Deep-sea aquaculture platforms — China moves fish farming offshore and instruments it
East-Asia (China) — Fujian, Guangdong, Shandong, Shenzhen
Content
Chinese aquaculture’s structural move of the 2020s is offshore: cage farming pushed into deeper water on instrumented platforms, with the state as programme framer and provincial operators as builders.
The national fleet, per MARA’s 十四五 progress summary (October 2025). Gravity cages 30,000+; truss-type cages and large enclosures 89; farming vessels 6. MARA describes the resulting structure explicitly — gravity cages as the technically mature and economically applicable mainstay, truss cages as auxiliary, and farming vessels as exploratory. Guangdong, the largest provincial builder, has put nine large truss platforms into the water, including 恒燚一号 (Hengyi-1, the province’s largest) and 明渔一号 (Mingyu-1, built with wind-fishery integrated smart equipment). One platform is reported to have survived a Force-14 typhoon within two months of completion — the operational argument for offshore structures stated as weather performance.
The instrumented case: Lianjiang county, Fujian. The National Data Administration’s “Hundred Platforms, Million Tonnes” (百台万吨) case study documents how a county-level programme instruments its platforms:
- Sensing: sonar, underwater robots, digital twins and IoT, monitoring water temperature, light, residual chlorine, pH, dissolved oxygen, turbidity, salinity, ammonia nitrogen and chlorophyll, plus physical parameters — currents, waves, tide level, wind speed and direction, pressure.
- Connectivity: China Mobile 700 MHz and China Unicom 900 MHz low-band for full nearshore 5G coverage, with 10 Gbps microwave backhaul and a 700 MHz + 2.6 GHz hybrid network; floating-platform 5G base stations.
- Platform layer: an “attended-free” operation model where data reaches the farmer’s phone; stress monitoring of mooring systems and platform structure feeding maintenance and future platform design.
- Data assets: inventory, quality evaluation, registration, rights confirmation and valuation of marine-fishery data, to convert data resources into legally effective, compliant data assets.
Reported outcomes (the case’s own figures). Around 70% labour cost saved; dissolved oxygen 30% higher in deep water with claimed improvements in fish survival and quality; 11 platforms deployed, one quarter of China’s truss-type platforms, nearly 180,000 m³ of farming volume, about 1,800 tonnes annual output, over CNY 200 million in annual output value; better feed and faeces dispersal reducing water-quality impact; and a downstream argument that mechanised, standardised output supports a unified trading centre and a seafood price index that dampens seasonality.
The industrial scale of the vessel route. Shenzhen’s first modern marine-ranch projects include four 100,000-tonne-class large intelligent farming vessels with total investment of about CNY 2.29 billion, targeting premium marine species (coral trout, hybrid grouper, golden pomfret, greater amberjack). Elsewhere: 海威2号 (Haiwei-2) with unattended operation and real-time remote monitoring, using a ballast system to raise and lower the platform for mooring at 15 m to deeper water; 格盛一号 (Gesheng-1) at 86 m × 34 m with 30,000 m³ of water and a full digital production system.
What this unit is doing in the taxonomy
The corpus’s second aquaculture unit and its first on the physical production layer in any country at this scale. It documents a jurisdiction building aquaculture infrastructure offshore with instrumentation designed in, rather than retrofitting sensors to existing ponds — the structural inverse of most aquaculture-AI stories in the corpus.
Distinct from:
china-smart-fisheries-fanli-llm.md— the model and dataset layer; this is steel, moorings and telemetry.canadian-aquaculture-ai.md— Canada’s commercial, mostly inshore and salmon-farm-side systems; China’s is a state-programmed offshore build-out.china-smart-agriculture-action-plan-2024-2028.md— names 智慧渔场 equipment (water-quality monitoring, auto aeration, intelligent feeding, fish-disease diagnosis, grading and counting); this unit is the deep-water realisation of that list.norway-salmon-ai-type material — Norway’s offshore aquaculture is commercially driven, whereas China’s is a provincial state-investment programme tied to marine-ranch policy.
Why it matters for talks
- Fleet composition is the honest headline: 89 truss platforms and 6 farming vessels against 30,000+ gravity cages. The exotic structures are real but marginal in count; the cage is still Chinese aquaculture.
- ~70% labour-cost reduction from instrumentation is the strongest labour figure in the corpus’s aquaculture material, and it comes from an official case study rather than a vendor.
- Design-for-typhoon is the adoption driver. Offshore structures are justified by survival in storms the nearshore cannot take — climate adaptation as the commercial case, not sustainability language.
- Data becomes a registered asset. The Fujian programme includes data-asset inventory, valuation and rights confirmation, which is the corpus’s first case of aquaculture telemetry being formally capitalised in a state system. Worth contrasting with the farmer-data-ownership debates elsewhere.
- Deep water is claimed to improve the product — +30% dissolved oxygen, better survival and quality. Treat as a claim; it is the kind of environmental argument that would be auditable and has not been.
Critical context
- Output is small relative to the fanfare. Eleven platforms producing about 1,800 tonnes a year and CNY 200 million is a rounding error against China’s aquaculture production. The programme is about capability and structure, not yet volume.
- The 70% labour saving has no baseline disclosure — what labour, at what scale, over what period. Single-figure efficiency claims in Chinese case studies typically compare a fully instrumented platform with a labour-intensive cage operation.
- Ecological claims are asserted, not measured. Reduced water-quality impact, better feed dispersal and higher dissolved oxygen are described in the case study without sampling data or independent monitoring.
- Capital intensity is extraordinary per unit of output. Four vessels at CNY 2.29 bn versus 1,800 tonnes of annual production across a county’s platforms — the economics rest on premium species prices and provincial investment, not on returns.
- “Unmanned” and “attended-free” are marketing-adjacent terms. The case describes remote monitoring and data-to-phone; it does not describe a platform operating without human intervention for a season.
- The NDA case study is a promotional genre. It is a government case library entry, selected for demonstration value, and it presents its own figures without external verification. V1 reflects a government primary source with named numbers, not an audit.