Microsoft has abandoned plans to use an artificial-intelligence data center being developed by Nscale in Loughton, Essex, after the local electricity network operator warned that adequate grid capacity might not be available until the early-to-mid 2030s.
The project had been presented as a major example of the shift by bitcoin-mining companies into AI infrastructure. It was supposed to provide 50 megawatts of computing capacity for Microsoft’s Azure cloud by early 2027, using 23,040 Nvidia chips. Instead, the site’s power connection has become the central obstacle, leaving Nscale to consider alternatives after clearing the location.
The development, reported by CoinDesk citing reporting from the Financial Times and the Guardian, shows how the expansion of AI computing is running into a basic infrastructure constraint: data centers need large amounts of reliable electricity, but grid connections can take years to secure.
A 2027 target collides with a 2030s grid timeline
The planned facility was located in Loughton, Essex, roughly six miles northeast of London. Nscale had intended to deliver the site for Microsoft’s Azure cloud with a target of operational computing capacity by early 2027.
UK Power Networks, the local grid operator, told Nscale that sufficient capacity might not be available until the early-to-mid 2030s. That timeline was incompatible with the project’s planned delivery schedule, prompting Microsoft to walk away from the arrangement.
Nscale has since cleared the site, which was still being used as a working scaffolding yard when the Guardian visited in February. The company is examining other ways to secure electricity for the project. It said its other contracts with Microsoft remain in place.
The project was announced last year by Nscale alongside Microsoft, Nvidia and OpenAI. Nvidia’s role was central to the proposed buildout: the facility was expected to house 23,040 of the chipmaker’s AI processors.
Why bitcoin miners are moving into AI
Nscale grew out of Arkon Energy, a company that developed facilities for bitcoin mining. Its founder, Josh Payne, applied experience in securing land and electricity to a business focused on supplying computing capacity to AI companies.
That strategy has spread across the publicly listed bitcoin-mining industry. Miners already operate power-intensive facilities and have experience developing large computing sites, making AI and high-performance computing an appealing area for expansion as the economics of bitcoin mining become more demanding.
The transition, however, requires much greater investment. Estimates cited in the source material put the cost of an AI-ready facility at $8 million to $15 million per megawatt, compared with approximately $700,000 to $1 million per megawatt for bitcoin mining. AI customers also expect servers to operate continuously.
The industry’s shift has involved both new contracts and asset sales. When CoinDesk reviewed the trend in March, public miners had announced more than $70 billion in AI and high-performance computing contracts. Those companies had also reduced their combined bitcoin holdings by more than 15,000 BTC from their peaks, partly by selling bitcoin to help finance construction and expansion.
Contracts are large, but operating capacity remains limited
The financial scale of announced AI deals does not yet match the amount of capacity that is generating revenue. In its latest mining report, fund manager CoinShares counted more than 4 gigawatts of AI capacity under contract among the miners it tracks. Only about 550 megawatts was actually billing customers.
The same report put contract backlogs above $100 billion, while annualized AI revenue was just over $1 billion. The figures illustrate the gap between signing agreements and building facilities that can be connected, equipped and operated.
Bitcoin mining has also been under pressure. The source material put the average listed miner’s cash cost of producing one bitcoin near $80,000 in the fourth quarter of 2025, while bitcoin was trading around $70,000 by March. AI contracts can provide recurring payments over several years, but the facilities needed to serve those customers are substantially more expensive and take time to develop.
Britain’s crowded connection queue
Nscale’s setback is part of a wider U.K. grid problem. Data center projects have requested connections representing about 73 gigawatts of demand, according to energy regulator Ofgem.
Ofgem is considering stricter rules for the queue because some applicants may never build their proposed projects. Those applications can nevertheless delay developments that are ready to proceed, adding to the uncertainty for companies seeking power-intensive sites.
For Nscale, the immediate question is whether it can secure electricity independently or persuade the grid operator to accelerate the connection. Without either outcome, the cleared Loughton site could remain unused for years, despite the original plan to provide Microsoft with AI capacity in 2027.
Moneyiar market picture
Moneyiar market data at 22:03-22:04 on Oct. 19, 2026, showed bitcoin at $81,990.46, up 0.57% over 24 hours but down 0.68% over seven days. Cronos was $0.05719, up 2.31% in 24 hours and down 4.27% over seven days.
Ethereum Classic stood at $7.7342, gaining 1.77% over 24 hours while falling 5.57% over seven days. Compound traded at $21.5138, up 3.05% over 24 hours and down 8.83% over seven days.
