Dr Beenish Ayaz (PhD, SMIEEE, CEng, FHEA)
Department of Electronic Engineering
Royal Holloway University of London

Artificial intelligence is often perceived as intangible — algorithms in the cloud and intelligence embedded in software. In practice, AI demands an immense physical backbone of so-called “AI warehouses”: data centres that consume growing amounts of electricity and water. In the UK, data centre demand is rising sharply alongside ambitions in AI, digital resilience, and high-performance computing. This creates a significant engineering and societal challenge: how can digital progress be aligned with environmental sustainability and social responsibility?

One emerging approach is the deployment of underwater data centres, cooled by the ocean’s naturally low and stable temperatures. Projects such as Microsoft’s Project Natick demonstrated that subsea cooling can significantly reduce the overhead energy associated with thermal management. However, cooling efficiency does not reduce the energy required for computation itself. True progress toward net-zero therefore depends not only on innovative cooling, but also on integration with renewable energy systems, lifecycle-aware design, and responsible operational practices.

Although Microsoft concluded its experimental programme in 2024, the concept has evolved from research prototype to strategic infrastructure. Several countries are now exploring commercial-scale deployments. China, for example, has established some of the earliest large-scale operational subsea data centres near Hainan and Shanghai, reporting substantial reductions in cooling energy demand. These developments illustrate technical feasibility, but they also highlight the need for shared standards for environmental protection and governance.

Relocating digital infrastructure offshore does not eliminate impact—it redistributes it. Subsea deployments interact with fragile marine ecosystems and with coastal communities that may have limited voice in how digital infrastructure is governed. Potential thermal plumes and broader ecological change raise the risk of an “out of sight, out of justice” scenario. At the same time, strategic competition around AI introduces questions of digital sovereignty and control over the physical foundations of the cloud. For the UK, with its extensive coastline and marine research capability, this represents both an opportunity and a responsibility. Here, technology can support trust-building, and intelligent underwater wireless sensor networks (UWSNs) become vital. Acting as a “nervous system”, they can monitor temperature, acoustics, and water chemistry in real-time. Yet technical capability alone is insufficient. Underwater networks face constraints such as limited bandwidth, power scarcity, biofouling, and harsh propagation environments. More fundamentally, they raise governance questions concerning data ownership, interpretation, and distribution of benefits. When monitoring data are confined within proprietary systems, transparency is diminished and public trust is undermined.

(Image created using Google Gemini)

A collaborative, solution-oriented approach aligned with responsible research and innovation is therefore essential. Open standards, interoperable platforms, and shared data frameworks can allow regulators, scientists, industry, and citizens to jointly assess environmental impact. In this way, underwater sensing becomes not only a performance tool, but a mechanism for accountability, inclusion, and societal benefit. Evidence from real-time marine sensing supports informed debate, policy engagement, and conservation-centred design, transforming technology from a closed system into a more inclusive public resource.

Ultimately, underwater data centres are not a simple route to net-zero computing. They offer a powerful test case for whether we can build AI infrastructure that is not only efficient, but also transparent, equitable, and environmentally responsible. The future of digital infrastructure is not merely computational. It is socioenvironmental and governance-driven. As engineers and researchers in the UK and beyond, we must design for bandwidth and biology, latency and justice, working collaboratively from the seafloor up.