The AI revolution is not just about algorithms and computing power—it also hinges on energy and physical infrastructure. Beyond chips and servers, the power systems fueling AI data centers, cooling technologies ensuring stable operations, and the metals required for construction are forming a new investment frontier.
According to a recent report by Bank of America Global Research, China's "non-IT" infrastructure capital expenditure market for AI is projected to reach 800 billion yuan by 2030. The report forecasts global AI-related capital expenditures to exceed $1.2 trillion by 2030, with China playing a pivotal role. The country's total AI capital spending is expected to grow at a 25-30% compound annual growth rate (CAGR) from 600-700 billion yuan in 2025 to 2-2.5 trillion yuan by 2030.
Energy supply is considered the cornerstone of AI development. Approximately one-third of these investments—amounting to 800 billion yuan—will be allocated to non-IT infrastructure supporting AI data centers. This includes power generation and transmission (38%), metals for construction (12%), advanced cooling systems (10%), and other engineering projects.
This trend presents clear investment opportunities in nuclear power, grid equipment, energy storage, backup power systems, and advanced cooling technologies, with ripple effects reaching upstream metal markets like copper, aluminum, and uranium.
As AI computing demand surges, the global AI race is shifting focus from pure computing power to energy. The International Energy Agency (IEA) predicts China's data center electricity consumption will rise from 102 TWh in 2024 to 277 TWh by 2030, an 18% CAGR.
**Powering AI: Five Key Opportunities** "No power, no AI." AI model training and inference require massive computational power, translating to equally enormous energy consumption. The report identifies three key drivers behind soaring electricity demand: 1. AI data centers rapidly replacing traditional ones. 2. High-performance chips like Nvidia’s Blackwell architecture, with its GB200 chip consuming 2.7 kW—far exceeding predecessors. 3. Rising server rack power density, with Nvidia’s next-gen Rubin Ultra NVL576 architecture potentially reaching 600 kW in thermal design power (TDP).
China holds advantages over the U.S. and Europe in power reserves, costs, renewable supply chains, grid infrastructure, and equipment. Estimates suggest China’s grid reserve margin will be ~30% in 2025, versus under 25% in the U.S. and ~15% in the EU. Additionally, China’s industrial electricity prices are 30-60% lower, with grid infrastructure averaging under 20 years old compared to over 40 years in Western markets.
This power advantage unlocks five major investment opportunities: 1. **Nuclear Power & Uranium**: Stable, efficient, and low-carbon, nuclear is ideal for AI baseload. China’s nuclear capacity is projected to grow from 60 GW in 2025 to 100 GW by 2030, driving uranium shortages and price hikes. 2. **Grid Equipment**: Global grid upgrades and AI-driven demand are fueling transformer shortages, with Chinese suppliers well-positioned to fill gaps. 3. **Energy Storage (ESS)**: Critical for stability, global ESS installations are forecast to grow at 21% CAGR (2024-2030), with Chinese firms seeing over 30% order growth. 4. **Diesel Generators**: As backup power, this market is expected to grow at 28% CAGR (2024-2027). 5. **Specialty Power Supplies**: High-voltage DC (HVDC) and power supply units (PSU) for AI servers face rising value and technical demands.
**Cooling AI: Liquid Cooling & Metal Demand Surge** High-performance chips generate intense heat—server reliability can drop 50% per 10°C temperature rise. With air cooling insufficient for AI’s power density, liquid cooling is becoming essential. Bank of America predicts China’s liquid cooling market will expand at 42% CAGR (2025-2030), reaching 45% penetration by 2030. Liquid cooling offers 20-50x better heat transfer efficiency and up to 30% energy savings versus air cooling. Emerging technologies like immersion cooling are gaining traction.
Metals like copper and aluminum are equally vital: - **Copper**: Essential for power transmission and cooling, AI data centers could drive nearly 1 million tons of copper demand in China by 2030 (5-6% of total demand). Grid and equipment needs will further amplify consumption. - **Aluminum**: Critical for structural components and cooling systems, with demand projected at 695,000 tons by 2030 (16% CAGR from 2025). - Rare metals like tungsten, tin, and gallium also play indispensable roles in chip manufacturing.
**Construction: Building AI’s Physical Foundation** Engineering and construction (E&C) account for 40% of non-IT data center costs—the largest segment after power systems. China’s national projects like "East Data West Computing" are driving demand for civil engineering, installation, and management services, encompassing buildings, grid connections, and fiber networks.
The report highlights how AI’s infrastructure boom extends beyond semiconductors and software, creating a vast ecosystem of power, industrial, and materials firms as foundational players. Market leaders in these sectors stand to benefit significantly.