Microsoft has decided not to exercise an option that would have more than doubled its existing deal with UK-based AI cloud provider Nscale for additional computing capacity at a Texas data center. The Financial Times reported that the add-on could have been worth tens of billions of dollars for Nscale, but a person familiar with the matter said Microsoft made the decision months ago, before Nscale filed for its September IPO. The existing Texas contract remains in place.
What this means for Nscale
Nscale is in the business of building expensive, long-lived infrastructure. Last year, it agreed with Microsoft to deploy about 200,000 Nvidia AI chips across facilities in Europe and the US. When a big customer hesitates to scale up, it raises questions about how much demand is firmly spoken for versus still being shopped around.
Nscale says there’s still plenty contracted. In its IPO filing, it disclosed statements of work with Microsoft worth up to about $43.8 billion through December 2033, not counting optional extensions. It also says its total disclosed contract value remains about $103 billion. But management is also trying to line up other buyers for capacity, after Microsoft previously stepped back from anchoring Nscale’s planned Loughton, Essex site because power-connection delays pushed the project’s timing into the 2030s.
Why the decision matters
Data-center builders live or die on “visibility” – how much future cash they can reasonably count on while they spend on power, chips, and construction. If a hyperscaler like Microsoft doesn’t take an expansion option, more of the build can become merchant capacity: space that still has to be financed and then resold to another customer.
Nscale says its Microsoft work orders run through December 2033 and it hasn’t changed its $103 billion disclosed contract value, so this isn’t necessarily a near-term revenue hole. But it does nudge the IPO narrative away from contracted backlog and toward the harder question investors price: how quickly the company can place the remaining Texas capacity, and on what terms, especially when power constraints have already reshaped plans like the Loughton project.
What it means for investors
For everyday investors, this is a reminder that even large, headline-grabbing AI contracts can have limits. When a major customer like Microsoft declines to expand, it doesn’t automatically mean the deal is falling apart – but it does signal that demand may not be as open-ended as some might hope.
For Nscale, the focus now shifts to its ability to secure new customers for the Texas capacity. The company’s IPO story will likely hinge on whether it can demonstrate that its contracted backlog is solid and that it can fill any remaining space without relying on optional expansions.
Microsoft’s decision also comes as Microsoft's AI bet pays off, with Azure surging 48% and margins holding. That suggests the company is still investing heavily in AI, but it’s being selective about where it commits capital.
For the broader market, this is a data point in the ongoing story of AI infrastructure spending. European stocks are set to rise as the news filters through, but the long-term implications for AI cloud providers remain to be seen.
Investors should watch how Nscale addresses the Texas capacity in its IPO roadshow and whether it announces new customers. The company’s ability to convert its $103 billion disclosed contract value into actual revenue will be key to its valuation.
In the meantime, the decision underscores the importance of power availability in the AI buildout. Power constraints have already reshaped Nscale’s plans, and they could continue to influence where and how AI infrastructure gets built.
