Can AI workloads follow renewable energy?

UK technology company Easy Compute is deploying AI computing hardware alongside renewable energy generation on farms, using electricity produced from agricultural waste to support AI workloads.

The model, operated through the company’s Green Compute network, places high-performance computing equipment close to local renewable generation rather than relying solely on power exported to the grid.

Initial deployments include UK farms using biogas, including a site in the North West that converts pig slurry into electricity.

The approach raises a wider infrastructure question: whether some AI workloads could be located closer to renewable energy sources, rather than requiring all power to be delivered into large centralised data centre campuses.

Bringing Compute Closer to Power

Most AI infrastructure investment continues to focus on hyperscale data centres, where operators can concentrate power, cooling, networking, and security at scale.

Easy Compute’s model takes a different approach. Instead of exporting surplus renewable electricity to the grid, farms can use that energy to run AI computing hardware on site through the Green Compute network.

According to the company, electricity exported to the grid typically generates around 8-12p per kWh. It claims that using the same electricity to power AI workloads can generate up to ten times more value.

Those figures are company estimates and have not been independently verified. However, they point to the commercial logic behind the model: connecting compute demand directly to renewable generation where that energy already exists.

Changing the Economics of Renewable Generation

Easy Compute has argues that its approach could improve the payback period for on-farm renewable infrastructure.

The company has claimed that revenue from AI computing could reduce the expected payback period for anaerobic digestion and associated equipment from around 12-15 years to approximately four years.

It has also been reported that some larger participating farms are generating tens of thousands of pounds per month from AI computing, although those figures remain company estimates.

The broader point is that AI demand is creating new ways to monetise renewable electricity. For farms and other distributed energy producers, compute workloads could provide an additional revenue stream where grid export alone does not fully capture the value of locally generated power.

A Different Deployment Model

The Green Compute network combines several commercial mechanisms.

Businesses can rent AI computing capacity on a pay-as-you-go basis. When commercial demand is lower, the company says computing resources can also contribute processing power to decentralised AI networks such as Bittensor, allowing participating farms to earn cryptocurrency rewards alongside conventional compute income.

Customers paying through the company’s Green Compute Coin receive a claimed 10% discount on compute credits.

These mechanisms are specific to Easy Compute’s platform, but they illustrate how new commercial models are emerging around AI infrastructure. Rather than treating renewable generation and computing demand as separate systems, the model links them more directly.

Verification Will Matter

Easy Compute has said that every participating site must verify its renewable energy before earning through the Green Compute network.

That requirement is important because as AI infrastructure faces closer scrutiny over energy use, claims about renewable-powered computing will need to be transparent, measurable, and credible.

Questions around verification, workload suitability, hardware lifecycle impact, and operational resilience are likely to become more important as distributed models develop.

The model may not be suitable for every AI workload.

Hyperscale data centres will continue to play the central role in supporting large-scale AI training, inference, and cloud services. They offer the density, networking, cooling, and operational control required for many advanced workloads.

Distributed compute may instead be more relevant where workloads can tolerate different latency, availability or orchestration requirements, and where renewable generation is already available.

Another Geography for AI Infrastructure

Easy Compute’s model does not replace hyperscale data centres. It does, however, provide an example of how AI demand is influencing where compute can be located.

Much of the current infrastructure discussion focuses on bringing power to large data centre campuses. Farm-based AI computing reverses part of that logic by placing compute closer to distributed renewable generation.

That positions the model as an additional deployment approach within a broader AI infrastructure landscape, rather than as a direct competitor to hyperscale infrastructure.

As AI demand continues to grow, the geography of compute may become more varied, shaped by data centre campuses and grid connections, and where usable energy is already being generated.