Key Points

• Microsoft rejected the opportunity to further expand its computing power cooperation with Nscale, with a potential scale of several billion dollars.
• The existing $43.8 billion agreement between the two parties is not affected.
• This decision highlights that the risks faced by AI infrastructure developers are on the rise.

It is reported that Microsoft has given up the opportunity to more than double the scale of its artificial intelligence computing power agreement with NVIDIA in Texas, and has withdrawn from a potential expansion plan worth billions of dollars.

Reuters disclosed this decision on November 5th, following a report by the British newspaper The Financial Times earlier in the week. A source familiar with the matter stated that Microsoft made this decision several months ago, prior to Nscale submitting its initial public offering (IPO) application to the United States in September.

It is worth noting that the existing contracts between both parties in Texas are still valid. This decision relates to additional optional capacity, rather than canceling the infrastructure expenditures that have already been committed to.

Microsoft abandons expansion in Texas AI, but the $43.8 billion commitment remains

The existing cooperative relationship between Nscale and Microsoft is already quite substantial.

According to documents submitted to the U.S. Securities and Exchange Commission in September by the company, agreements signed between September 2025 and April 2026 stipulate that, subject to meeting service delivery and availability requirements, the total related payments are expected to amount to approximately $43.8 billion by December 2033.

These contracts cover GPU infrastructure, data center hosting facilities, as well as cloud computing services. These agreements constitute a portion of the total contract value of approximately $103 billion for Nscale.

According to people familiar with the matter cited by Reuters, Microsoft's decision not to expand its production capacity in Texas will not reduce the total amount of the existing contracts mentioned above.

The two companies had previously agreed to deploy approximately 200,000 NVIDIA AI chips in their facilities in Europe and the United States.

However, before this latest development, Microsoft had already withdrawn from the data center project for Nscale in the UK; due to delays in power access, the expected time when the project could receive power supply was postponed to the 2030s.

Nscale faces financing doubts ahead of IPO

As investors become increasingly cautious about financing large AI computing power facilities, this latest setback has occurred.

Nscale has applied for listing in the United States and is advancing large-scale infrastructure expansion with the support of major technology clients.

However, the company's business still depends on its ability to deliver computing power on time and on its ability to continue to obtain financing.

Previous analyses regarding the infrastructure debt of AI have indicated that borrowing costs are on the rise, and the market is increasingly concerned about whether future revenues will be sufficient to justify the construction expenditures.

For those companies that have built facilities in advance before customers start paying for computing power services, this challenge is particularly significant.

Microsoft's decision also comes against the broader backdrop of more cautious spending in the AI sector, where technology companies are increasingly evaluating infrastructure costs based on anticipated needs.

However, it is reported that Nscale is in discussions with the remaining production capacity of the state and other large customers. This means that Microsoft's decision to refrain from expansion does not necessarily mean that these infrastructures will not find potential buyers.

What does this decision by Microsoft mean for the AI data center?

This development highlights an increasingly clear divergence between companies that purchase computing power and those that finance the construction of computing infrastructure.

Microsoft can adjust its future production capacity commitments without abandoning its broader AI strategy. At the same time, infrastructure developers still have to bear high construction costs and financing obligations.

As political restrictions surrounding data centers raise more questions about power supply capacity and future expansion prospects, this distinction is becoming increasingly important.