The Zhitong Finance App learned that Zhongtai Securities released a research report saying that the global AI arms race is driving a high increase in computing power capital expenditure, and the availability of electricity has become the primary bottleneck in AIDC construction. Although the unit cost of traditional gas turbines is lower, the host delivery and station construction cycle cannot match the time window for AI production, and self-provided power supplies change from optional to mandatory; with standardized modular superposition, SOFC can be quickly replicated on a scale of hundreds of MW to bypass grid-connection and transmission bottlenecks, and is being upgraded from alternative technology to the mainstream option for AI power supply.
The main views of Zhongtai Securities are as follows:
The global AI arms race has driven a high increase in computing power and capital expenditure, and the availability of electricity has become the primary bottleneck in AIDC construction
Global artificial intelligence is developing rapidly, and technology giants are in an arms race around computing power. Data center capital expenditure has risen sharply, and electricity demand has surged; however, the US power grid is aging, grid-connected queuing, and transmission capacity expansion is slow, and the delivery cycle of superimposed large-scale power generation equipment continues to lengthen, and the time for AIDC to wait for grid-connected and electrical equipment to arrive is often measured in years. Although the unit cost of traditional gas turbines is lower, the main engine delivery and the station construction cycle cannot match the time window for AI production. Electricity has become the first constraint for AIDC to start production. Self-owned power supplies have changed from optional to mandatory. Whoever can supply power faster can lock in computing power production capacity first, and this will hopefully lead to the AI era of card slots.
The data center launch window was narrowed, and SOFC solutions with immediate delivery capacity were the first to benefit
With mature technology and large-scale production capacity, gas turbines have become the preferred route for AIDC's on-site power supply, but the product delivery cycle is over 36 months, and orders before 2028 have been locked. Other on-site energy forms such as reciprocating internal combustion engines, aviation conversion and combustion, decommissioned aviation development and modification have limited production capacity, making it difficult to fill the demand gap; with standardized modular superposition, SOFC can quickly replicate on a scale of hundreds of MW to bypass grid-connection and transmission bottlenecks. The technical level can also match the future 800VDC AIDC power supply architecture. SOFC is choosing to upgrade from technology to AI power supply Of Mainstream options.
Pay attention to BE's industrial chain, suppliers with better card slots are expected to benefit significantly
The global SOFC player pattern is relatively concentrated. The US company Bloom Energy has deployed about 1.8 GW and has a backlog of about 20 billion US dollars. It is the core integrator in the SOFC industry chain. Based on Bloom Energy successively receiving major GW AIDC power supply orders of AEP up to 1 GW and Oracle's maximum 2.8 GW, it is expected to become the core catalyst for the SOFC industry chain in the future. At present, many companies have clearly formed direct or indirect product supply for BE's SOFC system. In the future, as BE's procurement scale in 2027 is gradually clarified, more companies with card slot advantages are expected to introduce the BE supply chain system, thereby significantly benefiting from the rapid increase in SOFC demand.
Investment advice
The focus is on Bloom Energy, Sanhuan Group, pilot base power, Chunhui Instrument, Zhenhua Co., Ltd., Jingquanhua, Yilian Technology, Dechang Electric Holdings, Weichai Power, Yilian Technology, Delta Electric, Ceres Power, Fuelcell Energy, Doosan Group, Yunyi Electric, Kaizhong Precision, Keliyuan, Zhongyuan internal distribution, ice wheel environment, etc.
Risk warning: risk of AI computing power demand and capital expenditure falling short of expectations; risk of technology route competition; BE order execution and revenue conversion risk; risk of US policy and tax credit; risk of geographical relationships and tariff uncertainty; risk of untimely update of research report usage information, etc.