Green Compute (SN110) Offers the Best Clean Inference on a Market That Runs on Cow Shit, and Is Now Powering Orion

Somewhere on 14 English farms, methane from anaerobic digesters is being burned into electricity that powers thousands of gaming GPUs training a 16-billion-parameter language model. This is not a metaphor!

IOTA (SN9) named Green Compute (SN110) as one of the suppliers behind Project Orion, and the physical setup is exactly what it sounds like: cow shit at the front of the pipeline, decentralized frontier AI training at the back.

A Different Kind of Compute Supplier

Green Compute doesn’t look like anyone else on the market, which is precisely why it fits IOTA’s architecture.

1. 14 small farm-based data centers: Scattered across working UK farms rather than concentrated in one facility.

2. Powered entirely by biogas: Anaerobic digesters convert agricultural waste into methane, which runs the generators, which power the racks.

3. Several thousand GPUs across three generations: A4000s, 4090s, and 5090s aggregated across the sites.

4. Independent operator, not a hyperscaler subsidiary: No procurement cycle, no corporate parent, no waiting list.

The UK grid cannot absorb all the electricity these farms generate, leaving roughly half of it unused on a typical day. Green Compute turns that stranded power into AI training capacity, replacing the traditional model of renting compute from centralized NVIDIA-backed data centers.

Why IOTA Needs This

Most training frameworks would treat a fourteen-site farm operation as a nightmare. iota was designed for exactly that.

1. Different GPU generations, one training run: A4000s and 5090s work alongside each other.

2. Inconsistent bandwidth handled at the protocol layer: No requirement for identical network conditions across nodes.

3. Geographic sprawl as a feature: Multiple sites contributing rather than one campus dominating.

4. No unified administrative plane: Independent operators keep their own operations while IOTA coordinates the output.

Project Orion’s training run combines compute from a globally distributed network of independent suppliers, each contributing available hardware.

Green Compute (SN110) supplies farm-powered RTX A4000 and RTX 5090 GPUs, while IOTA (SN9) unifies this heterogeneous infrastructure into a single training cluster.

Instead of relying on a hyperscaler-funded, uniform GPU fleet, IOTA trains frontier models by aggregating stranded compute from around the world.

The Wider Read

Training capacity is not inherently scarce; what is scarce is uniform, centralized compute controlled by hyperscalers. Project Orion demonstrates that frontier AI models can be trained by aggregating distributed compute instead of relying on a single data center.

Green Compute embodies that shift by turning methane-powered electricity from UK farms into GPU compute for a frontier machine learning pipeline.

➛ Read More on What Green Compute (SN110) Does Here:

How Green Compute (SN110) Is Turning Farmers Into Bittensor Miners