Underwater Data Centers

The ask

Design and deploy sealed, modular data centre pods submerged in Bangladesh's coastal waters or the lower Meghna estuary, exploiting passive seawater cooling to cut power usage effectiveness (PUE) to near 1.0 — and pitch the resulting green-compute premium to international hyperscalers seeking low-carbon hosting in South Asia.

Why now

Microsoft's Project Natick demonstrated that submerged data centres achieve PUE of 1.07 versus industry average 1.5–1.8, with dramatically lower failure rates (attributed to inert nitrogen atmosphere and absence of humidity cycling). AI inference workloads are creating a massive demand for compute that existing land-based data centre pipelines cannot satisfy fast enough; a novel, demonstrably lower-PUE format attracts premium contracts. Hyperscaler sustainability commitments — Microsoft, Google, and Meta have all pledged water-positive or carbon-negative operations — create a buyer for verified green-compute capacity.

Why Bangladesh

The Bay of Bengal coastline and Meghna delta offer year-round seawater temperatures of 20–30°C — warmer than ideal (North Sea deployments benefit from 8–12°C) but adequate for data centre cooling with a modest chilled-water loop. Bangladesh's emerging data centre policy and its position between India and Southeast Asia create a plausible argument for regional connectivity infrastructure. The country's shallow-water maritime expertise from offshore gas extraction provides a local talent base for deployment and maintenance. This is, however, a frontier technology with low Bangladesh-specific fit; the primary rationale is R&D and technology demonstration, not near-term commercial scale.

As a business

The near-term business is a technology demonstration and IP development vehicle: build one or two prototype pods, publish PUE and reliability data, and license the deployment methodology to larger infrastructure players in South and Southeast Asia. Revenue comes from R&D grants (BCSIR, ADB innovation funds), a pilot hosting contract with a domestic telco or cloud provider, and eventual licensing or joint-venture fees. This is a 7–10 year play to a trade sale or licensing deal, not a 3-year path to profitability.

Economics

Move the sliders to model your own underwater data centre venture. Defaults are order-of-magnitude estimates — pressure-testing them is part of what a founder pitches us.

What ZEPH would back

Honestly, we would not lead this round from Bangladesh — the technology risk and capital intensity exceed what makes sense for our geography and stage. We would engage as an advisor or angel if a founder with deep offshore-engineering credentials wanted to attempt a demonstration pod in the Bay of Bengal, and we would make an introduction to infrastructure DFIs. The idea earns its place in ClimateFair as an inspiration — but the founder needs to find a hyperscaler or government anchor before approaching us.

Impact

Submerged pods passively cooled by seawater achieve a PUE of ~1.07 versus Bangladesh's grid-powered land data centre average of ~1.6–1.8, cutting electricity per unit of compute by 30–40% — at Bangladesh's grid emission factor that equates to roughly 250–350 tCO₂e saved per pod per year. Eliminating freshwater evaporative cooling preserves a resource under acute stress in coastal Bangladesh. A green-compute premium from sustainability-committed hyperscalers generates hard-currency revenue for Bangladesh's digital economy while requiring minimal land use on an already land-scarce coastline.