Microsoft has chosen leasing over building as its primary way of expanding AI infrastructure. The company signed a multiyear agreement to buy GPU capacity from Nebius valued at $17.4 billion over five years, with options that could take the total to about $19.4 billion. Nebius will deliver dedicated capacity from a new data centre in Vineland, New Jersey, beginning later this year.
“Nebius’s core AI cloud business, serving customers from AI startups to enterprises, is performing exceptionally well,” Nebius founder and CEO Arkady Volozh said in the company announcement. “I’m happy to announce the first of these contracts, and I believe there are more to come”.
Microsoft executives have repeatedly described the company’s immediate problem as supply, not demand. “We have been short power and space,” CFO Amy Hood told analysts on the company’s fiscal Q2 2025 earnings call.
The filings explain how the arrangement works: Microsoft secures multiyear committed GPU capacity while Nebius finances and builds the data-centre footprint, using contract cash flows and debt secured against the agreement. “The capital expenses associated with this deal will be covered by cash from the deal and debt secured against the contract,” Nebius stated in its Form 6‑K filed with the SEC.
The Nebius agreement is the largest infrastructure contract of its kind announced this year. CoreWeave’s five-year agreement with OpenAI was reported at up to $11.9 billion, and other specialist arrangements in the market have been valued in the low-to-mid single-digit billions. At the same time, larger hyperscaler partnerships (most notably reporting around Oracle and OpenAI) have been disclosed as major strategic supply relationships.
Other hyperscalers have leaned into outright ownership of infrastructure. Alphabet raised its full‑year 2025 capital‑spending guidance and significantly stepped up data‑centre investment to support AI and cloud growth. Amazon reported heavy capital expenditures in 2025, with AWS driving an outsized share of the company’s spend as it scales data centres and servers. By contrast, Microsoft’s use of Nebius and earlier partnerships with CoreWeave illustrate a different approach: spreading risk and accelerating capacity by leasing from GPU‑rich specialists rather than relying solely on in‑house buildouts.
Microsoft has also been expanding its model suite. In 2025 the company announced proprietary models including MAI‑1and the Phi‑3 family, which are being integrated into Copilot and Office products. Mustafa Suleyman, head of Microsoft AI, told Semafor the company aims to build “the strongest models in the world” and that Microsoft’s home‑grown models are being deployed to meet specific workloads. That surge in consumption is directly tied to infrastructure demand.
Satya Nadella tied product and capacity explicitly in Microsoft’s investor materials: “Cloud and AI are the essential inputs for every business to expand output, reduce costs, and accelerate growth”. Those inputs require GPUs in volume, which Microsoft is securing not only by building but by leasing.
For Nebius, the Microsoft contract is both validation and lifeline. “The economics of the deal are attractive in their own right, but, significantly, the deal will also help us to accelerate the growth of our AI cloud business even further in 2026 and beyond,” Volozh said. The company plans to use revenue from the contract and secured debt to fund expansion; its Form 6‑K makes clear obligations commence once Nebius confirms it has secured any additional financing required under the agreement.
The announcement sent Nebius shares sharply higher in after‑hours and premarket trading, taking them to record levels as investors priced in a multiyear revenue stream. Analysts pointed out the contract exceeds Nebius’s pre‑deal market capitalization and framed it as proof that neocloud suppliers can be strategic partners for hyperscalers.
Behind Nebius’s ability to promise capacity sits a December 2024 private placement that raised $700 million from investors including Nvidia and Accel. That financing both funds Nebius’s expansion and aligns it with a leading GPU vendor; coverage notes Nvidia’s participation could translate into engineering collaboration and preferential access to new GPU generations—an important advantage when chips, not only racks, are the bottleneck. For Microsoft, a supplier that pairs capital, engineering experience and a chipmaker relationship reduces one dimension of supply‑chain risk.
Despite Microsoft’s own record capital spending, demand for AI workloads continues to outpace supply. CFO Amy Hood acknowledged constraints on power and space that the company is working to relieve. Against that backdrop, leasing from Nebius looks like a deliberate model: secure committed capacity quickly while the company continues to build out long‑lived assets.
