advancing SOFC capacity ramp-up; the report lists it as a favored Buy-rated name.

  Strengths

  Has an industrial foundation in engines and power equipment, participates in the SOFC partner ecosystem, and is positioned to benefit from data center and distributed power demand.

  Weaknesses

  Its SOFC business is still in the capacity ramp-up stage, and commercialization scale and the cost curve still need validation.

  Comparison

  Compared with Bloom, Weichai's SOFC positioning is more that of a partner and capacity expander; compared with traditional generator sets, it may benefit from re-rating under the clean power theme.

  Risks

  SOFC capacity release slower than expected, cyclical volatility in traditional engines, and AIDC power demand or policy support below expectations.

• Ceres Power

  An SOFC technology developer and licensor, with partners including Weichai, Doosan, and Delta.

  Strengths

  Covers South Korea, Taiwan, China, and other regional markets through a partner model, participating in the expansion of 2030 SOFC TAM.

  Weaknesses

  Commercialization progress depends on partner capacity expansion and end-order conversion.

  Comparison

  Unlike Bloom's system supply model, Ceres relies more on licensing and ecosystem partnerships, with upside from partner capacity expanding from about 100MW toward the 1GW level.

  Risks

  Partner capacity ramp-up below expectations, regional policy changes, and SOFC cost declines falling short of expectations.

• Doosan Fuel Cell、Delta

  Respectively leaders or partners related to SOFC in South Korea and Taiwan, China, benefiting from South Korea CHPS and Taiwan, China's RE100 and grid stability demand.

  Strengths

  Supported by local policy and demand drivers, including South Korea's clean hydrogen portfolio standard and Taiwan, China's semiconductor/AI power demand.

  Weaknesses

  There is still uncertainty around regional market size, subsidy continuity, and local deployment speed.

  Comparison

  Compared with the US AIDC market, demand in South Korea and Taiwan, China is more driven by policy, grid fragility, and semiconductor manufacturing constraints.

  Risks

  Subsidy or policy adjustments, microgrid project delays, and intensified local competition.

• Three-Circle、Kaori

  Key supply-chain companies for Bloom Energy, involved in electrolyte and hot box supply respectively.

  Strengths

  Benefit from Bloom's expansion and demand for key components in the SOFC system cost structure.

  Weaknesses

  Customer concentration and dependence on technical specifications may be relatively high.

  Comparison

  Compared with system suppliers, component companies are more directly exposed to capacity expansion and the path of declining BOP/stack costs.

  Risks

  Volatility in key material supply, slowdown in customer expansion, and onboarding of alternative suppliers.

Key data

• SOFC delivery cycle3-4 monthsShorter than about 1-2 years for engines and about 3-5 years for gas turbines, making it a key advantage for entering AIDC behind-the-meter power.
• SOFC electrical efficiency55%-65%The report says its high electrical efficiency and native DC output can reduce AC-DC conversion losses.
• SOFC 2030 TAM22GWBased on Ceres forecasts; the US accounts for about 24%, and data center applications about 50%.
• Bloom Energy capacity target2GW by the end of 2026, potentially 5GW by 2030The report believes Bloom's expansion plan is broadly de-risked.
• Ceres partner capacity targetAbout 100MW by the end of 2026, about 1GW by 2030Partners include Weichai, Doosan, and Delta.
• US data center BTM projects11GW of announced projects; 85% plan to use natural gas, 10% target renewable energy, 4% new-build nuclearRystad estimates about 40% of new US data center capacity in 2026-2030 will be behind the meter.
• Expected adoption of on-site powerAbout one-third of developers expect to adopt fully permanent on-site power by 2030; rising to 45% by the end of 2035From Bloom Energy's survey of US data center developers.
• Expected SOFC cost declineRystad expects system costs to fall 20%-25% by 2030; DOE 2030e target USD900/kWThe report also mentions current or baseline system cost of about USD3,000/kW, with BOP accounting for about 70% of system cost.

Impact & implications

The investment implication is that AI data center power bottlenecks may shift on-site primary power from emergency backup to long-term infrastructure deployment, benefiting SOFC suppliers and key component suppliers through higher order visibility and capacity expansion. If cost declines, policy support, and the pace of data center construction materialize, Bloom Energy, Weichai Power, and the Ceres ecosystem could see upward revisions to valuation and earnings expectations; conversely, if gas turbine and generator set lead times improve, or AIDC construction slows, SOFC penetration may be lower than expected.

Risks

• AIDC data center construction is slower than expected.
• SOFC capital expenditure remains high, and the cost reduction curve falls short of expectations.
• Gas turbine and gas/diesel generator set capacity expands faster than expected, weakening SOFC's lead-time advantage.
• Scandium, ceramic electrolytes, and Balance of Plant components become bottlenecks in supply.
• SOFC stack replacement every 5-7 years brings ongoing operating and maintenance costs.
• Changes in regional subsidies, carbon pricing, RE100, or clean hydrogen policies affect project economics.
• Natural gas as the main fuel for on-site power may face decarbonization constraints and fuel price volatility.

What to watch

• Whether Bloom Energy reaches 2GW of capacity by the end of 2026 as planned and advances toward 5GW before 2030.
• Whether Ceres partners Weichai, Doosan, and Delta achieve targets of about 100MW by the end of 2026 and about 1GW by 2030.
• Whether the 20%-25% system cost decline projected by Rystad and the DOE USD900/kW target are realized.
• The fuel mix of US data center BTM projects and the implementation progress of the 11GW of announced projects.
• Whether the adoption rate of permanent on-site power by data center developers rises from about one-third in 2030 to 45% in 2035.
• The pull on SOFC demand from South Korea CHPS, Taiwan, China's RE100 and subsidies, EU carbon pricing, and IMO shipping decarbonization policies.
• Whether delivery cycles for gas turbines, combined cycle, and reciprocating internal combustion engines shorten.