On October 18, 2026, multiple outlets reported that Microsoft has withdrawn from its agreement to serve as anchor tenant at Nscale’s planned AI data center in Loughton, Essex. The site was designed for roughly 50–90 MW and approximately 23,000 Nvidia GPUs with a target opening in 2027; UK Power Networks has indicated sufficient grid power is not expected until the early-to-mid 2030s. Nscale says it remains committed to the project and is exploring alternative power sources; its broader relationship with Microsoft is unaffected. Technology and infrastructure teams evaluating AI capacity should treat power availability and grid connection timelines as first-order constraints, not secondary details.
Microsoft’s exit from the Loughton project, reported October 18, underscores that grid connection delays can cancel even large contracted AI capacity years before chips are installed. Nscale, which has filed for a U.S. IPO and lists Microsoft and Anthropic among major contracted customers, now has no announced tenant for the Essex site. Operators planning multi-year AI infrastructure should add power-interconnection risk to due-diligence checklists.
- What changed: Microsoft terminated its anchor agreement for the Loughton data center; the site lacks a replacement tenant at the time of reporting.
- When: Reported October 18, 2026 by The Telegraph and confirmed in subsequent coverage. Original 2025 announcement targeted 2027 delivery.
- Who is affected: Nscale (IPO process and backlog), Microsoft’s UK Azure capacity plans, and any enterprise or neocloud customer that had modeled capacity from this site.
- What to do this week: Review any contracted or planned AI capacity that depends on new UK or European data centers; request updated power-connection timelines from providers; model scenarios in which grid delays push commercial operation into the 2030s.
What was the original Loughton plan?
In September 2025 Nscale and Microsoft announced that the Loughton, Essex site would host the UK’s largest AI supercomputer for Azure services. The facility was described as delivering 50 MW of AI capacity, scalable to 90 MW, and was planned to house approximately 23,040 Nvidia GB300 GPUs with delivery beginning in the first quarter of 2027. Nscale had pledged roughly £2 billion to the project, which the UK government highlighted as part of its AI strategy. BBC and company statements later confirmed the 2027 target and the scale of the planned deployment.
Why did Microsoft withdraw?
The Telegraph reported that Microsoft recently terminated the agreement. Neither company has issued a detailed public explanation of the commercial reasons. Reporting consistently cites repeated project delays and the fact that UK Power Networks does not expect to be able to supply sufficient power until the early-to-mid 2030s. Nscale has stated that the London-wide grid issue does not affect its other contracts with Microsoft and that it is exploring alternative power sources for Loughton while remaining committed to the UK.
What does this mean for Nscale’s IPO and backlog?
Nscale filed an S-1 registration statement with the SEC in September 2026 for a New York Stock Exchange listing under the ticker NSCL. Reporting on the filing indicates a total contracted value of approximately $103 billion, of which Microsoft and Anthropic account for roughly 85 percent (Microsoft up to about $43.8 billion through 2033; Anthropic up to about $44.6 billion). Only a small fraction—around $2.6 billion or 55 MW—was described as live operational capacity. Loss of the Loughton anchor tenant removes one concrete UK delivery vehicle and highlights the gap between contracted backlog and power-constrained, deliverable capacity. Technology buyers evaluating neocloud providers should distinguish between signed agreements and energized sites. A signed agreement does not create usable compute until the site has a firm grid connection and the equipment is installed and energized. Nscale has also reported delays at its flagship Norway facility.
How should technology teams adjust planning assumptions?
Power availability has become a binding constraint on AI infrastructure timelines in multiple markets. Teams that have included UK or European capacity in 2027–2028 roadmaps should request written confirmation of grid connection dates and any contingency power arrangements. Where capacity is contracted with neocloud providers, ask for disclosure of the specific sites and interconnection status supporting those commitments. Diversifying across providers and regions that already have energized capacity reduces single-site grid risk.
What to watch next?
Nscale’s response on alternative power (on-site generation, different interconnection points, or other UK sites) and any updated IPO timing or risk-factor language. Microsoft’s public comments, if any, on UK capacity plans. Broader UK grid-upgrade timelines that affect other announced AI campuses. The episode is consistent with earlier reporting on CoreWeave and other neoclouds facing financing and delivery pressure amid the AI capex cycle.
- Is the Loughton project cancelled? Nscale says it remains committed and is exploring alternative power sources; Microsoft is no longer the anchor tenant.
- When was power expected? Reporting states UK Power Networks indicated supply is not expected until the 2030s.
- Does this affect Microsoft’s other Nscale contracts? Nscale has stated that other contracts are unaffected.
- How large was the planned deployment? Approximately 23,000 Nvidia GPUs at 50–90 MW, targeting 2027.
- Why does this matter beyond one UK site? It illustrates that grid interconnection, not only chip supply or financing, can determine whether announced AI capacity is delivered on schedule.
Last Updated: October 19, 2026. Related reading: CoreWeave Bonds Reach Distressed Levels on October 15, 2026: Notes for Technology and Technology hub.
Discover more from Kurums | Business Intelligence
Subscribe to get the latest posts sent to your email.