Nscale’s Microsoft AI Partnership Collapses as Electricity Bottlenecks Persist

Microsoft has withdrawn from a planned U.K. AI data center project operated by Nscale after warnings that electricity supply to the site could be delayed until the 2030s. The Nvidia-backed company had intended to install 23,040 AI chips at the facility, but uncertainty over the grid connection has disrupted the timeline.

The proposed data center in Loughton, Essex, about six miles northeast of London, was expected to provide 50 megawatts of computing capacity for Microsoft’s Azure cloud service by early 2027. However, UK Power Networks reportedly told Nscale that sufficient grid capacity might not become available until the early-to-mid 2030s, according to The Guardian, which followed earlier reporting by the Financial Times.

Nscale announced the project last year with Microsoft, Nvidia, and OpenAI. When The Guardian visited the site in February, it was reportedly still being used as a scaffolding yard. Nscale has since cleared the location and is examining alternative ways to obtain power. The company said its other Microsoft contracts remain unaffected.

Bitcoin Mining Companies Shift Toward AI Computing

Nscale emerged from Arkon Energy, a company that built infrastructure for Bitcoin mining. Founder Josh Payne applied that experience in securing land and electricity to develop a business supplying computing capacity to AI companies. Several publicly traded Bitcoin miners are now following a similar path as they seek to diversify their income.

A CoinDesk report published in March found that publicly listed miners had announced more than $70 billion in AI and high-performance computing contracts. Their combined Bitcoin holdings had fallen by more than 15,000 BTC from their peaks, partly because companies sold tokens to help finance their infrastructure expansion.

The shift comes as traditional mining profitability faces growing pressure. The March report estimated that the average listed miner’s cash cost of producing one Bitcoin was close to $80,000 in the fourth quarter of 2025. Bitcoin was trading near $70,000 by March, highlighting the financial strain facing some operators.

AI contracts can provide a more predictable stream of revenue over several years, but the facilities required to support these workloads demand much larger upfront investments. Building an AI-ready data center can cost an estimated $8 million to $15 million per megawatt, compared with roughly $700,000 to $1 million per megawatt for a Bitcoin mining facility.

AI clients also expect uninterrupted computing services, making dependable electricity supplies and continuous server operations essential.

AI Infrastructure Deals Have Yet to Generate Comparable Revenue

Despite the large value of announced contracts, actual revenue from AI infrastructure remains well below the total capacity companies have committed to deliver.

CoinShares’ latest mining report identified more than four gigawatts of AI capacity under contract among the mining companies it tracks. However, only approximately 550 megawatts were actively generating customer billings. Contract backlogs exceeded $100 billion, while annualized AI revenue was slightly above $1 billion.

Nscale’s setback illustrates how electricity constraints can prevent companies from turning agreements into operating infrastructure. Even with major technology partners and planned investments, a delayed grid connection can postpone deployment and push expected revenue further into the future.

Unless the company secures another power source or obtains an earlier connection date, the cleared site in Loughton could remain unused for years.

U.K. Grid Bottlenecks Threaten Data Center Expansion

The project’s difficulties reflect a wider electricity capacity challenge across Britain. Data center developers are competing for grid connections as demand for computing infrastructure grows.

Energy regulator Ofgem has said proposed data center projects have requested connections representing around 73 gigawatts of demand. It is considering stricter requirements because some applicants may never proceed with construction, potentially tying up capacity that could be allocated to viable developments.

For Bitcoin miners transitioning into AI computing, securing electricity is becoming a critical factor alongside financing, customer demand, and access to advanced chips. Nscale’s stalled project demonstrates how grid delays can derail major technology plans and undermine the expected returns from AI infrastructure investments.