Zhongtai: SOFC has achieved a certain level of economic viability for use as primary power for AIDC; North America's AIDC power shortage opens up growth space.

Zhongtai released a research report stating that solid oxide fuel cells (SOFC), as a high-temperature fuel cell, use solid ceramic electrolyte as the core to directly convert chemical energy in fuel into electrical energy through electrochemical reactions, without combustion and unconstrained by the Carnot cycle, playing a key role as "on-site primary power" in the power supply system of AI data centers. North America's AIDC power shortage opens up growth space. According to the bank's estimates, under a neutral assumption, SOFC shipments will reach 8.5GW by 2030, with a market size of approximately US$17.8 billion. Zhongtai's main points are as follows: SOFC: Solid oxide fuel cells (SOFC), as a high-temperature fuel cell, use solid ceramic electrolyte as the core to directly convert chemical energy in fuel into electrical energy through electrochemical reactions, without combustion and unconstrained by the Carnot cycle, playing a key role as "on-site primary power" in the power supply system of AI data centers. Its power generation efficiency is 50%60%, combined heat and power efficiency can reach up to 90%, carbon emissions are more than 30% lower than traditional natural gas units, and it features near-zero NOx/SOx emissions and zero water consumption. In terms of technology routes, planar and medium-temperature oxygen-ion electrolytes are currently mainstream; the support structure is evolving from electrolyte-supported (represented by Bloom Energy) to metal-supported (represented by Weichai Power, Ceres Power), with the latter constituting the long-term direction due to advantages of low temperature, low cost, and rapid startup. 1) SOFC has achieved a certain level of economic viability for use as primary power for AIDC: The construction period can be as short as 90 days, grid connection requires 25 years or more, and the average delivery cycle for gas turbines is as long as 5 years; According to the U.S. DOE's calculation for a 100MW-class hydrogen-fueled SOFC, LCOE is approximately US$125/MWh (without ITC), and can drop to approximately US$100/MWh under a 40% ITC scenario, lower than simple-cycle gas turbines; Under the IRA framework, the maximum SOFC tax credit can reach 50%, and combined with cost reductions, the DOE targets system costs falling below US$900/kW before 2030. 2) North America's AIDC power shortage opens up growth space: On the demand side, the explosive growth of AI computing power in North America and the construction of high-density data center clusters have driven a surge in electricity consumption, putting enormous pressure on traditional local grid loads; On the supply side, the grid is aging and expansion is extremely slow, high volatility of wind, solar, and storage cannot provide sustained and stable power supply, gas turbine capacity is insufficient, and there is uncertainty over carbon emission policy after the 2028 U.S. election; Commercialization is achievablecases such as Oracle (up to 2.8GW), AEP (1GW framework), Equinix (100+MW), and CoreWeave (14MW) have been implemented successively. Bloom Energy estimates that its 800V DC power supply solution can reduce non-computing CAPEX of a 1GW AI data center by 27% and five-year TCO by 9%. According to the bank's estimates, under a neutral assumption, SOFC shipments will reach 8.5GW by 2030, with a market size of approximately US$17.8 billion. 3) Landscape and targets: Global competition is characterized by U.S. leaders dominating, with China, Japan, South Korea, and Europe accelerating their catch-up. Bloom Energy accounts for approximately 75% to 85% of the global SOFC share, with revenue of US$1.816 billion in 2026H1, up 149.76% year on year, a record-high gross margin of 32%, and a positive net margin; its "equipment sales + long-term service + asset-light cooperation" model opens up expansion space. Looking at the domestic industrial chain by segment: in the complete-system segment, it has cooperated with Ceres to build SOFC complete-system R&D and industrialization capabilities and leads domestic commercialization; in the core materials segment, it is the main supplier of membrane electrode assemblies to Bloom Energy; in the components segment, it is one of Bloom Energy's main temperature sensor suppliers with a share of over 70%; in the key resources segment, a single GW of connectors consumes approximately 5,000 tonnes of metallic chromium, with existing capacity of 30,000 tonnes (full production in 2027); in the testing segment, it won the bid for CGN's SOFC power generation system testing service project, positioning itself in third-party testing and verification; in the full-stack self-development segment, it has achieved independent supply across the complete chain of "powdercellstacksystem," with a 120kW demonstration project expected to be completed and put into operation by the end of 2026. Risk warnings: North American AI capital expenditure falling short of expectations; SOFC cost reduction and sintering yield improvement progressing slower than expected; deviations in penetration rate assumptions and market size estimates; grid expansion or gas turbine capacity recovery exceeding expectations; changes in carbon emission policy after the U.S. election; domestic SOFC industrialization progressing slower than expected; fluctuations in prices of raw materials such as metallic chromium; differences in statistical calibers in third-party data cited in estimates; risks of deviations in market space estimates; risks that public information used in the research report may be outdated or not updated in a timely manner.