Nebius Group Announces $27 Billion Meta Contract and Raises CapEx Guidance to $25 Billion as Demand Outstrips Supply
Q1 2026 Earnings Call, May 13, 2026 — Revenue Surges 684% Year-on-Year as AI-Native Hyperscaler Accelerates Buildout
The Meta Contract Is Bigger Than It Looks
The headline number from Nebius Group's first quarter earnings call is a $27 billion, five-year contract with Meta, but management was explicit that the actual economic value extends well beyond that figure. The deal is structured in two parts: a $12 billion committed tranche for dedicated compute capacity with delivery beginning in early 2027, and a $15 billion option tranche where Nebius retains discretion over whether to allocate capacity to Meta or sell it to its own AI cloud customers at potentially higher market prices. Chief Revenue Officer Marc Boroditsky was direct: "Meta is committed to buy up to $15 billion of any capacity in these clusters at our option during the entire 5-year contract. This commitment will likely allow us to finance the clusters with asset-backed financing at attractive terms, while selling them to our AI cloud customers at potentially higher market prices." In other words, the $15 billion option functions simultaneously as a financing backstop and a margin-enhancement mechanism. If spot cloud pricing remains elevated, Nebius can sell to third parties at a premium and simply not exercise the Meta allocation. The floor is protected; the upside is uncapped. "If the market remains strong, we should generate more than $27 billion in revenue from this great agreement," Boroditsky added.
CapEx Guidance Raised Sharply — This Is a 2027 Bet, Not a 2026 One
Nebius raised its full-year 2026 capital expenditure guidance to a range of $20 billion to $25 billion, up from the prior $16 billion to $20 billion range disclosed just one quarter ago. Management was careful to frame this not as a reaction to cost inflation but as a forward-looking commitment to 2027 capacity for which customer commitments are already in place. CFO Dado Alonso stated that more than 90% of the original CapEx range is already secured by cash and contractual commitments, and that the incremental spend above that range will be funded through asset-backed financing against Microsoft and Meta contracts, potential corporate debt, and the company's at-the-market equity program, which covers up to 25 million Class A shares and has not yet been utilized. Infrastructure head Andrey Korolenko confirmed the cost inflation impact was modest: "The impact of component inflation in our 2026 program was quite material, around low single digits as a percentage of total spend, also because we secured a lot of 2026 back in 2025 at the previous price levels." The capital raise trajectory has been aggressive: more than $6 billion raised so far in 2026, including $4.3 billion in convertible notes at coupons of 1.25% and 2.60%, and a $2 billion equity investment from NVIDIA.
Revenue and Margin Inflection Is Real, But Q2 Will Be a Speed Bump
Group revenue reached $399 million in Q1 2026, up 684% year-on-year and 75% sequentially. The core Nebius AI business, stripping out consolidated investments in autonomous driving venture Avride and online education platform TripleTen, grew 841% year-on-year to $390 million and now represents 98% of group revenue. Annualized run rate revenue for the AI business exited March at $1.9 billion, up from $1.25 billion at the end of Q4 2025. Group adjusted EBITDA was $130 million at a 32% margin, compared to $15 million and a loss of $54 million in the year-ago period. The Nebius AI business itself ran at a 45% adjusted EBITDA margin, nearly double the 24% posted in Q4. Management attributed the improvement to operating leverage, pricing gains, and strong utilization rather than any one-time items.
However, Alonso was explicit that Q2 margins will decline before recovering. The explanation is mechanical rather than structural: investments in headcount, acquisitions, and product development have already hit the cost base, while the capacity additions that will generate the offsetting revenue are predominantly back-end weighted. "We actually expect those margins in Q2 to go a little bit lower, returning to Q1 levels in Q3 and stepping even higher in Q4." Full-year group adjusted EBITDA margin guidance of approximately 40% was reiterated, implying a material step-up in the second half of the year that investors will need to take on faith until Q3 results confirm the trajectory.
Contracted Power Capacity Accelerating Beyond Initial Targets
One quarter ago, Nebius had contracted more than 2 gigawatts of power and was targeting 3 gigawatts by year-end. Today, contracted power stands at more than 3.5 gigawatts, and the new target is at least 4 gigawatts by end of 2026. A newly announced Pennsylvania site adds 1.2 gigawatts of potential capacity, making it the company's second owned gigawatt-scale U.S. site. Korolenko outlined the buildout timeline: Pennsylvania reaches first power with 250 to 300 megawatts by end of 2027, adding roughly 300 megawatts per year thereafter to reach full 1.2 gigawatt capacity by early 2030. He noted that more than 75% of total contracted capacity is now on owned sites, which management views as structurally more efficient for their full-stack platform model. On the previously reported New Jersey Vineland site, Korolenko confirmed all delivery commitments under the Microsoft contract have been met on schedule, with remaining tranches weighted toward Q3 and Q4 of this year.
Inference and the Software Stack Are Becoming the Differentiation Layer
Three acquisitions closed in 2026 — Tavily, Eigen AI, and Clarifai — reflect a deliberate effort to build a defensible software layer on top of the compute infrastructure. Eigen AI, recognized by NVIDIA as the number one speed inference provider, optimizes at the model level; Clarifai optimizes at the system level. Together they are being integrated into the company's Token Factory inference product. Tavily, an agentic search platform, was described as bringing developer adoption that "would, in general, take us meaningfully longer to build organically." Chief Product Officer Roman Chernin was clear that software is not being built as a standalone revenue stream: "The software first of all plays the role of unlocking the new capabilities for us, unlocking the new opportunities in the types of the workloads that are growing on the market and the types of the customers that we can address." The framing is that software expands the addressable customer base rather than generating independent margin pools, at least for now. On the agentic opportunity specifically, Chernin acknowledged the market is still forming: "It's still to be defined what is the final shape of the product and who will be the winners."
Demand Pipeline Growing Faster Than Capacity, With Pricing Tailwinds
Pipeline generation in Q1 grew 3.5 times sequentially versus Q4 2025, described as a record for the business. Critically, this figure excludes the Meta hyperscaler deal and reflects only the core AI cloud and Token Factory customer base across AI-native companies, software vendors, and enterprises. Boroditsky noted that pricing is moving in Nebius's favor: "We just raised prices again in the latest quarter, and we are still selling out across all chip types at the higher prices." Contract durations are lengthening, average contract values are rising, and prepayments are increasing across customer types including hyperscalers. The company is typically seeing four or more customers competing for each GPU brought online. Notable new customer wins disclosed include Revolut and monday.com for Token Factory, Sword Health in healthcare and life sciences, and 1X Technologies in physical AI robotics.
NVIDIA Partnership Extends Beyond Capital to Supply Chain Certainty
The $2 billion NVIDIA equity investment is being described in terms that go beyond the balance sheet contribution. Korolenko framed it as securing "multiyear partnership with our most important hardware supplier at the moment when access to GPU supply is a competitive advantage." Specifically, Nebius gains differentiated supply chain visibility on future Rubin and Vera CPU platforms, early deployment access as new SKUs become publicly available, and co-design collaboration with NVIDIA on future architectures. The company has now achieved NVIDIA Exemplar Cloud status on GB300 for training workloads, adding to prior certifications across multiple GPU generations, a distinction held by a very small number of cloud providers globally. The NVIDIA relationship also comes with a commitment framework extending to 5 gigawatts of capacity by end of 2030.
Balance Sheet and Financing Runway Are Adequate, But Equity Dilution Risk Remains Real
Cash and cash equivalents ended Q1 at $9.3 billion, supported by operating cash flow of $2.3 billion in the quarter, driven primarily by customer prepayments. The company has established the infrastructure for asset-backed financing against its Microsoft and Meta contracts, corporate debt issuance, and an at-the-market equity program. Management emphasized it has not yet tapped the equity program, but with CapEx now guided to as much as $25 billion against a $9.3 billion cash base, the financing math requires successful execution across multiple capital markets simultaneously. Net income of $621 million in Q1 was flattering, as it included a non-cash upward valuation adjustment tied to a recent ClickHouse funding round rather than operating performance. The Avride and TripleTen units continue to consume capital and weigh on group margins; management reiterated its intention to find strategic and financial partners for both businesses and eventually deconsolidate them.
Full-Year Guidance Maintained, But Back-Half Execution Is Load-Bearing
Annual targets were reiterated without change: annualized run rate revenue of $7 billion to $9 billion, full-year group revenue of $3 billion to $3.4 billion, and group adjusted EBITDA margin of approximately 40%. With Q1 annualized run rate already at $1.9 billion for the AI business alone, reaching the top end of the ARR range implies roughly 4.7 times growth from current levels by year-end, all contingent on the Q3 and Q4 capacity additions landing on schedule. Alonso summarized the key variables with appropriate candor: "Three key parameters will determine our growth profile and margin progression throughout the year — utilization, pricing, and capacity. At present, neither of the first two parameters is limiting our growth." The constraint is entirely on the capacity delivery side, making the second half of 2026 the pivotal period for validating the investment thesis.
Nebius Group N.V. Deep Dive
Business Model and Core Monetization Strategy
Nebius Group N.V. has aggressively transitioned from a diversified regional technology conglomerate into a pure-play, vertically integrated artificial intelligence infrastructure provider. Following a complex restructuring, the Amsterdam-headquartered entity repositioned itself as a leading neocloud, a specialized cloud computing operator catering exclusively to artificial intelligence and machine learning workloads. The core engine of the Nebius business model is a specialized graphics processing unit as a service framework. The company procures high-performance compute hardware, houses it in ultra-efficient data centers across Europe and North America, and leases the computational power to artificial intelligence developers. Revenue is generated via an elastic, consumption-based pricing model for short-term workloads, alongside lucrative, multi-year reserved capacity contracts for enterprise clients. Nebius distinguishes itself from traditional hyperscalers by avoiding the overhead and architectural compromises of general-purpose enterprise computing. Instead, it offers a full-stack computational ecosystem, integrating non-blocking networking fabrics, proprietary cluster orchestration software, and streamlined inference studios. By the end of the first quarter of 2026, this infrastructure-first commercial strategy drove a spectacular 684 percent year-over-year revenue increase to $399 million, pushing the company's annualized recurring revenue into hypergrowth as it targets a $7 billion to $9 billion long-term capacity envelope.
Product Ecosystem and Key Services
The foundational pillar of the corporate portfolio is the Nebius AI Cloud, a robust platform offering on-demand and reserved instances of the latest high-performance processing hardware. This architecture supports the entire machine learning lifecycle, from massive foundational model training runs utilizing tens of thousands of interconnected chips, to low-latency inference deployments managed through proprietary serverless interfaces. Beyond the core metal layer, the company maintains a portfolio of strategic, high-upside subsidiaries born from its corporate legacy. Avride functions as a venture-style autonomous mobility arm, developing both passenger robotaxis and sidewalk delivery robots governed by a unified artificial intelligence platform. Toloka, which transitioned to an equity stake following a deconsolidation transaction in 2025, serves the foundational model industry by providing human-in-the-loop data labeling and reinforcement learning from human feedback. TripleTen operates as an edtech and reskilling bootcamp platform, providing a steady, albeit non-core, tertiary revenue stream. Furthermore, the strategic acquisition of Eigen AI for $643 million in early 2026 has significantly bolstered the company's inference software capabilities, indicating a deliberate pivot toward the high-margin orchestration layer to complement its hardware density.
Customers, Competitors, and Supplier Dynamics
Nebius operates in a highly concentrated ecosystem defined by extreme hardware dependency and fierce capital arms races. The ultimate supplier and de facto kingmaker in this market is Nvidia. Nebius has secured its status as a preferred partner, receiving early and scaled access to coveted hardware architectures. Nvidia serves not just as a supplier but as a strategic investor, validating the Nebius technical architecture to the broader market. On the demand side, the customer base ranges from highly capitalized generative artificial intelligence startups to colossal enterprise operators. The firm recently validated its enterprise scale by securing a multi-billion dollar, long-term infrastructure supply agreement with Meta, confirming that massive foundational model builders are utilizing specialized neoclouds to augment their internal data center footprints. Competitively, Nebius faces off against a two-tiered adversary landscape. The first tier consists of legacy hyperscalers such as Amazon Web Services, Microsoft Azure, and Google Cloud Platform. These giants boast infinite capital but suffer from bloated, general-purpose architectures that fail to maximize chip efficiency. The second, more direct competitive tier comprises fellow neoclouds such as CoreWeave, Lambda Labs, and Crusoe. While CoreWeave remains the domestic heavyweight in the United States, Nebius has aggressively counter-positioned itself as the dominant sovereign alternative in Europe, all while aggressively penetrating the North American market through gigawatt-scale buildouts.
Market Share and Industry Dynamics
The global infrastructure market is undergoing an explosive, structural expansion driven by the computational demands of generative models. Industry forecasts project neocloud revenues to eclipse $250 billion by 2030, a staggering leap from roughly $42 billion in 2025. Legacy hyperscalers currently dominate the broader cloud computing market with an estimated 60 percent market share, but specialized neoclouds are rapidly capturing the dedicated artificial intelligence training segment. Within this high-growth niche, CoreWeave leads in gross domestic deployment, having reached a $1.6 billion quarterly revenue run rate by late 2025. Nebius, operating from a smaller initial baseline, is closing the scale gap with astonishing speed. With first-quarter 2026 revenues of $399 million, Nebius commands an expanding mid-single-digit percentage of the global neocloud market. More importantly, the company has secured a dominant, near-monopoly market share of the pure-play sovereign artificial intelligence infrastructure sector in Europe. The push for digital sovereignty by European regulators has created a commercial moat, driving domestic enterprises to favor providers governed outside of the United States. As the company brings a massive 1.2 gigawatt artificial intelligence factory in Pennsylvania and a 310 megawatt facility in Lappeenranta, Finland, online, Nebius is structurally positioned to secure a robust double-digit global market share within the next 24 months.
Competitive Advantages and Strategic Moats
The primary moat encircling Nebius is its immense concentration of human capital, an inheritance from its history as a dominant search and machine learning conglomerate. Following its restructuring, the company retained a specialized workforce of over 1,300 engineers, representing a talent density that is virtually impossible to replicate in today's constrained labor market. This deep engineering DNA allows Nebius to achieve superior operational efficiency compared to peers who merely assemble off-the-shelf servers. By writing its own cluster management software and optimizing network topologies, the firm extracts maximum compute performance per megawatt, offering a highly attractive total cost of ownership to end users. Furthermore, Nebius leverages a formidable structural cost advantage regarding energy consumption. Its flagship data centers in Finland and Iceland capitalize on abundant, inexpensive Nordic power. Proprietary liquid-cooling systems and advanced heat-recovery technologies export excess thermal energy to heat local municipalities, which dramatically lowers the power usage effectiveness ratio. This mechanical efficiency protects gross margins in an industry where electricity represents the single largest variable operating cost. The direct equity investment and strategic partnership with Nvidia crystallize these advantages, securing priority allocation in a perpetually supply-constrained silicon market.
Opportunities, Threats, and Industry Dynamics
The clearest opportunity for Nebius lies in the industry-wide transition from model training to model inference. As large language models mature and reach production scale, the computational burden shifts from periodic, massive training runs to continuous, high-volume inference queries. The recent acquisition of Eigen AI positions the company perfectly to capture this higher-margin software orchestration layer, transitioning the business from a pure hardware lessor to an integrated platform provider. Additionally, the proliferation of autonomous robotics presents an asymmetric upside. Through the Avride subsidiary, Nebius has partnered with Uber in a multifaceted agreement involving up to $375 million in strategic funding to roll out autonomous delivery bots and a robotaxi fleet in Texas. If Avride successfully scales this hybrid network, it will generate an entirely parallel, high-margin mobility revenue stream. Conversely, the strategic threats are severe and deeply tied to capital intensity. Building gigawatt-scale data centers requires immense upfront expenditure, with Nebius deploying approximately $2.5 billion in capital expenditures in the first quarter of 2026 alone. This aggressive spending compresses near-term margins and forces a perpetual reliance on debt and equity capital markets. A broader industry threat looms in the potential commoditization of compute power. If the supply of advanced chips eventually eclipses enterprise demand, pricing power will evaporate rapidly, leaving high-capex infrastructure providers structurally exposed and fighting in a race to the bottom on price.
New Products and Technological Growth Drivers
To stay ahead of the relentless obsolescence curve inherent to hardware, Nebius is rapidly deploying the vanguard of computational technology. The company has moved decisively beyond previous generations and is actively pre-ordering and rolling out Nvidia Blackwell architectures. These next-generation, liquid-cooled systems deliver a massive, discontinuous leap in parallel processing capabilities and are expected to serve as the foundational substrate for the next era of multimodality and agentic workflows. By securing early allocations for its new facilities in New Jersey, Missouri, and Europe, Nebius commands premium pricing and guarantees utilization from top-tier research labs. Beyond pure silicon, the Avride subsidiary is emerging as a powerful, differentiated technological driver. Rather than focusing exclusively on complex urban robotaxis, Avride utilizes a dual-platform strategy that harmonizes the software stack between sidewalk delivery robots and retrofitted autonomous passenger vehicles. This unique cross-pollination of sensor data allows the underlying artificial intelligence platform to learn from highly diverse edge cases across different modalities, accelerating the timeline for full commercial autonomy and deepening the strategic integration with global ride-hailing and delivery networks.
New Entrants and Disruptive Technologies
The lucrative economics of the artificial intelligence infrastructure boom have inevitably attracted highly capitalized new entrants seeking to disrupt the established neoclouds. Heavily funded sovereign wealth vehicles in the Middle East and major global energy conglomerates are beginning to fund their own purpose-built data centers, leveraging their direct access to captive power generation to bypass grid constraints. In Europe, well-funded regional startups like Nscale are attempting to replicate the Nebius sovereign cloud playbook. More importantly, a technological disruption is emerging at the silicon layer. A cohort of nimble semiconductor startups, alongside internal engineering teams at major hyperscalers, are designing application-specific integrated circuits tailored explicitly for inference workloads. These custom chips strip away the generalized architecture of standard graphics processing units, offering exponentially higher energy efficiency for specific, rigid tasks. If inference workloads become highly standardized, the industry could pivot away from the expensive, general-purpose chips that currently fill Nebius data centers. While the company's software orchestration layer provides some defense, a rapid industry shift toward specialized silicon architectures represents a credible, long-term disruptive threat to the entire infrastructure business model.
Management Track Record and Corporate Restructuring
Few management teams have successfully navigated a corporate crucible as severe as the one recently faced by Nebius. Founder and Chief Executive Officer Arkady Volozh built a formidable technology empire over two decades before geopolitical realities forced a radical corporate bisection. Volozh demonstrated exceptional strategic pragmatism by successfully divesting toxic domestic assets in mid-2024, decisively ring-fencing the international cloud, artificial intelligence, and autonomous mobility intellectual property. Crucially, management orchestrated the relocation of over a thousand elite engineers to Western hubs, preserving the institutional brain trust that serves as the company's ultimate competitive advantage. Following his removal from international sanctions lists, Volozh returned to the helm to provide vital operational stability. Retaining approximately 15 percent of the company equity and holding super-voting control, his financial incentives are inextricably aligned with minority shareholders. Under his strict guidance, the firm rapidly relisted on the public markets, secured multi-billion-dollar enterprise contracts, and pivoted the entire corporate structure toward global infrastructure with clinical precision. This track record of resilience, aggressive capital preservation, and flawless operational execution provides a robust foundation of institutional trust.
The Scorecard
Nebius Group N.V. represents one of the most compelling, aggressively managed assets in the global artificial intelligence infrastructure landscape. By leveraging a battle-tested engineering workforce and deep strategic ties with ultimate silicon supplier Nvidia, the company has successfully carved out a highly defensible niche as a premier neocloud operator. The clinical execution of its corporate restructuring allowed the firm to retain invaluable intellectual property and pivot entirely toward a service model that is currently experiencing hyper-demand. First-quarter 2026 results confirm that this strategy is yielding extraordinary top-line expansion, while monumental gigawatt-scale capacity build-outs across North America and Europe suggest that management holds immense conviction in the longevity of the compute cycle. The strategic optionality embedded within subsidiaries like Avride provides a secondary vector for disruptive mobility growth, fundamentally differentiating the firm from pure hardware lessors.
However, the structural realities of the neocloud business model mandate extreme analytical caution regarding capital intensity. The sheer volume of capital expenditure required to secure energy and populate data centers with next-generation clusters introduces significant balance sheet risk and perpetual margin pressure. Nebius is locked in an escalating capital arms race against domestic rivals like CoreWeave and the infinitely resourced legacy hyperscalers. If the broader market transitions into an environment of silicon oversupply, the resulting commoditization of compute could severely punish operators holding massive physical data center footprints. The ultimate success of the firm will depend on management's ability to seamlessly transition from hardware provisioning to higher-margin software inference, balancing the company's elite technological execution against the inherent financial fragility of front-loading billions in infrastructure costs.