Connect with us

AI

Nebius Q2 Turns AI Cloud Demand Into Prepaid Growth Machine

Nebius Group posted 454 percent revenue growth and 50 percent AI cloud margins after locking four billion-dollar deals with prepayments that cut payback under.

Published

on

Nebius Group N.V. reported Q2 2026 revenue of $582.3 million, up 454 percent from $105.1 million a year earlier, as its AI cloud business locked four landmark contracts averaging more than $1 billion in total value each and pushed annualized run-rate revenue to $3 billion.

Adjusted EBITDA swung to a $236.2 million profit from a $21 million loss, while the pure AI cloud segment delivered a 50 percent adjusted EBITDA margin. The numbers mark more than volume growth. They show specialized providers now extract prepaid, high-yield contracts that fund their own expansion.

Q2 Turns the Growth Story Into Cash Flow

Consolidated results for the three months ended June 30, published in the company’s second quarter 2026 financial results, cover Nebius AI cloud plus Avride autonomous vehicles and TripleTen edtech. AI cloud itself generated about $575 million, up 514 percent year over year.

Metric (USD millions) Q2 2025 Q2 2026 Change
Revenue 105.1 582.3 +454%
Adjusted EBITDA (21.0) 236.2 n/m
Net cash from ops (cont.) (167.7) 2,246.1 n/m
Capex (PPE + intangibles) 510.6 5,657.4 +1008%

First-half revenue reached $981.3 million, up 529 percent. Operating cash flow turned massively positive on customer prepayments. Capex exploded as the company raced capacity online, yet 50-60 percent of associated build costs on new deals came covered by those same prepayments.

Adjusted net loss narrowed to $33.2 million from $91.5 million. Net income from continuing operations showed a $190.4 million loss in the quarter after heavy depreciation and other items, while the first half still posted $430.8 million of net income. The core business is now cash-generative at the operating level even as the balance sheet absorbs multi-gigawatt construction.

Four Billion-Dollar Deals Reset the Price of a Megawatt

Founder and CEO Arkady Volozh’s shareholder letter framed Q2 as the quarter demand inflected into contracted economics. The company closed four landmark AI cloud deals with average total contract value above $1 billion each. Annual contract value per megawatt stepped to more than $20 million, and in some cases $20-25 million, versus a prior base nearer $12 million.

  • Average TCV per landmark deal: greater than $1 billion
  • ACV yield: $20-25 million per MW on the big wins
  • TCV growth of Q2 closes: nearly 4x quarter over quarter; new-customer TCV more than 9x
  • Prepayment share: roughly 70 percent of deals included them
  • Expected payback: 1 year 10 months, down from two-to-three years

Volozh wrote that the company closed its largest AI Cloud deals on its strongest terms to date, at prices that represent a step-change in the economics of the business. Management said it could sell its entire 2027 capacity on these terms today but is deliberately holding some back for higher-value near-term needs.

Named customers in the letter include Reflection for open-source model training, Cohere for enterprise agentic solutions, a US AI lab for frontier development, and a large quantitative trading firm. Vertical wins range from AMI (co-founded by Yann LeCun) to biological foundation models at Basecamp Research and Prima Mente, plus Higgsfield’s AI feature film trained on the platform.

We closed our largest AI Cloud deals on our strongest terms to date, at prices that represent a step-change in the economics of our business.

Arkady Volozh, founder and CEO, said that in the Q2 letter to shareholders.

The UK Buildout and the Race to 5 GW

Capacity targets keep rising. Nebius raised its year-end 2026 contracted power guidance to 5 GW from prior levels above 4 GW. It plans to deploy more than 1 GW per year starting in 2027. Sites now span the US (Minnesota, Kansas City, Pennsylvania, New Jersey, Missouri, Oklahoma, Alabama) plus Europe (Iceland, UK, France, Spain, Finland, Estonia) and Israel.

In June the company committed approximately £1.7 billion (about $2.3 billion) to a £1.7 billion UK capacity buildout. Three new NVIDIA-powered deployments plus the existing Blackwell Ultra site will reach a combined 65 MW when fully ramped in 2027. The move aligns with the UK government’s AI Opportunities Action Plan and adds local sovereign compute for fintech, healthcare and research customers such as Revolut and Prima Mente.

UK AI Minister Kanishka Narayan welcomed the addition of significant AI compute that lets companies train and run advanced systems at home. NVIDIA’s Paolo Guglielmini called the UK one of Europe’s most ambitious AI markets and said Nebius expands local access to the full-stack AI factory platform.

The same scarcity dynamic that lifts pricing also forces geographic choices. European and UK demand for low-latency, data-sovereign capacity creates pockets where a focused player can out-execute global hyperscalers on speed and terms. That dynamic sits alongside broader shifts toward hybrid data center architectures that mix central training clusters with edge inference.

Prepayments and the $775 Million Debt Template

In July Nebius closed its first $775 million senior secured facility, backed by deployed GPU infrastructure and contracted cash flows from an investment-grade customer. Pricing was SOFR + 2.50 percent, maturity October 31, 2030. The facility plus customer cash flows covers more than 100 percent of the related capex.

Management described a repeatable framework. With more than $40 billion of additional contracted revenue already in place from investment-grade customers such as Microsoft and Meta, the company expects further asset-level financings on similar terms. One earlier Meta agreement carries a contract value of up to approximately $27 billion over five years for dedicated capacity starting 2027.

Cash ended Q2 at roughly $8 billion, including $2.3 billion of positive operating cash flow in the half. Prepayments are expected to exceed $9 billion across full-year 2026. The model turns long-term contracts into both revenue visibility and construction finance, a structure that increasingly treats specialized AI compute the way Wall Street now treats AI compute into a $500 billion asset class.

COO Ophir Nave said the financing reinforces confidence in a disciplined, diversified approach of owned data centers and asset-light partnerships that can deliver durable margins.

Who Is Buying Nebius Capacity

Demand spans AI labs, open-source developers, enterprises and regulated industries. Token Factory inference workloads more than tripled in the quarter, increasingly on open-weight models such as Kimi K3, GLM 5.2 and Nemotron Ultra. Acquisitions of Eigen AI and Clarifai brought inference optimization in-house. The platform shipped Aether 3.6 (Nebius AI Cloud 3.6) and added the Echo agent for workload management.

Contract types now deliberately mix durations:

  • Short three-to-six-month premium deals for acute needs (first one signed in early Q3)
  • Core one-to-three-year mid-term contracts with ambitious AI companies
  • Long-term investment-grade agreements that underwrite financing

An early capacity auction pilot cleared the highest Blackwell prices the company has seen. The mix lets Nebius capture rising spot demand while locking multi-year cash flows that banks will lend against.

Avride’s autonomous fleet nearly tripled year-to-date with more than 1 million miles and over 60,000 commercial Uber rides. TripleTen continues reskilling. Equity stakes include ClickHouse and Toloka. These remain secondary to the AI cloud engine.

From Yandex Spinout to Pure-Play AI Cloud

Nebius Group N.V. is the renamed and restructured Dutch parent that once held Yandex. After the 2022 trading suspension and the 2024 sale of Russian assets, the company retained international operations, roughly 1,000 former Yandex engineers, and the Nasdaq listing. Trading resumed in October 2024 under the NBIS ticker with Volozh as CEO. Headquarters sit in Amsterdam.

The post-split thesis was simple: take deep infrastructure talent and capital, shed geopolitical overhang, and build a full-stack AI cloud from silicon-adjacent hardware through training, inference and agentic tooling. Two years later the commercial proof points are the $3 billion run-rate, 50 percent AI cloud margins, and a financing template that hyperscalers themselves helped validate by signing the largest offtake deals.

Investors on X highlighted the simultaneous beat on revenue and adjusted EBITDA, the ARR jump from roughly $1.9 billion at the end of March to $3 billion, and the capacity guidance raise. The crowd also noted that the capex bill remains enormous and that execution risk on multi-GW construction has not vanished. The prevailing read is that demand still outruns supply and that Nebius’s pricing and prepay terms now reflect that imbalance.

The Math That Keeps Scaling

Full-year 2026 guidance was reaffirmed across metrics. Management expects the landmark Q2 deals to contribute primarily to 2027 revenue as the underlying capacity arrives in late 2026 and beyond. Short-term and auction capacity can still capture near-term upside at premium prices.

The second-order shift is structural. When ACV per megawatt doubles, when 70 percent of deals carry prepayments that cover half the capex, and when those contracts support SOFR-plus-250 debt that fully funds the build, growth stops being a pure capital sink. It becomes a flywheel. Nebius can raise its contracted power target to 5 GW, plan 1 GW-plus annual deployments, and still show 50 percent adjusted margins on the core cloud while sitting on $8 billion of cash.

That is the story the 454 percent headline only begins to tell. The market is paying specialized AI infrastructure providers for scarcity, speed and software stack, and those providers are converting the payments into self-reinforcing capacity. For Nebius the Q2 print is evidence the conversion is working at scale.

Logan Pierce is a writer and web publisher with over seven years of experience covering consumer technology. He has published work on independent tech blogs and freelance bylines covering Android devices, privacy focused software, and budget gadgets. Logan founded Oton Technology to publish clear, no nonsense tech news and reviews based on real hands on testing. He has personally tested and reviewed dozens of mid range and budget Android phones, written extensively about app privacy, and built and managed multiple WordPress publications over the past decade. Logan holds a bachelor's degree in English and studied digital marketing at a certificate level.

Continue Reading
Click to comment

Leave a Reply

Your email address will not be published. Required fields are marked *

Trending