Microsoft is adopting a more selective and controlled approach to its projects with specialized AI cloud companies such as Nebius, IREN, and Nscale, focusing on computing capacity that it directly owns and controls.
Microsoft’s review of its arrangements with Nebius, IREN, and Nscale
Microsoft’s official response and the combination of owned facilities and partners
IREN’s quarterly disclosures and the postponement of the $4 billion target to early 2027
Non-activation of the Texas agreement expansion option with Nscale and the risks of the specialized cloud model
Analytical reading of the strategic and economic dimensions of the event
Frequently asked questions
Microsoft’s review of its arrangements with Nebius, IREN, and Nscale
Microsoft is currently reviewing certain arrangements and contracts with specialized AI cloud computing providers, including Nebius, IREN, and Nscale, according to a report published by Bloomberg News and picked up by Reuters on November 5, 2026. The report noted that the company is showing hesitation in expanding certain projects facing delivery or operational risks, while sharpening its focus on computing capacity that it directly owns and controls. Reuters attributed these details to Bloomberg without independently confirming every internal decision.
In response to inquiries from Reuters, Microsoft explained that its strategy combines its own data centers, leased facilities, and contractual relationships with third-party providers. This clarification is significant because tightening selective oversight does not mean a wholesale withdrawal from relying on external AI infrastructure, as an infrastructure contract may be extended, modified, delayed, or halted for various reasons.
Microsoft’s official response and the combination of owned facilities and partners
Separately and simultaneously, IREN revealed in its quarterly financial results that requests for network design adjustments from Microsoft have caused delays in parts of its planned computing deliveries. The company emphasized that these specific delays will not incur financial penalties.
IREN has shifted its target date for achieving part of its annualized run-rate revenue goal of over $4 billion to January 2027, with a potential extension to February, noting that annualized run-rate revenue is not equivalent to actual audited annual revenue earned. IREN also announced that its quarterly AI cloud revenue reached approximately $200 million, jumping from $71 million in the previous quarter.
IREN’s quarterly disclosures and the postponement of the $4 billion target to early 2027
Reports also indicated that Microsoft decided not to exercise a contractual option that would have allowed it to more than double the size of a Texas computing agreement with Nscale.
A distinction must be made here between declining to exercise a future expansion option and canceling an existing binding contract. Concerns related to electrical grid connections and local building codes are also impacting other projects, though precise outcomes must be tied to developments named for each project.
Non-activation of the Texas agreement expansion option with Nscale and the risks of the specialized cloud model
The broader issue lies in specialized cloud providers relying on a relatively small number of large enterprise customers, expensive computing accelerators, and reliable access to electricity. This makes their business models highly vulnerable when a single delivery schedule slips, and the value of long-term contracted revenue depends heavily on the actual date when computing capacity comes online.
This report complements IREN’s earnings and new delivery disclosures because they underscore the same operational pivot. Important future evidence will include contract amendments, actively operated capacity, customer base diversification, cash flows, and final delivery schedules. These reviews illustrate that cloud giants like Microsoft have shifted from a phase of rapid contracting to secure any available capacity to a phase of rigorous engineering scrutiny regarding network efficiency and power-connection timelines.
Analytical reading of the strategic and economic dimensions of the event
Microsoft’s review of its arrangements with specialized AI cloud providers such as Nebius, IREN, and Nscale—along with its decision not to activate the Texas agreement expansion option with Nscale—reveals that the computing infrastructure market has entered a new phase of financial and engineering discipline. At the height of the chip shortage crisis over the past two years, tech giants rushed to contract with any external provider capable of supplying GPUs. Today, Microsoft is prioritizing capacity it directly owns and controls while tightening standards for network design and delivery deadlines.
IREN’s quarterly disclosures clearly illustrate this reality. Despite a surge in cloud revenue from $71 million to $200 million in a single quarter, the company was forced to delay part of its $4 billion annualized revenue target to January or February 2027 to implement engineering adjustments requested by Microsoft in the network design. This highlights the fragility of specialized cloud providers’ business models, which borrow billions of dollars to purchase processors and rely on a very limited number of major clients, making any technical or electrical delay directly impactful on their projected cash flows.
Frequently asked questions
Question: Who are the specialized cloud providers whose arrangements Microsoft is reviewing?
Answer: The review includes specialized AI cloud companies Nebius, IREN, and Nscale, with Microsoft focusing on capacity it directly owns and manages alongside partners.
Question: What did IREN disclose regarding its contracts with Microsoft and its quarterly revenue?
Answer: IREN clarified that Microsoft’s requests to modify network design delayed certain deliveries without financial penalties, and it postponed part of its $4 billion annualized revenue target to January or February 2027, while its quarterly revenue jumped from $71 million to $200 million.
Question: What decision did Microsoft make regarding Nscale’s project in Texas?
Answer: It decided not to activate a contractual option that would have more than doubled the size of the Texas computing agreement with Nscale, which is distinct from canceling an existing contract.
Question: What are the most prominent risks facing specialized AI cloud companies?
Answer: These companies rely on a limited number of large clients, the purchase of expensive processors, and securing electricity supplies, making them sensitive to any delays in operational timelines.
