Microsoft has rejected an option to more than double its existing Nscale compute rental agreement at a Texas data centre, the Financial Times reported on 5 November 2026. The decision removes a potential agreement worth tens of billions of dollars from the UK start-up’s pipeline after it filed for a US initial public offering in September.
Texas capacity option
The option concerned computing power at a Texas data centre. The available details concern the expansion option, not the full terms of the underlying Texas agreement.
Nscale is backed by NVIDIA and sells AI computing capacity. Microsoft and Nscale did not immediately comment outside business hours.
For enterprise infrastructure teams, the point is the dependency chain behind AI capacity procurement. Large commitments depend on data centre delivery, power availability and customer demand that can change before capacity comes online.
Essex project delay
The Texas decision follows another setback for Nscale involving Microsoft. Weeks earlier, Microsoft pulled out of a deal to be the anchor customer for Nscale’s Loughton data centre project in Essex.
Power connection delays pushed the expected supply date for that project into the 2030s. The Essex withdrawal and the Texas expansion decision are separate, but both concern the availability and timing of AI compute capacity.
IPO and customer concentration
Nscale filed for a US initial public offering in September. Microsoft and Anthropic were expected to be among its major customers.
Reuters reported last month that Nscale was in talks to sell AI computing power to Anthropic, the company behind Claude, in a deal that could be worth about USD 45 billion. That makes the loss of a larger Microsoft commitment material to how investors and customers read Nscale’s near-term capacity plans.
The case also underlines a wider issue for cloud buyers. AI capacity is often discussed as a software or model question, but the commercial risk often sits in data centre build-out, grid connections and long-term rental commitments.
When an AI compute rental option can change at a Texas data centre while a separate Essex project is pushed into the 2030s by power connection delays, ECS's Infrastructure topic deals with the same constraints behind Azure architecture, hybrid cloud, cost and platform engineering. Infrastructure sessions give attendees practitioner experience on capacity planning, platform choices and cost controls for cloud workloads. Bring the assumptions behind your own compute plans and use the conference to test them with people who run these platforms.
Photo: Mike Mozart, CC BY 2.0, via Wikimedia Commons, cropped
This article was written by the ECS editorial team with the assistance of Amal Agent, run.events' AI assistant, which supported the research, drafting and proofreading.