Microsoft Reaffirms $30 Billion UK Investment After Leaving Nscale Deal

Microsoft reaffirmed its planned $30 billion UK investment through 2028 after withdrawing from an agreement to anchor Nscale’s proposed Loughton AI data center, saying the site decision would not change the wider plan.

For 2025 through 2028, Microsoft has allocated half of the $30 billion total to cloud and AI infrastructure capital spending and half to operating costs. The total combines infrastructure spending with operating costs; it is not presented as $30 billion in capital spending alone.

Darren Hardman, CEO of Microsoft UK and Ireland, described the Loughton project as a small part of the package. “We are still fully committed to that investment package,” he said.

Microsoft characterized its withdrawal as a single-site decision. It has not publicly identified power delays as the reason for its exit. The company’s withdrawal from the agreement to anchor the project left Nscale’s proposed site without a publicly announced customer.

Nscale says it remains committed to Loughton and is exploring alternative power sources. Its Sept. 16, 2025, plan called for 50 megawatts of initial AI computing capacity, expandable to 90 megawatts, and 23,040 NVIDIA GB300 GPUs scheduled for delivery in the first quarter of 2027 to support Microsoft Azure. Those specifications and the GPU schedule were part of the original plan; they do not establish a current delivery schedule or customer arrangement. Nscale has not announced a replacement customer.

The site’s planned opening and GPU delivery were separate milestones. An earlier Nscale schedule targeted a fourth-quarter 2026 opening, while the later plan set GPU delivery for the first quarter of 2027. No revised opening date has been announced.

The project also faces a wider UK grid-connection constraint. UK Power Networks said it could not discuss individual connections and that many data centers are waiting for work on the national transmission network, which it does not own. The UK government identifies timely grid connections as the biggest obstacle to establishing AI Growth Zones.