Nebius
AI-infrastructure company building a full-stack cloud for developers and enterprises.
Website: https://nebius.com/
Cover Block
Open sources
| Field | Value |
|---|---|
| Name | Nebius |
| Tagline | AI-infrastructure company building a full-stack cloud for developers and enterprises. [Crunchbase] |
| Headquarters | Amsterdam, Netherlands [Crunchbase] |
| Founded | 2022 [Crunchbase] |
| Stage | Public [TechCrunch, November 2024] |
| Business model | API / Developer Platform |
| Industry | Deeptech |
| Technology | AI / Machine Learning |
| Geography | Global / Remote-First |
| Growth profile | Venture Scale |
| Founding team | Corporate Spinout |
| Funding label | $100M+ |
| Total disclosed funding | ~$7,815,000,000 [Nebius, December 2024] [Reuters, March 2026] [Nebius, July 2026] |
Links
Open sources
- Website: https://nebius.com/
- X / Twitter: http://nebius.com/careers
What an Investor Needs First
PUBLIC Nebius is building an AI-focused cloud platform for developers and enterprises, and it merits attention now because it has paired unusual capital access with unusually large disclosed customer commitments for a company only formed in 2022 [Crunchbase] [Reuters, March 2026] [Nebius, July 2026]. The company emerged from the restructuring of Yandex N.V., is headquartered in Amsterdam, trades on Nasdaq under NBIS, and is led by Yandex co-founder Arkady Volozh, whose return to active company-building followed his 2024 removal from the EU sanctions list and his 2026 renunciation of Russian citizenship [TechCrunch, November 2024] [The New York Times, March 2024] [Bloomberg, February 2026].
The product thesis is straightforward: Nebius is not trying to be a general-purpose cloud, but a purpose-built AI infrastructure stack spanning GPU compute, networking, storage, managed services, and model-serving, with public product signals that include per-second GPU billing, large-cluster NVIDIA deployments, and the Token Factory inference platform [nebius.com] [docs.nebius.com] [NVIDIA Newsroom, August 2026]. That positioning matters because the company is leaning into owned capacity as well as cloud abstraction, including a disclosed plan for a Pennsylvania AI factory with up to 1.2 GW of secured power and land, a scale marker that places execution and capital discipline at the center of the underwriting case [SEC filing, 2026].
The leadership story is one of experienced operators rather than first-time founders: Volozh co-founded Yandex in 1997, John Boynton has served as chairman of the board since 2016, and Marc Boroditsky is identified publicly as chief revenue officer [Reuters, July 2024] [nebius.com, Retrieved 2026] [LinkedIn, Retrieved 2026]. The financing profile is equally distinctive, with a $700 million strategic equity raise in December 2024 involving Accel, NVIDIA, and Orbis-managed accounts, followed by a $4.34 billion debt raise reported by Reuters, a reported $2 billion NVIDIA investment, and a $775 million secured debt facility in July 2026 [Nebius, December 2024] [Reuters, March 2026] [Nebius, March 2026] [Nebius, July 2026].
What bears watching over the next 12 to 18 months is less product launch cadence than conversion of contracted demand into durable economics. Reuters reported 2026 capital expenditure plans of $16 billion to $20 billion and management's expectation that customer prepayments would fund about 60% of growth, while Nebius has said it has more than $40 billion of additional contracted revenue from investment-grade customers such as Microsoft and Meta, claims that make delivery risk, counterparty concentration, and financing structure more important than category demand itself [Reuters, March 2026] [Nebius, July 2026].
Partially corroborated -- Core company identity, funding events, and leadership are corroborated by named public sources, but several material operating claims in this summary, especially contracted revenue and some product specifics, rely partly on company disclosures.
Taxonomy Snapshot
| Axis | Value |
|---|---|
| Stage | Public |
| Business Model | API / Developer Platform |
| Industry / Vertical | Deeptech |
| Technology Type | AI / Machine Learning |
| Geography | Global / Remote-First |
| Growth Profile | Venture Scale |
| Founding Team | Corporate Spinout |
| Funding | $100M+ (total disclosed ~$7,815,000,000) |
Inside the Company
PUBLIC
Nebius presents as an unusual public-market AI infrastructure company because its operating identity is tied less to a conventional startup origin story than to a corporate restructuring. The company says it was created from the post-Yandex restructuring of the former Yandex N.V. business, is headquartered in Amsterdam, and trades on Nasdaq under the ticker NBIS [Crunchbase] [nebius.com]. Crunchbase lists the company as founded in 2022 and describes it as an AI-centric cloud business focused on NVIDIA-powered infrastructure [Crunchbase].
The public chronology that can be supported from company and database sources is still fairly compact. Crunchbase places the founding in 2022 [Crunchbase]. By December 2024, Nebius was publicly framing itself as a full-stack AI infrastructure provider and announced a $700 million strategic equity financing to accelerate rollout of that platform [Nebius, December 2024]. In May 2025, the company said it participated in ClickHouse's Series C and described that move as part of funding the hypergrowth of its core AI infrastructure business [Nebius, May 2025].
The next milestones are about capital formation at unusual scale for a recently formed platform company. In September 2025, Nebius published a financing update that referenced broader funding diversification [Nebius, September 2025]. By July 2026, the company announced a $775 million first secured debt financing to accelerate global buildout, a sign that the story had moved from launch and repositioning into asset-heavy expansion [Nebius, July 2026]. The legal entity structure is not fully laid out in the provided public-source set, so the cleaner reading is that Nebius is best understood here as an Amsterdam-headquartered, Nasdaq-listed AI cloud platform emerging from a larger corporate separation rather than as a de novo venture-backed startup [nebius.com] [Crunchbase].
Partially corroborated -- Confirmed primarily by Crunchbase and Nebius company materials; legal entity detail is only partially established from the cited public sources.
Under the Hood
Platform scope
MIXED Nebius is positioning itself as a full-stack AI cloud rather than a narrow GPU reseller, and the public record is reasonably clear on the broad shape of that stack even when product-level detail remains company-led. The company describes the platform as covering model training, data processing, production inference, and deployment for developers and enterprises [nebius.com] [TechCrunch, November 2024]. Its own site also points to accelerated compute that ranges from single-node instances to thousand-GPU clusters, built on NVIDIA networking, with on-demand access to GPU families including GB300, B300, GB200, B200, H200, H100, RTX PRO 6000, and L40S [nebius.com].
That description is partially corroborated by public usage and infrastructure details outside the marketing layer. Nebius documentation says running GPU virtual machines are billed in one-second increments [docs.nebius.com], and secondary coverage indicates the company bills per second for GPU usage and gives a 60-second warning before forced shutdown of spot instances [gpufinder.dev] [btw.media]. Taken together, that suggests a product aimed at both bursty developer workloads and larger reserved or contracted enterprise demand, though the operational characteristics beyond those pricing and interruption mechanics are still mostly described by the company itself [docs.nebius.com] [Nebius, July 2026].
Differentiation and technical posture
MIXED The more interesting technical claim is not simply access to GPUs, but Nebius's effort to pair owned capacity with managed services and specialized inference infrastructure. Public company and press materials describe a strategy that combines owned AI factories, networking, storage, managed services, and model-serving endpoints [Nebius, August 2026] [nebius.com]. In August 2026, NVIDIA said Nebius Token Factory was the first AI cloud to adopt NVIDIA Groq 3 LPX for Vera Rubin NVL72 systems, framing the deployment as a way to raise inference token generation rates [NVIDIA Newsroom, August 2026]. That is one of the cleaner externally verified datapoints showing Nebius trying to differentiate on inference performance, not only raw training capacity.
Nebius has also used M&A to widen the platform surface. Bloomberg reported in February 2026 that the company agreed to acquire Tavily for $275 million to add agentic search capabilities to the Nebius AI Cloud platform [Bloomberg, February 2026]. The website separately highlights use cases such as training biology foundation models and running drug discovery pipelines, which signals vertical ambition, but those workflow claims should still be treated as company-described capability rather than independently validated market adoption [nebius.com].
Claim stands unchecked -- Core product capabilities are primarily confirmed by Nebius's website and newsroom, with limited independent corroboration from TechCrunch, NVIDIA Newsroom, docs.nebius.com, gpufinder.dev, btw.media, and Bloomberg.
Market Research
PUBLIC
The market matters now because Nebius is selling into a narrow but rapidly expanding layer of the AI stack, where capital availability, power access, and enterprise demand for dedicated GPU capacity appear to be moving faster than the supply of deployable infrastructure [Reuters, March 2026] [SEC filing, 2026].
Public evidence in the source set does not support a clean TAM, SAM, or SOM build from named third-party market studies, so the more defensible approach is to anchor market context in observed demand signals around AI compute procurement rather than force a synthetic sizing model. On that standard, Nebius's disclosed posture is notable: Reuters reported in March 2026 that the company planned $16 billion to $20 billion of capital expenditures for 2026 and expected customer prepayments to fund 60% of growth, principally from Microsoft and Meta [Reuters, March 2026]. That level of planned spend, paired with customer-backed financing, suggests a market where large buyers are willing to commit capital in advance to secure capacity, which is closer to infrastructure reservation behavior than ordinary cloud consumption [Reuters, March 2026].
The strongest tailwind in the public record is concentration of demand among investment-grade counterparties that need large clusters, long-duration commitments, and speed to deployment. Nebius said in July 2026 that it had more than $40 billion of additional contracted revenue from investment-grade customers including Microsoft and Meta, although that figure remains company-reported and should be treated cautiously until reconciled across filings [Nebius, July 2026]. Separately, the company's first-quarter 2026 materials disclosed that it had secured up to 1.2 GW of power and land for a new owned AI factory in Pennsylvania, which points to the other defining constraint in this market: power interconnection and site readiness now matter almost as much as GPU sourcing [SEC filing, 2026].
Adjacent markets are also relevant because Nebius is not competing only in raw compute rental. Its public materials position the platform across training, inference, storage, networking, and managed services, while Bloomberg reported that Nebius agreed to acquire Tavily for $275 million to add agentic search to its cloud platform [Bloomberg, February 2026]. That broadens the addressable surface from infrastructure-as-a-service toward AI platform services, but it also places Nebius next to substitute options such as hyperscaler GPU instances, enterprise on-prem deployments, and model-layer vendors bundling inference with application tooling [nebius.com] [Bloomberg, February 2026].
Regulation and macro conditions cut both ways. The supportive side is straightforward: sovereign and enterprise sensitivity around where AI workloads run can favor providers that can offer dedicated capacity and clearer contractual commitments. The constraint side is more practical than legal in the current source base, namely debt-funded expansion, customer concentration, and exposure to semiconductor and electricity supply cycles; Reuters' reporting on multi-billion-dollar debt and capex plans makes clear that this is a market where financing conditions can shape share outcomes as much as product design can [Reuters, March 2026]. Founder history also remains part of the institutional diligence context, given Arkady Volozh's prior sanctions episode and subsequent removal from the EU list, though the public record indicates those restrictions were lifted in March 2024 [The New York Times, March 2024].
| Cited market signal | Value | What it suggests |
|---|---|---|
| Planned 2026 capex | $16B to $20B | AI infrastructure demand may justify utility-scale buildout when capacity is pre-contracted [Reuters, March 2026] |
| Growth expected to be funded by customer prepayments | 60% | Large customers may be underwriting supply expansion directly [Reuters, March 2026] |
| Power and land secured for Pennsylvania AI factory | Up to 1.2 GW | Power access is a gating factor in AI cloud supply [SEC filing, 2026] |
| Additional contracted revenue from investment-grade customers | More than $40B | Enterprise commitment levels may be unusually large, though this figure is company-reported [Nebius, July 2026] |
Taken together, the table points to a market defined less by abstract software TAM and more by scarce physical inputs, power, financing, and buyer willingness to sign forward contracts. That is attractive if demand holds, but it means market sizing should be read through infrastructure bottlenecks rather than standard cloud multiples.
Reasoned from indirect evidence -- Section relies on Reuters and SEC materials for corroborated market signals, but lacks independent third-party TAM research and includes one material company-reported demand figure from Nebius.
Competition and Substitutes
MIXED Nebius is positioned less as a general-purpose cloud and more as a capital-intensive AI-compute specialist, which places it in practical competition with hyperscalers for large training and inference budgets, and with GPU cloud specialists for developer workloads [TechCrunch, November 2024] [nebius.com].
At the top end, the relevant incumbents are the hyperscalers, especially Microsoft, because enterprise buyers already purchase compute, storage, networking, and managed services from them at scale, and because Microsoft appears in Nebius's public materials as both a major customer and a source of contracted demand [Nebius, July 2026] [The Motley Fool, October 2026]. That creates a complicated posture: Nebius can win as a specialist supplier into the AI buildout, but it also sits beside much larger platforms that can bundle adjacent services and absorb lower margins for strategic accounts [Reuters, March 2026].
Below that tier, the challengers are specialist AI clouds and GPU rental platforms, even if the company has not named them directly in the source set. Nebius's own product posture, GPU-dense infrastructure, thousand-GPU clusters, purpose-built networking, and per-second billing, suggests the comparison set is providers competing on access, performance, and time-to-capacity rather than on broad enterprise software suites [nebius.com] [docs.nebius.com] [gpufinder.dev]. Adjacent substitutes include on-premise deployments and direct procurement from NVIDIA-based infrastructure partners, especially for customers large enough to justify reserved capacity or custom buildouts. Nebius's push into owned AI factories, including up to 1.2 GW of power and land secured in Pennsylvania, indicates management believes control of physical infrastructure is becoming a primary competitive variable, not just a procurement detail [SEC filing, 2026].
Nebius does have a visible edge today, but it is an edge tied to capital access and supply aggregation more than to software lock-in. The company has assembled an unusually large financing stack for an AI-cloud specialist, including a $700 million strategic equity financing in December 2024, a reported $4.34 billion debt raise in March 2026, a $2 billion NVIDIA investment announced the same month, and $775 million of secured debt in July 2026 [Nebius, December 2024] [Reuters, March 2026] [Nebius, March 2026] [Nebius, July 2026]. That matters because the near-term market is constrained by access to GPUs, power, and balance sheet capacity. Its differentiated position is reinforced by product choices that matter to developers, including per-second GPU billing and very large cluster configurations, plus NVIDIA-aligned positioning such as being the first AI cloud to adopt NVIDIA Groq 3 LPX in its Token Factory offering [docs.nebius.com] [NVIDIA Newsroom, August 2026]. Still, this edge looks perishable unless contract conversion and facility rollout keep pace with financing. Capital can buy a window; it does not by itself create durable customer captivity.
The company is most exposed where a larger platform controls the customer relationship and can internalize the same workload. Microsoft is the clearest named example because Nebius's disclosed demand profile appears meaningfully tied to investment-grade counterparties such as Microsoft and Meta, and because Reuters reported that customer prepayments were expected to fund 60% of growth [Nebius, July 2026] [Reuters, March 2026]. If those buyers decide over time that equivalent capacity should sit inside their own cloud or procurement stack, Nebius faces both pricing pressure and concentration risk. The same logic applies on the product side. Nebius can compete on raw AI infrastructure and managed services, but the source set does not show evidence that it owns a broad application platform, enterprise distribution channel, or proprietary workflow layer that would keep customers from shifting workloads once compute becomes less scarce [TechCrunch, November 2024] [nebius.com].
Over the next 18 months, the most plausible competitive scenario is bifurcation rather than winner-take-all. Microsoft looks like the winner if enterprise AI demand remains concentrated in a handful of very large buyers who prefer bundled procurement, integrated software, and balance-sheet certainty, because it can purchase external capacity while still defending the primary customer interface [Reuters, March 2026] [Nebius, July 2026]. Nebius looks like the winner if GPU scarcity, power bottlenecks, and time-to-deployment remain the binding constraints, because its financing cadence and owned-capacity strategy are designed for exactly that environment [SEC filing, 2026] [Reuters, March 2026]. The likely loser if supply normalizes faster than expected is any specialist AI-cloud provider whose differentiation rests mainly on access rather than on software or distribution. Nebius is not exempt from that risk, even with its current scale of disclosed financing.
Opportunity
PUBLIC
The prize here is unusually large: if Nebius executes on its current buildout, it has a plausible path to become one of the few scaled independent AI clouds outside the incumbent hyperscalers, with enough contracted demand to matter at public-market size [Reuters, March 2026] [Nebius, July 2026].
The headline opportunity is not simply selling GPU time. It is becoming a specialized capacity layer for model builders and large enterprises that want frontier compute, managed infrastructure, and contract-grade supply without relying exclusively on one general-purpose cloud [TechCrunch, November 2024] [nebius.com]. That outcome looks reachable, not merely aspirational, because the public record already shows three ingredients that rarely appear together this early: large financing capacity, including a $700 million strategic equity round in December 2024 and a $4.34 billion debt raise in March 2026, named strategic relationships with NVIDIA, and multi-year customer commitments tied to Microsoft and Meta in company and media reporting [Nebius, December 2024] [Reuters, March 2026] [Nebius, July 2026]. The infrastructure footprint also appears to be moving beyond resale economics, with Nebius disclosing up to 1.2 GW of power and land for an owned AI factory in Pennsylvania, which is the sort of asset base required if the company intends to serve sustained, very large training and inference demand rather than periodic burst usage [SEC filing, 2026].
The upside breaks into a few distinct paths, all of which depend on the same underlying premise: that reliable supply and vertically integrated service matter enough in AI infrastructure to support a non-hyperscaler winner [Reuters, March 2026] [NVIDIA Newsroom, August 2026].
| Scenario | What happens | Catalyst | Why it's plausible |
|---|---|---|---|
| Anchor-tenant utility | Nebius becomes a preferred capacity partner for a small set of very large investment-grade buyers, monetizing dedicated GPU and data center buildouts under long-duration contracts. | The reported Microsoft and Meta commitments, plus customer-prepayment funding dynamics, turn capex into contracted supply faster than the market expects [Reuters, March 2026] [Nebius, July 2026]. | Reuters reported that Nebius expected 60% of growth to be funded through customer prepayments, and Nebius said it had more than $40 billion of additional contracted revenue from investment-grade customers such as Microsoft and Meta [Reuters, March 2026] [Nebius, July 2026]. |
| Full-stack AI cloud | Nebius moves from raw infrastructure into a higher-value platform for training, inference, model serving, and managed workloads, improving wallet share per customer. | Product expansion around thousand-GPU clusters, managed services, and Token Factory pulls customers up the stack [nebius.com] [NVIDIA Newsroom, August 2026]. | Nebius publicly markets accelerated compute from single-node instances to thousand-GPU clusters, and NVIDIA said Nebius Token Factory was the first AI cloud to adopt NVIDIA Groq 3 LPX for Vera Rubin NVL72 inference performance [nebius.com] [NVIDIA Newsroom, August 2026]. |
| Agentic infrastructure layer | Nebius uses the Tavily acquisition and model-serving infrastructure to position itself as a cloud substrate for agentic applications, not only model training. | The Tavily acquisition broadens the product from compute toward search-enabled AI application infrastructure [Bloomberg, February 2026] [Nebius]. | Bloomberg reported Nebius agreed to acquire Tavily for $275 million, and Nebius said the deal was aimed at bringing agentic search into its AI cloud platform [Bloomberg, February 2026] [Nebius]. |
What compounding looks like is fairly straightforward, and that is part of the appeal. Large customer contracts support prepayments and debt capacity, debt and equity fund more owned infrastructure, owned infrastructure can improve service levels and economics, and better performance then helps Nebius win the next large workload [Reuters, March 2026] [SEC filing, 2026]. The company is already hinting at the beginnings of that loop: Reuters tied the 2026 capex plan of $16 billion to $20 billion to committed customer demand and prepayments, while Nebius and NVIDIA have both highlighted differentiated infrastructure around next-generation NVIDIA systems and high-performance inference [Reuters, March 2026] [NVIDIA Newsroom, August 2026]. If that loop holds, Nebius does not need to win every developer. It needs to become a trusted home for a concentrated but expanding set of very large AI workloads.
The size of the win is easier to frame than to pin down. Reuters reported a $4.34 billion debt raise on top of earlier financing, and Nebius has also disclosed very large contracted-demand figures, though those remain partly company-reported and should be treated carefully [Reuters, March 2026] [Nebius, July 2026]. A conservative way to express upside is this: if Nebius proves it can convert those contracts into durable infrastructure revenue and becomes one of the market's few scaled independent AI clouds, it could plausibly justify an enterprise value measured in the tens of billions of dollars (scenario, not a forecast). That is not a formal valuation model, but the public inputs support the shape of the opportunity: multi-billion-dollar capital formation, hyperscale-class customer commitments, and physical power access at GW scale are the markers of a company pursuing category-level infrastructure, not a narrow software niche [Nebius, December 2024] [Reuters, March 2026] [SEC filing, 2026].
Partially corroborated -- This section relies on Reuters and SEC materials for financing, capex, and infrastructure capacity, but several material upside claims, including contracted revenue figures and parts of the product roadmap, remain company-sourced.
Sources
Open sources
[Crunchbase] AI-centric cloud powered by the latest NVIDIA GPUs | https://www.crunchbase.com/organization/nebius/growth_outlook
[TechCrunch, November 2024] The curious case of Nebius, the publicly traded AI infrastructure 'startup' | https://techcrunch.com/2024/11/24/the-curious-case-of-nebius-the-publicly-traded-ai-infrastructure-startup/
[Nebius, December 2024] Nebius announces oversubscribed strategic equity financing of USD 700 million to accelerate roll-out of full-stack AI infrastructure. | https://nebius.com/newsroom/nebius-announces-oversubscribed-strategic-equity-financing-of-usd-700-million-to-accelerate-roll-out-of-full-stack-ai-infrastructure
[The New York Times, March 2024] Russian Tycoon Arkady Volozh Is Removed From E.U. Sanctions List | https://www.nytimes.com/2024/03/13/world/europe/arkady-volozh-russia-eu-sanctions.html
[Bloomberg, February 2026] Bloomberg Billionaires Index - Arkady Volozh | https://www.bloomberg.com/billionaires/profiles/arkady-y-volozh/
[nebius.com] The Ultimate AI Cloud | https://nebius.com/
[SEC filing, 2026] Nebius reports first quarter 2026 financial results. | https://www.sec.gov/Archives/edgar/data/1513845/000110465926059872/tm2614392d1_ex99-1.htm
[Reuters, July 2024] Russian tech entrepreneur Volozh says AI venture Nebius aims to be bigger than Yandex | https://www.reuters.com/world/europe/russian-tech-entrepreneur-volozh-says-ai-venture-nebius-aims-be-bigger-than-yandex-2024-07-22/
[Nebius, July 2026] Nebius raises $775 million in first secured debt financing to accelerate global buildout | https://nebius.com/newsroom/nebius-raises-775-million-in-first-secured-debt-financing-to-accelerate-global-buildout
[Nebius, May 2025] Nebius participates in ClickHouse Series C, crystallizes value of asset to fund hypergrowth of core AI infrastructure business. | https://nebius.com/newsroom/nebius-participates-in-clickhouse-series-c-crystallizes-value-of-asset-to-fund-hypergrowth-of-core-ai-infrastructure-business
[Nebius, September 2025] Nebius provides financing update. | https://nebius.com/newsroom/nebius-provides-financing-update
[Reuters, March 2026] Nebius says ‘well-funded’ for AI race after closing $4.3 billion debt raise. | https://www.reuters.com/technology/nebius-says-well-funded-ai-race-after-closing-43-billion-debt-raise-2026-03-23/
[Nebius, August 2026] Nebius reports second quarter 2026 financial results. | https://nebius.com/newsroom/nebius-reports-second-quarter-2026-financial-results
[NVIDIA Newsroom, August 2026] Nebius Token Factory is the first AI cloud to adopt NVIDIA Groq 3 LPX | https://nvidianews.nvidia.com/news/nebius-token-factory-first-ai-cloud-to-adopt-nvidia-groq-3-lpx
[Bloomberg, Retrieved 2026] Bloomberg Billionaires Index - Arkady Volozh | https://www.bloomberg.com/billionaires/profiles/arkady-y-volozh/
[TechCrunch, June 2022] Russian tech giant Yandex removes national borders from Maps app | https://techcrunch.com/2022/06/09/yandex-maps-no-borders/amp
[Nebius] Nebius announces agreement to acquire Tavily to add agentic search to its AI cloud platform | https://nebius.com/newsroom/nebius-announces-agreement-to-acquire-tavily-to-add-agentic-search-to-its-ai-cloud-platform
[The Motley Fool, October 2026] How Nebius Could Beat Palantir, SoundHound, and BigBear.ai in the AI Arms Race | https://www.fool.com/investing/2026/10/06/how-nebius-could-beat-palantir-soundhound-and-bigb/
[LinkedIn, Retrieved 2026] Arkady Volozh | https://www.linkedin.com/in/arkady-volozh/
[nebius.com, Retrieved 2026] John Boynton | https://group.nebius.com/board-of-directors/john-boynton/
Articles about Nebius
- Nebius's $40 Billion in Contracts Funds a New AI Factory in Pennsylvania — The publicly traded spinout from Yandex is funding a $20 billion capex plan with customer prepayments, betting on a new model for hyperscale AI infrastructure.