Skip to main content

Jun 01, 2026

Meeting Japan's power demand: the case for a new energy approach

Next year, if it all goes to plan, a floating data centre may be visible in Tokyo harbour.

With space at a premium in downtown Tokyo and data centres needing to be close to major transactional hubs to cut latency, expanding into the sea has been explored as a viable option.

This site of a 73 MW, 120-meter water-based data centre would say a lot, not just about Tokyo’s pressured real estate market, but also about its expanding need for compute infrastructure to power the new generation of Artificial Intelligence (AI), visualisation and other high-performance applications.

Increasing demand

Japan’s data centre capacity is expanding fast, with IT power capacity forecast to grow from around 2.3 GW in 2025 to approximately 3.7 GW by 20301, reflecting sustained demand growth driven by cloud and AI. Forecasts by energy analysts Wood Mackenzie go further, projecting that data centre energy need and peak power demand will continue to grow beyond 2030, with peak loads reaching approximately 6.6 to 7.7 GW by 20342. These numbers are consistent with experiences around the world, with the World Economic Forum saying that data centres could account for up to 12% of total US electricity consumption by 20303.

This level of demand begins to translate into very large increments of generation capacity, providing a clearer sense of the scale of infrastructure required to support it. One way to contextualise this is to compare it with the output of large-scale generation assets such as nuclear. Given that a nuclear power station, such as Sizewell C in the UK, has an output of 3.2 GW4, meeting the projected increase in power demand in Japan would require two new nuclear reactors to be built in the next decade just to power data centres5. That is a daunting task for any country, but especially challenging in Japan, where nuclear build has been paused since the 2011 Fukushima disaster and has only just restarted its nuclear power programme6.

Japan does not have an abundance of fossil fuels, so meeting this demand-supply delta through conventional power generation could expose the country to oil and gas market volatility and price fluctuations resulting from geopolitical instability.

Time to power

Alongside the scale of capacity required and the exposure to external energy markets, there is another significant challenge to meeting rising power demand: time. Conventional grid and existing alternative microgrid systems can encounter lengthy interconnection queues and suffer transmission delays, with connection to power grid capacity in Japan taking from 5 to 10 years7 in some cases. Given that new data centre developments are being built on 12 – 36 month timelines8, there is a clear need for alternatives that can power these data centres in the required timeframes while also meeting the massive High-Performance Computing (HPC) and AI compute demand at pace.

Japan is not unique in facing this problem. According to the International Energy Agency (IEA), global data centre power use is expected to roughly double by 2030, reaching around 945 TWh or 3% of global demand9, and grid connection delays are evident in many countries, with minimum delays ranging from 3 years in Italy10 up to 10 years in the United Kingdom11.

Overcoming the challenges

Japan’s latest national energy strategy offers an early glimpse of the future and how the rest of the world might also look to solve its energy crisis. Facing surging data centre loads, high renewable costs and uncertainty about nuclear restarts, Japan has indicated a renewed emphasis on long-term liquefied natural gas (LNG) contracts, is adding new gas fired capacity and has described LNG as a “realistic transition fuel” on the path to 2050 net zero12.

The problem is that this raises questions about alignment with Japan’s stated decarbonisation policy. The Japanese government has formally committed to achieving “net-zero” greenhouse gas emissions by 2050 across its entire economy, which inherently includes decarbonising its grid. This is underpinned by interim targets to reduce greenhouse gas emissions by 46% by 2030, 60% by 2035 and 73% by 204013 (relative to 2013 levels).

Aligning approach and policy may not be an easy task, but part of achieving it is to consider alternative technologies such as solid oxide fuel cells (SOFCs). These offer fuel flexibility, using natural gas or biomethane today, with the potential to transition to hydrogen tomorrow as low-cost green hydrogen becomes available at scale. This technology can also generate power more cleanly; SOFC systems can reduce CO₂ emissions per MWh versus many conventional combustion-based distributed generation solutions and produce minimal NOₓ, SOₓ and particulates. SOFC systems are also able to produce a highly concentrated CO2 stream that can support efficient carbon capture and sequestration.

While SOFCs have been proven across a range of industrial and distributed energy applications, scaling deployment to meet hyperscale data centre demand depends not only on the core technology, but also on the surrounding ecosystem. This includes local manufacturing capability, established supply chains, and long-term service and maintenance infrastructure, all of which are critical to reliable, continuous operation at scale.

Within this context, data centres represent a particularly relevant application for SOFC systems. Data centres require high availability power, often targeting ‘five nines’ (99.999%) or better uptime and are extremely power dense, with campus level loads in the tens to hundreds of megawatts. This imposes tight requirements on power quality, fault ride through and latency. SOFCs deliver premium quality energy as an uninterruptible power supply and their modularity means that they can scale to meet the demands of the largest data centres while providing rapid ramp-up to deliver the on-demand capacity that cloud services must provide.

Where local network capacity is constrained, substations and distribution networks can become overwhelmed; SOFCs can help reduce reliance on such infrastructure. As independent, modular and self-contained systems, they can be deployed without extensive grid reinforcement, while helping to address local concerns about air quality, noise and the use of diesel back-up generation.

Deployment in industry

Japan’s government is supporting the deployment of technologies such as SOFCs through a new programme launched this month14 to fund development and demonstration projects for low-carbon technologies to power data centres.

Ceres’ SOFC offering provides companies access to market opportunities such as this in Japan, and globally, through its licensing model. By accessing the Ceres EnduraTM stack platform, which uses a highly differentiated steel-supported SOFC cell technology, global manufacturers can gain a 10-year development advantage compared to starting from scratch.

Ceres EnduraTM is designed for rapid and low-cost manufacturing, delivering robustness and reliability for real world deployment and production at scale. Lower operating temperatures than other SOFCs unlock cost-efficient, widely available materials and enhance robustness for both manufacturing and deployment. Ceres’ technology enables rapid adjustment of power output, with a ramp up rate of 4.6% per second at stack level, that supports responsive load following as computing demand fluctuates in real time. 

The technology design in parallel with Ceres’ licensing model has attracted global manufacturers such as Doosan Fuel Cell in South Korea, Delta Electronics in Taiwan and Weichai Power in China, who are developing manufacturing sites and ramping up production. As production capacity increases, so too will economies of scale.

Taken together, these factors suggest that Ceres EnduraTM could play a role in helping meet Japan’s growing power demand on advantageous timescales, while supporting energy security and emissions reduction objectives.

Let's talk