The core challenge in artificial intelligence infrastructure construction is shifting: from a singular pursuit of GPU quantity to a race for the speed of power supply. As "Time-to-Power" becomes the most severe bottleneck in AI development, Chinese power solution providers, with their rapid delivery capabilities and advantages in mass production, are encountering a historic opportunity for value reassessment.
According to information, a Goldman Sachs team led by Jacqueline Du stated in a report released on the 14th that the average lifespan of power grids in the United States and the European Union has reached 35 to 40 years. Faced with the explosive energy demand from AI data centers (AIDCs), this infrastructure appears increasingly fragile. Currently, domestic U.S. power equipment production capacity can only meet about 40% of local demand, while grid connection wait times have extended to nearly five years. This supply-demand mismatch is forcing U.S. utility companies and data center operators to break with tradition and begin accepting non-traditional suppliers to fill this multi-year supply gap.
This structural shortage is reshaping pricing power within the supply chain. Goldman Sachs emphasized in its report that for qualified Chinese suppliers, the decisive advantage is no longer just low cost, but significantly shorter delivery lead times. Driven by severe shortages, Chinese companies' products in overseas markets command a substantial premium compared to domestic sales, with the premium ranging from 10% to 80%, providing these companies with high earnings visibility.
Based on this, Goldman Sachs has expanded its coverage of the AIDC power supply chain, focusing on companies that can solve the "power delivery" bottleneck. The firm believes that Chinese enterprises possessing advanced high-voltage direct current (HVDC) expertise, high-density power conversion capabilities, and mature OEM/ODM relationships will leverage this spillover demand to achieve rapid expansion.
U.S. Power Bottleneck: Supply Gap to Persist Until 2030 U.S. data center power demand is undergoing explosive growth. Goldman Sachs forecasts that U.S. data center electricity consumption (including both AI and non-AI) will increase by approximately 175% from 2023 levels by 2030. This will contribute about 120 basis points to overall U.S. electricity demand, pushing the average annual growth rate to 2.6%.
The power supply shortage is particularly acute in the United States. Goldman Sachs data indicates that between 2025 and 2030, the U.S. summer peak effective reserve power capacity is expected to decline further, while China's reserve capacity is projected to rise. Even if U.S. domestic power transformer suppliers complete their announced capacity expansion plans, they will only be able to meet about 40% of local demand by 2027.
On the power generation side, gas turbines represent the most critical bottleneck. Goldman Sachs points out that approximately 60% of AI data center power is expected to come from natural gas, due to its fast construction times, stable output, lower carbon emissions, and ability to enable on-site generation that bypasses grid constraints. Order backlogs for major gas turbine manufacturers like Siemens Energy, General Electric, and Mitsubishi Heavy Industries have reached 4.5 to 5 years, with Siemens Energy explicitly stating its gas turbine capacity is sold out through 2028.
In the transmission and distribution segment, power transformers face the most urgent supply shortage. Siemens Energy predicts a transformer shortage rate of around 30% in the EU and U.S. for 2025, expected to ease to about 10% by 2030. Because each transformer must be custom-built to unique specifications for impedance, cooling, tap changers, overload capacity, and seismic standards, the production process is labor-intensive and lengthy.
39% CAGR: 800V DC Architecture Drives Technology Upgrade Cycle Goldman Sachs estimates that the total addressable market for AI data center power products will expand at a compound annual growth rate (CAGR) of approximately 39% between 2025 and 2030. This encompasses seven major product categories: gas turbines, power transformers, uninterruptible power supply systems, server power supplies, electrical components, and liquid cooling systems.
This growth is underpinned by three key drivers: ongoing capacity build-out, continuously increasing power density, and a structural shift from predominantly alternating current (AC) to direct current (DC) architectures. As AI rack power escalates from 100kW to higher levels, even approaching 1MW, engineering complexity surges across rectification, protection, heat dissipation, and energy storage integration, creating a larger value pool for DC-ready components and grid-connection assets.
The 800V DC power distribution architecture is becoming the standard for most new-build AI data center greenfield projects, with some pilot projects already exploring ~800V DC busbars. Goldman Sachs notes that compared to traditional AC topologies, adopting higher voltage DC distribution can save approximately 5-15% in energy consumption at the facility level. Key players like NVIDIA have fully committed to an 800V DC architecture roadmap, and major ecosystem participants are aligning their product roadmaps towards DC-ready power conversion, protection, and battery energy storage system integration.
This architectural upgrade creates a structural opportunity for suppliers of DC power infrastructure components. Goldman Sachs is particularly positive on Hongfa (宏发股份宏发股份) at the high-voltage DC relay level and Jianghai (江海股份江海股份) at the capacitor level, believing that existing AIDC-related orders provide a more direct boost to their core businesses. As duration requirements increase, product portfolios should tilt towards higher-specification HVDC relays and supercapacitors (with electric double-layer capacitors growing first, followed by lithium-ion capacitors).
Delivery Speed Decides the Winner: The Core Competitiveness of Chinese Suppliers Goldman Sachs' analysis of eight key product categories indicates that for qualified Chinese suppliers capable of capturing spillover demand, the decisive competitive advantage is not just lower cost, but crucially, shorter delivery lead times.
Faced with wait times of 3-5 years for critical components, data center operators and utility companies have prioritized delivery speed as the primary decision factor, which can outweigh preferences for historical suppliers. Although challenges remain in global service capabilities, the value proposition of rapid delivery combined with acceptable quality is highly compelling in the current supply-constrained environment.
Taking power transformers as an example, Sieyuan Electric's (思源电气思源电气) short delivery cycle enables it to gain market share in the U.S. Goldman Sachs expects the proportion of U.S. market revenue in Sieyuan's overseas revenue to increase from 26% in 2026 to 28% in 2028. The firm notes that only a few Chinese companies can combine high quality with long-term commitments, navigating the market through rigorous certification processes, sustained upfront investment, and proven overseas track records, areas where Sieyuan excels.
In the gas turbine blade sector, Yinglite (英力特英力特), a leading domestic manufacturer of high-end precision castings, currently holds less than 1% of the global market share, indicating substantial growth potential. The company has signed long-term agreements with Baker Hughes, Ansaldo, and GE Aviation. Goldman Sachs believes Yinglite will benefit from the gas turbine supply shortage, growing as a supplementary supplier.
Pricing Power Emerges: Overseas Order Gross Margin Premiums of 10%-80% Due to severe supply shortages, Chinese suppliers can achieve significant price premiums in overseas markets, ranging from 10% to 80% compared to domestic sales. Goldman Sachs points out that although production remains concentrated in China to maintain the critical delivery cycle advantage, the resulting margin uplift remains significant even after accounting for increased costs related to tariffs and logistics.
This opportunity is not universally available but is concentrated among a select few top-tier Chinese industrial enterprises possessing scale, quality control, and the capability to serve international customers. For instance, in power transformers, Sieyuan Electric's products sold in the U.S. achieve gross margins of approximately 45%, compared to around 30% for domestic sales. Huaming Power's (华明电力华明电力) overseas sales gross margin is about 40%, versus roughly 25% domestically.
In the uninterruptible power supply system segment, Kehua's (科士达科士达) overseas ODM model commands a 25-50% pricing premium compared to orders from domestic hyperscale customers. Goldman Sachs forecasts that Kehua's overseas high-power electrical system sales will surge from RMB 100 million in 2025 (2% of total revenue) to RMB 800 million in 2026 and RMB 1.782 billion in 2028 (11% and 15% of total revenue, respectively), primarily driven by order execution for European-based clients serving end-users in the United States.
Goldman Sachs calculations show that with exposure to the U.S. market, the covered companies could achieve an average sales CAGR of 23% between 2025 and 2030, compared to 17% without U.S. exposure. By 2030, the contribution of overseas AIDC market revenue (primarily from the U.S.) to these Chinese companies' total revenue is projected to average 23%, with an average global market share of 4%.
Product Priority Ranking: Gas Turbine Blades > Power Transformers > Electrical Components Goldman Sachs provides a clear preference ranking for Chinese power supply-related product categories: Gas Turbine Blades > Power Transformers > Electrical Components > UPS/Power Shelves > Liquid Cooling Systems > Server Power Supplies.
Gas turbine blades rank highest due to extremely high barriers in material science and manufacturing—particularly single-crystal high-temperature segment technology—coupled with severely constrained global effective capacity. Power transformers follow closely, benefiting from labor-intensive manufacturing and lengthy certification cycles. Electrical components benefit from strong demand visibility and broad exposure to both AI and grid infrastructure build-out, but possess less structural advantage in delivery cycles or differentiation due to easier scalability of manufacturing.
UPS/Power shelf systems benefit from differentiated ODM business models that enable companies like Kehua to enter the U.S. market; liquid cooling systems also enjoy structural growth as both volume and content value increase with rising power density, but face relatively intense global competition. Server power supplies rank lowest because the product/scale gap compared to global peers remains significant, and rapid technological iteration introduces uncertainty regarding potential disruption.
Comments