Underwater Data Center in Shanghai: AI Infrastructure Goes Underwater, This Time by Design

Key Takeaways

• 32‑meter steel cylinder underwater, powered by offshore wind >95% renewable
• Investment ≈ $224 million for 24 MW capacity, first phase 2.3 MW
• Claimed PUE ≤1.15, 22.8% lower power use vs land data centers
• Expected 90% land savings, zero water use, but corrosion risks remain

Pulse Analysis

The underwater data center in Shanghai reflects a broader shift toward integrating renewable energy directly with compute infrastructure. By situating servers beneath the sea, operators exploit the ocean’s stable 15 °C temperature, dramatically reducing the need for traditional chillers and cooling towers. Coupled with a dedicated offshore wind link that supplies more than 95% of the power, the design pushes the industry’s PUE benchmark toward the theoretical ideal of 1.0, promising substantial operational cost savings and a smaller carbon footprint.

Beyond environmental gains, the project addresses a strategic bottleneck for AI workloads: reliable, high‑density power in space‑constrained megacities. The 24 MW target capacity, with an initial 2.3 MW phase, is earmarked for GPU‑heavy tasks such as large‑language‑model training and massive data annotation. By decoupling compute from land‑bound grids, the Ling‑gang hub could attract cloud providers seeking to expand AI services without competing for scarce urban real estate, thereby accelerating China’s domestic AI ecosystem.

However, the maritime model introduces new operational complexities. Saltwater corrosion, underwater cable maintenance, and limited physical access raise questions about long‑term reliability and total cost of ownership. Ecological impacts on marine habitats also demand rigorous monitoring. If these challenges can be managed, the Shanghai facility may become a blueprint for repeatable offshore data centers, prompting other coastal economies to explore similar green‑compute corridors as the demand for AI processing continues to surge.