Overview of the AIDC Industry Chain

The Artificial Intelligence Data Center (AIDC) industry chain is entering a phase of rapid expansion, driven by the escalating computational demands of AI models. According to a recent research report by CITIC Construction Investment, the convergence of advanced chip iterations and power architecture upgrades is catalyzing a new investment cycle. The widespread penetration of 800V DC power supply systems in data centers is set to unlock massive billion-dollar market opportunities, fundamentally reshaping the power equipment landscape for AI infrastructure.

Technical Details: Vera Rubin and 800VDC Architecture

The core technical catalyst for the next AIDC power cycle is the transition to Nvidia's Vera Rubin architecture. Following the Blackwell generation, which introduced the 200kW-class NVL72 full-rack delivery and 33kW Power Shelves, the upcoming Vera Rubin VR200 and VR300 (Rubin Ultra) chips will utilize the advanced Kyber rack architecture (NVL 144). This evolution mandates a shift towards 800V DC power distribution. High-power Power Shelves and external HVDC Sidecars will become the mainstream deployment models. The 800VDC penetration is the primary driver pushing power value from in-rack Power Shelves to external Sidecars, significantly elevating the overall power density and value per rack.

Market Context: Surging Power Capacity Demand

The exponential growth of AIDC construction has triggered unprecedented electricity capacity requirements. CITIC Construction Investment estimates that the power capacity demand driven by AI in the US will experience a Compound Annual Growth Rate (CAGR) of approximately 55% from 2025 to 2028. Over the next three years, the cumulative demand is projected to exceed 150GW. North America is currently facing a prominent power shortage, exacerbated by the steady or accelerated retirement of thermal power units and grid-connection delays for new capacity. Consequently, on-site power generation has become a prevailing trend among AI giants like xAI, Google, and Meta.

Industry Impact: The Gas Turbine Supply Chain

To address the severe power deficit, gas turbines have emerged as the priority solution for AIDC primary power due to their rapid response, high power adaptability, lower generation costs, and high reliability. However, the global gas turbine market is experiencing a significant supply-demand mismatch. Leading manufacturers have order backlogs far exceeding their current production capacities. Heavy-duty gas turbine delivery cycles have extended to over five years. This bottleneck presents substantial opportunities for domestic component manufacturers specializing in blades, castings, and heat recovery steam generators (HRSG). Furthermore, the supply gap is accelerating alternative solutions, including aero-derivative gas turbines and marine diesel engine conversions.

Implications for AI Data Centers: Power Equipment Opportunities

The transition to higher power rack architectures fundamentally alters the AIDC power equipment market. As 800VDC supply methods gradually penetrate the market, the external Sidecar market is projected to reach a billion-yuan scale by 2028. Mainland Chinese enterprises, which did not fully capture the previous Blackwell cycle dominated by Taiwanese firms, are poised to seize higher market shares in the Rubin and 800VDC cycles. Companies are already locking in materials and ramping up production for high-power in-rack power supplies, 800V external power supplies, and liquid cooling innovations, positioning themselves to capitalize on the surging Average Selling Price (ASP) and shipment volumes.

Future Outlook: 2027-2028 Growth Trajectory

The period from 2027 to 2028 represents the critical window for concentrated volume growth in AI external power supplies. With Vera Rubin entering mass production ramp-up in the second half of 2026 and scaling in the first half of 2027, 800V Sidecars will enter a rapid growth trajectory. The Sidecar market space is estimated to expand from 850 million yuan in 2026 to 103.5 billion yuan in 2028. However, stakeholders must remain vigilant regarding potential risks, including:

  • Fluctuations in cloud providers' capital expenditures and AI terminal demand.
  • Upstream supply chain constraints in power semiconductors, storage, and metals.
  • Geopolitical shifts, tariff changes, and stringent energy approval policies.