Fuse brings AI support in-house on NVIDIA hardware

Fuse Energy is bringing its AI compute in-house on NVIDIA infrastructure. The shift begins with its customer-facing AI and marks a broader collaboration between the two companies.

The UK energy supplier is moving its customer support AI from closed-source systems to open-weight models running on NVIDIA hardware. This is the first live use case in a wider plan to run current and future AI workloads on NVIDIA systems.

Since launch, Fuse has applied an in-house operating model across its energy business, covering generation, trading and retail supply rather than relying on outside providers for parts of that chain. It is now extending that approach to AI infrastructure.

Customer support AI

Its customer AI currently supports users through the company's app. Fuse says the system delivers 99% customer satisfaction and an 80% self-containment rate, meaning four in five queries are resolved without being passed to a human agent.

Fuse expects the move to open-weight models and in-house compute to maintain that level of performance. The customer service system is designed to provide round-the-clock support through the app rather than through traditional call centre processes.

The decision adds a technology layer to a business that has sought to differentiate itself in Britain's energy market through vertical integration. Fuse says its structure, spanning solar and wind generation through to trading and direct supply, has helped it deliver efficiency gains of about 17% compared with legacy providers.

That model has been central to the company's argument that tighter control over operations can help lower costs for customers, particularly during periods of wholesale market volatility. The latest move suggests Fuse sees a similar case for direct control over the computing resources used in customer service and wider internal systems.

Business expansion

In June, the company reported group-level profitability three years after launching publicly. It now supplies energy to more than 450,000 UK households.

Fuse was founded in 2022 by Alan Chang and Charles Orr, both former Revolut executives. The company says it also owns and operates low-carbon generation assets, with 22 MW active and more than 1 GW in development across the UK, the US and Europe.

The NVIDIA agreement also points to broader AI uses beyond customer support. NVIDIA says the work could extend to operational decisions tied to electricity demand, renewable output and market volatility, areas where energy retailers and generators often depend on large data sets and rapid forecasting.

For Fuse, the immediate step is narrower: moving an existing customer-facing tool from one model architecture to another while changing the underlying compute setup. But the tie-up indicates the company wants to build a longer-term AI base inside the business rather than depend on external model providers.

AI infrastructure

That reflects a wider shift among companies seeking more direct control over AI costs, data handling and deployment choices as usage grows. Open-weight models can offer more flexibility than proprietary systems, although they also require companies to take on more responsibility for tuning, hosting and operating them.

Competition in the UK energy retail market has pushed suppliers to look for ways to cut service costs while improving response times. Digital-first challengers have used apps and automated support to distinguish themselves from established players that rely more heavily on telephone support and outsourced systems.

Fuse says its app-first design has enabled 24/7 customer support without hold queues or call transfers. That approach has become an important part of its identity as it expands from a start-up into a larger supplier with hundreds of thousands of household accounts.

Speaking about the decision to shift AI systems in-house, Alan Chang, Chief Executive Officer and Co-Founder of Fuse Energy, said the move was consistent with the way the company had built its wider business.

"We built Fuse on one simple principle: own the stack, control the cost and deliver cheaper, abundant energy for everyone. Now we're betting our AI future on that same principle," said Chang.

"Working with NVIDIA gives us the flexibility to run AI workloads on our own terms, on infrastructure we control, as we continue to scale at pace. We're tackling the problems incumbents have failed to solve for decades and directly translating that into better value for our customers," added Chang.

NVIDIA framed the partnership as a combination of energy sector expertise and AI infrastructure.

"Fuse has built its business around controlling costs and delivering value across the entire energy chain, from source to socket. By combining that deep energy expertise with NVIDIA's accelerated computing and open-weight models, Fuse is building intelligence into the foundations of its operations - from customer-facing agents to complex decisions around demand, renewable output and market volatility," said Santhosh Gopinath, Managing Director, Energy EMEA, NVIDIA.