Key takeaways

  • Latos Data Centres plans a 28,000 sq ft Manchester edge facility, part of a "neural Edge" strategy targeting 40 UK sites by 2030.
  • Home-based GPU pilots are already paying residents to host distributed AI compute, with speed, reliability, security and hybrid feasibility still under evaluation.
  • EdgeCore Networks is pitching an all-optical fabric to connect GPUs, servers and storage across sites, while arguing that each campus should fund its own power infrastructure.

Manchester's first AI-focused edge hub

Latos Data Centres has lodged a planning application with Manchester City Council for a two-building campus on a vacant brownfield parcel off Bower Street in Newton Heath. The build-out would occupy roughly 28,000 sq ft (2,601 sqm).

If the council grants permission, Latos expects the site to be operational by 2030. The developer frames the project as part of a broader "neural Edge" strategy, with 40 sites planned across Britain by 2030 so that no location sits farther than 50 miles (80.4 km) from a low-latency AI node.

The Manchester application follows a £100 million ($136m) planning approval Latos received last year for a data center near Middlesbrough, UK.

Home-based GPU networks: crowdsourced compute at the edge

A separate approach looks at whether households can host AI-accelerated workloads. Pilot programs are already paying homeowners to run GPUs in their homes, turning residential connections into a distributed compute layer. Coverage of these pilots examines their speed, reliability, security, and how they might work alongside traditional edge facilities — open questions rather than settled outcomes.

EdgeCore's power-first stance and all-optical fabric

Edgecore Networks says distributed AI infrastructure will become more important as data centers run into limits on power, space, and cooling. Speaking at the 2026 OCP APAC Summit, the company outlined an all-optical networking approach designed to connect GPUs, servers, and storage across multiple sites and regions.

The developer-operator also says it will fund the power infrastructure its own campuses need, though it acknowledges the precise definition of "full cost" for electricity provision remains unclear.

How the three strands intersect

  • Geography – Latos' plan to keep AI services within 50 miles of users mirrors the broader industry push toward distributed, latency-sensitive topologies — whether through purpose-built edge sites or home-based nodes.
  • Power – EdgeCore's position that operators should self-fund power infrastructure sits alongside the capital commitments already implied by Latos' UK build-out, including its earlier £100 million Middlesbrough approval.
  • Networking – An all-optical fabric, as EdgeCore describes it, is designed to link distributed compute across sites and regions — though no source ties EdgeCore's technology directly to the Manchester project or to the home-GPU pilots.

Conclusion Manchester's edge data-center proposal, the emerging home-GPU pilot model, and EdgeCore's power-first, all-optical pitch each address a different piece of the AI compute buildout — and remain, for now, separate initiatives at different stages of maturity. Latos still needs planning approval; the home-GPU pilots are still being evaluated for viability; and EdgeCore's fabric has not been publicly tied to a named deployment. Investors and engineers tracking distributed AI infrastructure should watch for the Manchester planning decision, pilot results from the home-GPU trials, and whether any of these separate threads end up interconnected in practice.

Sources

This article was researched and fact-checked against the following sources: