Luffy AI, an Abingdon startup building what it calls neuroplastic AI for the real-time control of physical machines, has raised £8.1m in a Series A round, the company said on Tuesday.

The round was led by BGF, which bills itself as the most active equity investor in the UK and Ireland, and the money is earmarked for turning pilots into commercial deployments.

Luffy is a spinout in the fullest sense. Its founders, Dr Matthew Carr and Dr Alex Meakins, are former nuclear physicists from the UK Atomic Energy Authority, and the company still sits on the Culham Campus near Oxford, a site better known for fusion research than for the kind of university deeptech spinouts now spilling out of it.

The pitch turns on a problem the wider AI boom has mostly stepped around. Large language and image models lean on vast data, cloud compute, and constant connectivity, none of which suits a pump or a conveyor belt on a factory floor, and it is a world away from the physical-AI bets that have drawn Britain's biggest cheques, such as Wayve's $1bn round.

Luffy's answer is a different kind of network. Its sparse neural networks are trained in simulation, sidestepping the scramble for large training sets that conventional deep learning demands, then refined against the real machine, an approach the company says can be up to 400 times more efficient than traditional deep learning.

The architecture is meant to be small enough to sit on the hardware itself. Because the models self-refine from live feedback rather than retraining from the cloud, Luffy says they can run on the edge and continuously tune themselves to whatever they are controlling.

The first target is the electric motor. Around half the world's electricity is consumed by electric motors, most of them running inefficiently, and Luffy is deploying its models into motor control and variable-frequency-drive applications such as industrial pumps, fans, and conveyors.

The commercial promise is plug-and-play. Adaptive control, the company says, lets a motor tune itself to its load and operating conditions in the field, cutting energy use, shortening commissioning time, and lifting performance without a specialist engineer on site.

“AI has been transformative for language and image generation, but has yet to make a substantial impact in industry beyond predictive maintenance and dashboards,” said Carr, Luffy's co-founder and chief executive. Factories and motors, he added, need AI that is “small, fast and adaptive in real time,” rather than cloud-dependent and hungry for data and compute.

Joining BGF were MIG Capital, the Munich deep-tech investor, through its MIG Fonds, alongside existing backers Bow Capital, Chrysalix, Momenta, and UKI2S. Kate Ronayne, an early-stage investor at BGF, said Luffy was “disrupting an industry norm that has stood for 100 years” by embedding specialised AI directly into physical systems, and reducing the reliance on specialist engineers to commission them.

For MIG, the appeal was efficiency as much as ambition. “Luffy does more with far less data and compute, which is precisely what makes AI workable inside physical machines,” said Dr Nicolas Rose-André, an investment manager at the firm, pointing to the scale of the electricity that motors consume as an opportunity in its own right.

Luffy did not disclose a valuation for the round or its revenue to date. The funding will push its proofs of concept and pilots toward partnerships with larger industrial brands, and further out the company sees the same control technology extending to robotics and drones, thermal process control, and other physical-AI uses.

It is a large ambition for a modest round, and one that rests on whether a self-tuning motor can prove itself on a real factory floor rather than in simulation.