HiCloud / Highlander (Hailanyun)

Status: Alive — the only operator anywhere with underwater datacenters in commercial operation. Not a graveyard entry; the counter-example the rest of the graveyard is judged against. The catch: "commercial" is heavily state-scaffolded.

Who they are

Beijing Highlander Digital Technology (Shenzhen-listed, founded 2001) is a maritime electronics company, not a datacenter operator by origin. The underwater business runs through subsidiaries Shenzhen HiCloud and Shanghai Hailanyun (海兰云), with CNOOC's offshore-engineering arm as build partner. Notably, they incorporate tech from their acquired Canadian subsidiary OceanWorks despite subsidiaries sitting on the US Entity List.

Deployment 1 — Hainan (Lingshui, "first commercial UDC")

First 1,300-ton cabin submerged Dec 2022 at ~35 m, launched with customers Mar 2023; second cabin (400+ servers, pitched at AI inference) added Feb 2025. Execution reality: 2 cabins, ~1,200 servers total, three years into a plan for 100 cabins under an RMB 5.6B (~$880M) framework with the Sanya government. Named customers: China Telecom, SenseTime, Tencent (early reports), "about ten companies" by 2025. Workloads skew mundane and latency-tolerant — inference, annotation, consumer app backends. Light Reading's read: effectively a Hainan Free Trade Port demonstration project, not operator-funded colo.

Deployment 2 — Shanghai Lingang (24 MW, offshore-wind-powered)

The flagship. Launched Jun 2025, built by Oct 2025, full operation May 2026 — commercial rollout in under 30 months, which one UC Davis researcher calls "something Natick never attempted." ~10 m depth, 6–10 km offshore, direct connection to an adjacent 50+ turbine wind farm (97% wind-powered, bypassing the grid). 198 racks, ~2,000 servers. Claimed PUE ~1.15, ≥30% less electricity than land per a CAICT assessment, zero freshwater. Financing: CNY 1.6B (~$222M), state/SOE-heavy (Lingang Special Area committee, state investment holding group, CCCC). Tenants: China Telecom (GPU clusters) and LinkWise.

Pipeline and export talk

Jun 2025: cooperation agreement targeting a 500 MW underwater deployment (Shenergy, Shanghai Telecom, INESA, CCCC) — no location or timeline. Export ambitions (Japan, HK, Singapore, Korea) are talk; no signed overseas deal or shipment as of Jul 2026. Company claims an 80% cost reduction from shallow-water (<50 m) redesign vs. Natick-style pressure vessels.

The skeptical read

• Every named tenant is a state carrier or Chinese AI firm. No disclosed pricing, occupancy, contract values, or segment revenue — ever.
• Both flagship sites sit in free-trade/special zones with government co-investment; MERICS traces the driver to state energy-efficiency mandates and AI-infrastructure policy, not market demand.
• Scale is small: 198 racks vs. 2,000–10,000+ for typical land facilities; Hainan is at 2% of its announced cabin count.
• Unanswered engineering questions (per CleanTechnica): failed-server replacement philosophy, cable landing, marine-risk ownership. The "zero failures" claim is unverified.
• Environmental: warm low-oxygen outlet water during marine heat waves; the Hainan site was deliberately sited in an upwelling zone (<24.5°C year-round) — i.e., the model depends on favorable oceanography.

What it means for the segment

This is the only place on Earth to test whether anyone pays for underwater compute — and the honest answer so far is "the Chinese state pays, with real workloads running." The engineering achievement is real and fast. The demand signal is not yet market-priced. If the 500 MW project gets a site and a date, or a single export order lands, this stops being a demo and starts being an industry.

Sources

offshorewind.biz: full operation May 2026 · Scientific American analysis · Tom's Hardware specs · MERICS policy read · DCD: exports ambition · DCD: 500 MW agreement · Light Reading skeptical take · CleanTechnica critique · SCMP: Hainan launch