Locai raises £1m pre-seed led by Fuel Ventures to move AI inference off the cloud
April 22, 2026
Locai, a London-based deep-tech startup, has raised £1 million in a pre-seed round led by Fuel Ventures to commercialise infrastructure that runs AI model inference on end-user devices rather than in the cloud.
The company is targeting a structural problem in how AI products are priced. Cloud-based inference charges per interaction, which means SaaS and desktop AI vendors face costs that scale linearly with user growth and route customer data through third-party servers. For products like meeting tools, writing platforms, code assistants and customer-support apps, that combination puts persistent pressure on margins and limits what kinds of customers, particularly in regulated sectors, the product can serve.
Locai's platform is pitched as a drop-in replacement for cloud AI APIs. The company exposes an OpenAI-compatible endpoint that lets developers migrate existing applications onto on-device models with a single line change. Inference then runs on the user's own laptop, workstation or dedicated hardware, with prompts and data staying local. The technical premise rests on recent hardware advances that allow modern devices to run 7–13 billion parameter models locally, an envelope wide enough to cover a long tail of enterprise use cases without round-tripping to the cloud.
The startup was founded by Joseph Ward and Saif Al-Ibadi, who previously built SmallSpark Space Systems, a deep-tech company applying generative design to defence and aerospace engineering, including work on the UK Ministry of Defence's first generatively designed rocket engine. Locai joined the initial cohort of the Google for Startups Accelerator in 2025 and has demonstrated the approach through a multi-year contract with B2Space, where it deployed AI agents at the edge of space and reportedly cut bandwidth costs by more than 90%.
The pre-seed funding will be used to scale go-to-market, with a focus on SaaS and desktop AI product companies whose unit economics are most exposed to cloud inference bills. Mark Pearson, partner at Fuel Ventures, said the firm was backing the team for its track record deploying AI in resource-constrained environments and the broader pull towards sovereign AI infrastructure as enterprise demand for on-premise and on-device options grows.
Sources
Thank you to our Patrons & Partners:
