Whether you're looking for an angel investor, a growth advisor, or just want to connect — I'm always open to great ideas.
Get in TouchAI, startups & growth insights. No spam.
TL;DR: On March 11, 2026, Nvidia announced a $2 billion strategic equity investment in Nebius Group, the Amsterdam-based AI cloud infrastructure company and Yandex spin-off, bringing Nebius's total outside funding to $2.7 billion. Nvidia will acquire shares at $94.94 each, taking an approximately 8.3% stake. The deal goes well beyond a check: Nebius and Nvidia will co-develop AI factory architecture, build a shared inference and agentic AI stack, and jointly pursue a target of deploying more than 5 gigawatts of AI compute capacity by 2030. Nebius stock surged 16% on the announcement. Coming days before Nvidia GTC 2026 (March 16–19), the investment is a clear signal that Nvidia's ecosystem strategy now runs through the neocloud operators it is financing to absorb the next wave of GPU supply.
The headline number — $2 billion — is the largest single strategic equity investment Nvidia has made in an AI cloud operator to date. The structure is a direct equity purchase: Nvidia agreed to buy Nebius shares at $94.94 per share, acquiring approximately 8.3% of the company. This is disclosed in an SEC filing, giving the terms a specificity that typical strategic partnership announcements often obscure.
The price per share matters. At the time of the announcement, Nebius shares were trading at roughly $94, meaning Nvidia purchased at approximately market price — not a discounted secondary transaction, but a commitment to current equity value. The announcement sent shares up 16% to $109.72, meaning Nvidia's stake was immediately worth more than the entry price.
This investment is not the first time Nvidia has written a check to Nebius. In December 2024, Nebius closed a $700 million fundraising round that included Nvidia as a participant alongside Accel Partners — at the time, Nvidia acquired approximately 0.5% of the company. That earlier entry was a small strategic bet. The March 2026 commitment is categorically different: an 8.3% stake at $2 billion is a significant financial commitment and a very public endorsement of Nebius's model.
| Metric | Detail |
|---|---|
| Investment amount | $2 billion |
| Price per share | $94.94 |
| Nvidia stake acquired | ~8.3% |
| Nebius stock reaction | +16% (to $109.72) |
| Previous Nvidia investment | $700M round (Dec 2024), ~0.5% stake |
| Total Nebius funding raised | ~$2.7 billion |
| Analyst 12-month price target | $147.45 (consensus) |
The total funding figure of $2.7 billion combines the December 2024 round and this investment. With $3.7 billion in cash and equivalents already on the balance sheet (as reported in Q4 2025 results), Nebius enters the partnership phase from a position of financial strength that few infrastructure startups of any age can claim.
Nebius's origin story is unusual enough to warrant explanation, because it directly shapes the company's engineering DNA and European positioning.
Nebius Group N.V. is the renamed successor to Yandex N.V., the Dutch holding company that was the parent of Russia's dominant search and technology conglomerate. After Russia's invasion of Ukraine in 2022, Yandex N.V. entered a prolonged and legally complex restructuring. In mid-2024, the company sold all of its Russian-based business assets to a consortium of local investors for approximately $5.4 billion. What remained in the Dutch entity were the international businesses that had operated outside Russia — including an AI cloud infrastructure division built on the engineering capabilities Yandex had developed over decades.
The company rebranded as Nebius Group N.V., led by CEO Arkady Volozh, the founder of Yandex who had relocated to Israel and had been removed from EU sanctions lists in 2024. In October 2024, Nebius resumed trading on the Nasdaq under the ticker NBIS, returning to public markets as an AI infrastructure company with approximately 1,000 former Yandex engineers.
The Yandex background is significant for a specific reason: Yandex had built its own GPU cloud infrastructure and machine learning platform for internal use at a scale comparable to mid-tier hyperscalers. Nebius inherited that engineering capability — the ability to design, deploy, and operate GPU clusters at scale — and oriented it toward the external market as a neocloud operator. Unlike companies that learned GPU cloud operations from scratch, Nebius started with years of institutional knowledge in AI infrastructure management.
Nebius's Q4 2025 financial results, reported in February 2026, established the business metrics that made the $2 billion investment make sense as a return-seeking commitment rather than purely a strategic hedge.
The headline numbers from Q4 2025:
For 2026, management guided $3.0–$3.4 billion in revenue with a group adjusted EBITDA margin of approximately 40%. The longer-range target — $7–9 billion ARR by end of 2026 — is aggressive but grounded in contracted hyperscaler agreements: Nebius has deals with Meta (fully deployed) and Microsoft (tranches underway, full run rate expected in 2027) valued at billions of dollars combined.
Critically, Nebius has been sold out of available capacity since Q3 2025 and entered 2026 already sold out for Q1. The constraint is not demand — it is the speed at which new data center capacity can be brought online. This capacity-constrained growth dynamic is precisely why Nvidia's investment carries strategic weight: it comes with early access to hardware that can fill new capacity the moment it comes online.
The company plans $16–20 billion of capital expenditure in 2026, funded by the cash on hand, operating cash flow, and presumably future debt financing. The scale of that planned CapEx — relative to the company's current revenue — signals the ambition and the risk simultaneously. Nebius is building a very large infrastructure company very quickly.
The $2 billion check is accompanied by a structured technology and operational partnership across four domains, as outlined in the joint press release from both companies:
1. AI Factory Design and Support Nebius gains access to Nvidia's partner design materials, design review processes, early hardware samples, system software support during bring-up, and ongoing technical reviews. In practical terms: Nebius gets to design and build AI factories to Nvidia's specifications and with Nvidia's direct engineering support. This compresses the time from hardware delivery to revenue-generating capacity.
2. Inference and Agentic AI Stack The companies will jointly develop what they describe as a best-in-class inference and agentic AI stack for developers and enterprises, using Nvidia's latest software technologies, optimized models, and libraries. This is the software layer — the services that run on top of the raw GPU infrastructure. As the market shifts from pure training workloads toward inference and agentic applications, having a co-developed software stack with Nvidia's latest APIs and optimization libraries is a differentiated offering.
3. AI Infrastructure Deployment Nebius will be an early adopter of Nvidia's next-generation computing architectures, including the Nvidia Rubin platform, Nvidia Vera CPUs, and Nvidia BlueField storage systems. Early access to hardware before general availability allows Nebius to offer capacity to customers on the newest generation while competitors are still waiting in queue. In a supply-constrained environment, this early access is a direct revenue advantage.
4. Fleet Management Nvidia will deploy its GPU health monitoring and software recommendation systems across Nebius's entire fleet. For a company managing thousands of GPUs across multiple geographies, automated fleet health optimization reduces downtime, extends hardware utilization, and lowers operational overhead. This is the least glamorous of the four pillars and arguably the most operationally important.
Nebius operates across multiple geographies, with a footprint that spans Europe and North America — positioning it as an infrastructure provider with cross-regional reach that most neoclouds lack.
Finland (Mäntsälä): Nebius's founding data center, located 60 kilometers from Helsinki. The facility houses a GPU supercluster and was the company's primary compute base during the transition from Yandex. Finland's political stability, access to renewable energy, and physical location near transatlantic fiber routes make it a natural anchor for European AI workloads.
France (Paris, Saint-Denis): A GPU cluster at Equinix's PA10 campus in the Saint-Denis district of Paris. This facility was among the first in the world to deploy Nvidia H200 GPUs at production scale, demonstrating the preferential hardware access that the Nvidia relationship facilitates.
United States (Kansas City, Missouri): Nebius has an existing colocation presence in a repurposed Kansas City data center and has announced plans for a 2.5 million square foot campus in Jackson County on approximately 400 acres — a facility capable of delivering 800 megawatts of AI compute capacity. Power delivery for the first phases is expected in the second half of 2026.
United States (Vineland, New Jersey): A 300-megawatt facility under construction via a partnership with DataOne, accelerating Nebius's US expansion beyond the Midwest.
Emerging locations: Nebius also has capacity in London, Israel, and Keflavik, Iceland — the last of which offers access to geothermal energy and sub-zero ambient temperatures that dramatically reduce data center cooling costs.
| Location | Status | Capacity |
|---|---|---|
| Mäntsälä, Finland | Operational | Existing supercluster |
| Paris, France (Equinix PA10) | Operational | H200 GPU cluster |
| Kansas City, Missouri | Existing operational + planned campus | Up to 800 MW (planned) |
| Vineland, New Jersey | Under construction | Up to 300 MW |
| London, UK | Leasing capacity | Undisclosed |
| Keflavik, Iceland | Active | GPU cluster |
The contracted power guidance stood at over 3 gigawatts as of Q4 2025 reporting, with a target of 800 MW to 1 GW of available (operational) data center capacity by year-end 2026. The 5 GW target by 2030 — shared with Nvidia as part of the partnership announcement — is the longer arc these infrastructure commitments are building toward.
The Nebius investment arrived three days after a comparably sized announcement from its most direct European rival: on March 9, 2026, UK-based Nscale closed a $2 billion Series C — the largest Series C ever closed in Europe — led by Aker ASA and 8090 Industries, valuing the company at $14.6 billion.
The timing was not a coincidence. The European AI infrastructure market is seeing a funding wave in early 2026 that parallels the US neocloud buildout of 2024–2025, driven by a combination of demand tailwinds (enterprises and governments seeking non-US AI infrastructure options) and supply tailwinds (Nvidia GPU availability improving as manufacturing scales).
How the two companies compare:
| Metric | Nebius (NBIS) | Nscale |
|---|---|---|
| Headquarters | Amsterdam, Netherlands | London, UK |
| Structure | Public (Nasdaq: NBIS) | Private |
| Total equity raised | ~$2.7 billion | >$4.5 billion across all rounds |
| Latest round size | $2B (Nvidia strategic) | $2B (Series C) |
| Valuation | ~$24 billion (implied at $109.72/share) | $14.6 billion |
| Nvidia relationship | Strategic equity partner (8.3% stake) | Investor in Nscale's Series C |
| Key investors | Nvidia, Accel | Aker ASA, Nvidia, Citadel, Dell |
| Geographic focus | Europe + North America | Europe + global |
| Notable contracts | Meta (deployed), Microsoft | Undisclosed enterprise |
One crucial difference: Nebius is public and Nscale is not. The Nebius-Nvidia relationship is also fundamentally different in depth — Nvidia holding 8.3% of a public company with co-designed AI factory architecture is a more committed relationship than Nvidia participating as one of many investors in a Series C. For Nebius, Nvidia is the strategic partner. For Nscale, Nvidia is a check writer among many.
Both companies are building toward the same destination: multi-gigawatt European-anchored AI infrastructure. The question is whether the market can sustain two well-capitalized competitors at this scale, or whether consolidation is inevitable as the AI training market eventually reaches capacity saturation.
To understand the $2 billion Nebius investment, it helps to understand Nvidia's broader problem: it is manufacturing GPUs faster than the hyperscalers can deploy them.
Amazon, Microsoft, and Google are buying enormous quantities of Nvidia hardware. But their procurement and deployment cycles are governed by internal processes, regulatory approvals for data center construction, and the constraint that each company must deploy infrastructure globally across dozens of regions. Hyperscalers do not move at startup speed.
Neocloud operators — Nebius, CoreWeave, Lambda Labs, and others — exist to fill the gap. They move faster, focus exclusively on GPU cloud infrastructure, and can deploy new capacity in months rather than years. They buy Nvidia GPUs and resell the compute as cloud services to AI companies, researchers, and enterprises that need GPU access without building their own data centers.
From Nvidia's perspective, neoclouds are an important additional channel for GPU sales. Every dollar of neocloud infrastructure is GPU-denominated. The faster neoclouds can finance and deploy capacity, the more GPUs Nvidia sells. By investing $2 billion in Nebius, Nvidia is not just making a financial bet — it is effectively providing growth capital that will be recycled back into GPU purchases.
The early hardware access component of the partnership amplifies this dynamic. When Nebius gets early samples of the Rubin platform and Vera CPUs, it deploys them immediately in revenue-generating capacity. This creates real-world demand signal for Nvidia's next-generation hardware before it reaches general availability — accelerating Nvidia's own platform adoption curve.
Nebius occupies a specific and valuable position in the European AI infrastructure conversation: a Nasdaq-listed company headquartered in Amsterdam, operating under Dutch and US regulatory oversight, with substantial compute presence in Finland and France.
This matters because the EU AI Act, GDPR, and a growing body of European data residency requirements are creating genuine demand for AI compute infrastructure that is not hosted by American hyperscalers. European enterprises and governments processing sensitive data — healthcare records, financial data, citizen services — face legal and reputational pressure to keep that data within European jurisdiction and preferably on infrastructure operated by entities subject to European law.
Nebius is positioning itself as the answer to this demand. Its Finnish and French data centers are geographically and legally European. Its corporate structure under Dutch law, combined with Nasdaq listing and US regulatory transparency, gives it a credibility profile that a purely European startup lacks. It can serve European sovereign AI requirements while offering the capital markets access and financial reporting standards that large institutional buyers expect.
The lifting of EU sanctions on Arkady Volozh in 2024 resolved the major reputational obstacle to this positioning. With that resolved, Nebius can market itself in Brussels and Frankfurt without the Yandex-Russia association creating deal-killing hesitation.
Nscale is pursuing a similar angle from the UK. But Nebius's head start in Finland and France, combined with the Nvidia endorsement, gives it a stronger claim on the core European market in the near term.
The announcement landed on March 11, 2026 — five days before Nvidia GTC 2026 (March 16–19, San Jose, California), Jensen Huang's annual showcase for Nvidia's platform roadmap and ecosystem partnerships.
This timing is unlikely to be coincidental. GTC is where Nvidia contextualizes its technology investments for developers, enterprise customers, and investors. Announcing the Nebius investment the week before GTC effectively sets a narrative: Nvidia is not just building chips, it is building the cloud operators who will deploy those chips at scale.
GTC 2026 is themed around the "Agentic AI Inflection Point" and will feature the formal debut of the Vera Rubin architecture — the platform Nebius will be an early adopter of under the new partnership. Other partnership announcements at or around GTC include a strategic deal with Thinking Machines Lab to deploy at least one gigawatt of Vera Rubin systems. The Nebius deal and the Thinking Machines deal together represent Nvidia's strategy of creating a network of large-scale infrastructure operators committed to early deployment of each successive GPU generation.
Jensen Huang's keynote will also address the full stack — chips, software, models, and applications — which directly maps to the four partnership pillars Nvidia and Nebius announced jointly. The Nebius deal is not a sideshow to GTC: it is an illustration of the ecosystem model Nvidia intends to describe on stage.
The Nebius-Nvidia deal, set alongside the Nscale $2 billion Series C just days earlier, marks a structural moment in the AI infrastructure market.
Neocloud operators are becoming a primary GPU deployment channel. The era of only hyperscalers deploying GPU infrastructure at scale is ending. Neocloud operators with direct Nvidia relationships, preferential hardware access, and billions in capital are now capable of building gigawatt-scale AI factories. The total addressable market for GPU cloud services — previously dominated by AWS, Azure, and GCP — is fragmenting.
Nvidia's investment strategy is about supply chain certainty as much as financial return. By taking 8.3% of Nebius, Nvidia creates a partner with both the capital and the incentive to deploy Nvidia hardware as fast as it can be manufactured. The return on the $2 billion is partially financial (if Nebius scales as projected) and partially strategic (guaranteed absorption of GPU supply into revenue-generating capacity).
European AI infrastructure is investable at scale. The same week saw $4 billion committed to European-anchored AI infrastructure companies between Nebius and Nscale. This suggests the geographic diversification of AI compute — long discussed as a regulatory and strategic necessity — is now happening at capital scale.
The inference and agentic AI layer is becoming a battleground. The partnership's explicit focus on building a joint inference and agentic AI stack signals that raw GPU rental is not the long-term business model. The margin is in the software layer — in providing optimized inference APIs, model serving infrastructure, and agentic workflow tooling on top of the compute. Nebius and Nvidia are building that together, which is a more defensible business than renting H100s by the hour.
For enterprises making AI infrastructure decisions in 2026, the Nebius-Nvidia partnership offers something that pure commodity GPU clouds cannot: a co-engineered stack from the silicon to the API layer, with a cloud provider financially committed to deploying the latest generation hardware before anyone else can access it.
Nebius Group N.V. is an Amsterdam-based AI cloud infrastructure company, incorporated in the Netherlands and publicly traded on the Nasdaq under the ticker NBIS. It was formed from the international operations of Yandex N.V. after that company sold all of its Russian business assets in 2024 and rebranded. The company employs approximately 1,000 engineers, many of them former Yandex employees with deep AI infrastructure experience.
Nvidia agreed to purchase Nebius shares at $94.94 per share, for a total commitment of $2 billion, acquiring approximately 8.3% of the company's outstanding shares. The terms are disclosed in an SEC filing.
Before the Nvidia $2 billion investment, Nebius had raised approximately $700 million in a December 2024 round that included Nvidia (at ~0.5% stake) and Accel Partners. Combined with the March 2026 investment, total outside equity raised reaches approximately $2.7 billion.
As of Q4 2025 financial results (reported February 12, 2026), Nebius had $3.7 billion in cash and equivalents, in addition to generating $834 million in operating cash flow in Q4 alone.
Q4 2025 group revenue was $228 million, up 547% year-over-year. Core AI cloud annual recurring revenue reached $1.25 billion by end of 2025, up from $90 million at end of 2024. The company is guiding $3.0–3.4 billion in revenue for full-year 2026.
Nebius will be an early adopter of multiple upcoming Nvidia platforms, including the Nvidia Rubin platform (the successor to Blackwell), Nvidia Vera CPUs, and Nvidia BlueField storage systems. Early access means Nebius can offer this hardware to customers before it is generally available.
Nvidia and Nebius announced a joint objective for Nebius to deploy more than 5 gigawatts of AI compute capacity by the end of 2030. As of Q4 2025, Nebius had contracted power of over 3 GW and was targeting 800 MW to 1 GW of operational capacity by year-end 2026.
Nebius operates data centers in Mäntsälä, Finland (existing supercluster), Saint-Denis, Paris (Equinix PA10 campus, H200 GPUs), and has major US construction projects in Kansas City, Missouri (up to 800 MW, 400 acres) and Vineland, New Jersey (up to 300 MW). Additional capacity is in London, Iceland (Keflavik), and Israel.
Both are neocloud operators focused on GPU cloud infrastructure, but they differ in several ways. CoreWeave is US-based and private; Nebius is Europe-headquartered and publicly traded. Nebius has significant data center presence in the EU, which positions it for European sovereign cloud demand. CoreWeave has been primarily US-focused and has a deeper relationship with Microsoft as an anchor tenant. Both have direct Nvidia relationships, but Nvidia's 8.3% equity stake in Nebius is a deeper commitment than what has been disclosed in the CoreWeave relationship.
Nscale is a UK-based AI cloud infrastructure company that raised a $2 billion Series C on March 9, 2026 — just two days before the Nebius-Nvidia announcement — valued at $14.6 billion. Nscale has raised over $4.5 billion total across all rounds but remains private. Like Nebius, it is pursuing European AI infrastructure with sovereign cloud positioning. Nvidia participated in Nscale's Series C as one investor among many, a significantly smaller commitment than the 8.3% strategic stake it took in Nebius.
Nebius shares surged 16% on March 11, 2026, rising to $109.72 from pre-announcement levels around $94. The analyst consensus 12-month price target is $147.45. Nebius's 52-week range spans $18.31 to $141.10, reflecting the dramatic re-rating the company has undergone as it transitioned from a Yandex restructuring story to a credible AI infrastructure growth story.
Nvidia manufactures GPUs and needs channels to deploy them into revenue-generating use at scale and speed. Neocloud operators like Nebius are dedicated to deploying GPU infrastructure and can move faster than hyperscalers, which have slower internal processes and broader internal capital competition. By investing in Nebius, Nvidia creates a well-capitalized partner with both the incentive and the capability to absorb large GPU orders and deploy them immediately into revenue-generating capacity — accelerating the feedback loop between Nvidia's manufacturing and revenue from GPU sales.
Nvidia GTC 2026 (March 16–19, San Jose) is Nvidia's flagship ecosystem event for the year, themed around the agentic AI era and featuring the formal debut of the Vera Rubin architecture. The Nebius announcement landing five days before GTC positions Nebius as a showcase example of Nvidia's neocloud ecosystem strategy — a partner that will deploy the Vera Rubin platform at scale as an early adopter, giving Jensen Huang a concrete illustration of how Nvidia's technology goes from chip to cloud.
The Nvidia Rubin platform is the next-generation GPU architecture succeeding Blackwell, expected to enter production in 2026–2027. As a partner under the new agreement, Nebius gains early access to samples and bring-up support — meaning it can integrate and test Rubin hardware before general market availability. This gives Nebius a first-mover advantage in offering Rubin-class compute to customers who need the latest generation for frontier model training and inference.
No. Nebius's business is currently built almost entirely on Nvidia GPUs — as is most of the neocloud industry — but there is no disclosed exclusivity clause. The partnership enhances Nebius's preferential access to Nvidia hardware, but Nebius remains free to evaluate hardware from AMD or other vendors. In practice, Nvidia's 8.3% equity stake and the deep technical co-development relationship create strong incentives for Nebius to remain an Nvidia-first operator.
Nebius operates data centers in Finland and France, both EU member states, under corporate governance subject to Dutch law and US Nasdaq listing requirements. The EU AI Act and GDPR create genuine demand from European enterprises and governments for AI compute infrastructure that stays within EU legal jurisdiction. Nebius's European infrastructure, combined with its Nasdaq-level financial transparency and Nvidia endorsement, positions it as a credible sovereign cloud option for European institutional buyers who cannot or will not use US hyperscaler infrastructure.
Arkady Volozh, the founder of the original Yandex, serves as Nebius's CEO. He relocated from Russia to Israel following the invasion of Ukraine and was removed from EU sanctions lists in 2024 — resolving a significant reputational obstacle for Nebius's European business development. Volozh has described Nebius as the company he wanted to build after the constraints of operating as a Russian internet company were lifted.
For enterprises evaluating AI cloud providers, the Nebius-Nvidia partnership means early access to the latest Nvidia hardware, a co-developed inference and agentic AI stack optimized for Nvidia's software ecosystem, and a cloud provider with Nvidia's direct engineering support during data center bring-up. Enterprises that are building on Nvidia's CUDA platform — which is most enterprise AI buyers — benefit from infrastructure partners who have first access to Nvidia's latest tools, libraries, and hardware generations.
As of March 2026, consensus analyst price target is $147.45, representing approximately 34% upside from the post-announcement trading price of ~$109. With $3.7 billion in cash, $3–$3.4 billion in guided 2026 revenue, a 40% EBITDA margin target, and sold-out capacity through Q1 2026, the fundamental business is executing. The primary risks are execution risk on the aggressive capital expenditure plan ($16–20 billion in 2026), the concentration of revenue in a small number of hyperscaler customers (Meta and Microsoft), and the pace at which new data center capacity can be brought online to meet demand.
Meta plans four new MTIA chips through 2027, aiming to cut AI compute costs by up to 30% and reduce reliance on Nvidia hardware.
Larry Fink warns America is running out of electricians. BlackRock's $100 million Future Builders Initiative aims to train 50,000 skilled tradespeople over five years to build the physical infrastructure AI depends on.
Legal AI platform Legora closes a $550 million Series D led by Accel at a $5.55 billion valuation — the largest legal tech funding round ever, signaling that vertical AI platforms are winning the enterprise race.