Power supply capacity is becoming an invisible ceiling for data center construction.
Shenwan Hongyuan Securities released a deep industry report titled &#34;Computing Power Demand Growth Exacerbates Power Gaps, Opening Growth Space for Gas Turbine Markets&#34; on April 16. It points out that the accelerated construction of North American AI Data Centers (AIDCs) is pushing power supply contradictions to a critical point.
In the short term, gas turbines and reciprocating internal combustion engines have become the core power supply solution for AIDCs at this stage, leveraging advantages in rapid start-stop and flexible deployment. Aero-derivative gas turbines require only 5-10 minutes to start, while heavy-duty gas turbine combined cycles can achieve efficiencies exceeding 60%. In the medium to long term, SMRs (Small Modular Reactors), controlled nuclear fusion, and SOFCs (Solid Oxide Fuel Cells) are considered ideal directions for long-term AIDC power supply.
<h2>Gas Turbines: The Most Effective &#34;Firefighting&#34; Solution Now</h2>
Facing the immediate power supply needs of AIDCs, Shenwan Hongyuan believes that gas turbines and reciprocating internal combustion engines are the most feasible core solutions at this stage.
This is due to the high compatibility of their technical characteristics. Photovoltaic and wind power belong to intermittent energy sources and struggle to meet all-day stable power supply demands; nuclear power plants and large-scale hydroelectric projects typically have construction cycles of over 4-5 years, failing to match the construction rhythm of AIDCs. In contrast, gas turbines respond quickly, offer flexible regulation, and have short construction cycles, making them the &#34;mainstream solution for supporting continuous loads and emergency backup power for AIDCs at this stage.&#34;
Within gas turbines, two types of models have distinct focuses:
Aero-derivative Gas Turbines are derived from modified aircraft engines, requiring only 5-10 minutes to start, with typical output ranging from 5-60MW. They feature small footprints and mobile deployment, adapting to scenarios such as distributed energy supply and emergency peak shaving. Representative models include GE's LM2500+ series and LM6000 series, Siemens Energy's SGT-A35, and Mitsubishi Heavy Industries' FT4000 based on the Pratt &amp; Whitney PW4000 engine.
Heavy-Duty Gas Turbines offer larger power outputs, with typical ratings of 50-500MW and start-up times of 30-60 minutes, suitable for baseload power supply requirements of large AIDCs. After matching with combined cycle systems, power generation efficiency can exceed 60%, representing a ceiling-level technology route for fossil fuel power generation. Classified by turbine inlet temperature, heavy-duty gas turbines range from Class D (power &lt;100MW) to Class J (&gt;1600°C, power 300-400MW), improving progressively.
Reciprocating Internal Combustion Engines complement gas turbines. They start up even faster (reaching full load in 5-10 minutes), have single-unit capacities of 3-20MW, and construction cycles of 15-24 months with controllable costs. They mainly cover power supply for small-to-medium AIDCs and extreme backup power scenarios. For instance, Meta's actual deployment at its Ohio data center plant utilized a combination of Solar Titan 250 and Siemens Energy SGT 400 gas turbines alongside CAT 3520 gas internal combustion engines.
<img src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/634118c9-622b-4deb-a915-847b394a355d.png?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="" original-src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/634118c9-622b-4deb-a915-847b394a355d.png"/>
<img src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/efc387ce-34ba-4216-9528-b3ff3b32bf9a.png?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="" original-src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/efc387ce-34ba-4216-9528-b3ff3b32bf9a.png"/>
<h2>Medium to Long Term: Three Technical Routes Open Long-Term Space</h2>
The Shenwan Hongyuan report indicates that while gas turbines are a &#34;firefighting&#34; solution, the long-term power supply demand for AIDCs points to three frontier technology routes: SMRs, controlled nuclear fusion, and SOFCs.
SMR (Small Modular Reactors)
The International Atomic Energy Agency (IAEA) defines SMRs as nuclear reactors with a single unit rated output between 10-300 MWe. Compared to traditional large-scale nuclear power plants, SMRs adopt modular designs allowing &#34;Lego-like&#34; construction, significantly reducing construction risks and costs. IAEA's latest forecast shows that by 2050, nuclear power generation capacity will increase to 2.5 times the current level, with SMRs accounting for 25%.
Tech giants have entered the field intensively. According to the report: Amazon is collaborating with X-energy to deploy over 5 gigawatts of SMRs in the US, targeting completion by 2039; Google signed a power purchase agreement with Kairos Power, aiming to build its first SMR by 2030; Microsoft signed a 20-year power purchase agreement with Constellation to restart the Three Mile Island nuclear plant; Meta funded TerraPower's two nuclear reactor projects with a total generation capacity of 690 megawatts and signed an agreement with Oklo to develop a 1.2-gigawatt nuclear energy technology park in Ohio; Oracle is designing a data center expected to be powered by more than 1 gigawatt from three small nuclear reactors.
The report suggests that the economic viability of SMR power stations is expected to rival natural gas plants after 2030.
<img src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/c40b6171-1520-442b-8654-f2e0a604c1fb.png?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="" original-src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/c40b6171-1520-442b-8654-f2e0a604c1fb.png"/>
Controlled Nuclear Fusion
Nuclear fusion is regarded as the &#34;ultimate energy source,&#34; currently divided into magnetic confinement and inertial confinement technical paths. Shenwan Hongyuan believes magnetic confinement fusion &#34;has greater engineering feasibility and potential for medium-to-long-term industrialization,&#34; with representative facilities including ITER (global) and EAST (China).
Tech giants are also actively laying out strategies. Microsoft is partnering with Helion to advance pulsed-field reversed-field configuration technology, signing the world's first fusion power purchase agreement, with completion expected by 2028; Google is developing high-temperature superconducting tokamaks with CFS and hydrogen-boron fusion with TAE; NVIDIA invested in CFS, focusing on AI + fusion simulation directions.
<img src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/f29c2954-04b2-4ddb-b1d2-b1e16783f2fb.png?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="" original-src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/f29c2954-04b2-4ddb-b1d2-b1e16783f2fb.png"/>
SOFC (Solid Oxide Fuel Cells)
SOFCs operate at temperatures between 600-1000°C, directly compatible with various fuels such as natural gas, coal-derived gas, and biogas. With power generation efficiency of 45%-60%, combined heat and power systems can achieve overall energy utilization efficiency exceeding 90%. US-based BloomEnergy has achieved large-scale commercial application, deploying multiple distributed power generation systems for Apple, Walmart, Google, and other enterprises, with power generation efficiency around 60% and combined heat and power efficiency nearing 90%.
The common characteristics of these three routes are: environmental friendliness, high energy density, and long-cycle stable power supply, highly matching the long-term needs of AIDCs. Shenwan Hongyuan believes that 2030 is expected to be a key node for commercial implementation.
<img src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/5c89e713-0b92-4d44-9841-4547c72c958c.png?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="" original-src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/5c89e713-0b92-4d44-9841-4547c72c958c.png"/>
<h2>Structural Mismatch in Power Supply</h2>
The report judges that current data center construction does not face a comprehensive power shortage but rather a structural mismatch. According to EIA (US Energy Information Administration) data, total US power generation in 2024 was 4.31 trillion kilowatt-hours, with total consumption at 3.98 trillion kilowatt-hours, indicating no overall shortage. However, 70% of US transmission lines and transformers have been operating for over 25 years, and 60% of circuit breakers have operated for over 30 years.
Meanwhile, AIDC electricity consumption is highly concentrated in tech industry clusters like Silicon Valley in California, Texas, and Virginia, creating a dual mismatch in time and space between aging grids and explosive demand.
Shenwan Hongyuan proposed a noteworthy judgment in the report: &#34;Power supply capacity largely affects the scale and implementation pace of AIDC construction. The importance of power supply solutions may even surpass computing hardware, becoming a bottom-level constraint condition in AI large model training and commercialization promotion that cannot be ignored, rather than simply an auxiliary supporting link.&#34;
The US Department of Energy (DOE)'s &#34;Resource Adequacy Report&#34; also explicitly warns that due to the cumulative effect of power plant retirements and power load growth, US blackouts could increase by 100% by 2030.
<img src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/bf615004-b06b-42ed-9d69-0e1ee0936e14.png?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="" original-src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/bf615004-b06b-42ed-9d69-0e1ee0936e14.png"/>
Shenwan Hongyuan proposes two main investment tracks in the report.
First Track: Focuses on mature power generation equipment at this stage, including manufacturing of complete gas turbines and reciprocating internal combustion engines, core components, system integration, and operation and maintenance services. The report mentions companies with assembly capabilities such as Aviation Power Dynamics, China Power, Dongfang Electric, as well as core links in the industry chain including Yingliu Shares, Wanzhe Shares, Hangyu Technology, and Longda Shares.
Second Track: Forward-looking layout of frontier power generation technologies, focusing on R&amp;D and industrial implementation of SMR nuclear reactor design and manufacturing, nuclear fusion core equipment, and SOFC fuel cell technologies. Related targets for nuclear fusion include Hefei Smart Machinery, Lianchuang Optoelectronics, and Western Superconducting Technologies.
The report also highlights three types of risks: slower-than-expected progress in AIDC construction, technology route substitution risks, and slower-than-expected market expansion (including potential impacts of geopolitical factors on overseas markets).
<img src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/33f25ad0-3632-4972-a46d-26f6ca9adb66.png?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="" original-src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/33f25ad0-3632-4972-a46d-26f6ca9adb66.png"/>

The power gap in North American AI data centers continues to widen. In the short term, gas turbines and reciprocating internal combustion engines have emerged as core emergency solutions due to their rapid start-stop capabilities and flexible deployment. In the medium to long term, Small Modular Reactors (SMRs), controlled nuclear fusion, and Solid Oxide Fuel Cells (SOFCs) are viewed as ideal directions. Power supply capacity is becoming a fundamental constraint on AI development, potentially surpassing the importance of computing hardware itself

- The report by Shenwan Hongyuan Securities highlights the urgent power supply challenges faced by the rapidly growing North American AI data centers (AIDC), identifying gas turbines and reciprocating internal combustion engines as the primary short-term solutions.  
- In the long term, the report suggests that small modular reactors (SMR), controllable nuclear fusion, and solid oxide fuel cells (SOFC) are the most viable technical pathways for meeting AIDC's energy demands.  
- Structural mismatches in electricity supply, particularly in tech hubs, could severely impact AIDC growth, with projected increases in power outages by 2030 due to aging infrastructure and rising demand.

AIDC Power Supply: Short-Term Reliance on Gas Turbines for Emergency Response, Long-Term Driven by Three Key Routes