Is Nebius ($NBIS) a neocloud? Yes — a deep dive into the NVIDIA GPU data centers (Finland, US Vineland NJ, UK Surrey, France, Iceland) on an interactive map, the Meta and Microsoft anchor contracts and backlog, the full supply chain, and the risks behind the neocloud's 2026 run.
More Macroplane supply-chain guides · Baskets · Investment theses · Macro trends
Nebius ($NBIS) is a pure-play AI cloud — a "neocloud" — that builds and operates its own NVIDIA GPU data centers and rents that compute to AI labs and enterprises. Spun out of the old Yandex group and now Nasdaq-listed at roughly a $50B market cap, it is one of the few independent operators with the capital, the NVIDIA relationship, and the data-center pipeline to scale into the GPU shortage rather than just resell it. This is a look at what Nebius actually owns: the hardware, the data centers, the customer backlog, and the risks.
👉 See it in context: $NBIS sits in the AI Cloud / Neoclouds basket on Macroplane, alongside $CRWV, $IREN, and the other operators renting out accelerated compute.
Nebius stock ($NBIS) has been one of the loudest AI-infrastructure trades of 2026. The shares ran roughly 30% in a week and more than 40% over the past month heading into mid-June 2026, lifting the market cap to around $50B as a run of catalysts stacked up:
The valuation is the other side of the trade. Nebius does not trade on earnings; like its neocloud peers it trades on a forward revenue multiple and on confidence that the contracted backlog converts to cash. Wall Street's average rating is a "moderate buy," though some analysts called the stock too expensive even after the UK expansion. The number that matters more than today's price is whether each new gigawatt of capacity comes online on schedule and gets paid for — which is what the rest of this piece maps out.
A neocloud is only as good as its physical footprint, and Nebius has moved fast from one Finnish site to a multi-region buildout across the US and Europe.
The footprint, region by region:
The most recent expansion is the UK. Nebius is taking purpose-built, liquid-cooled data-hall capacity at Ark Data Centres' Longcross Park in Surrey, with an initial deployment of 4,000 NVIDIA Blackwell Ultra GPUs going live around Q4 2025 (Nebius announcement). It is a small share of total fleet megawatts, but it matters strategically: it puts sovereign, in-country AI compute next to UK enterprise and government demand, where data-residency rules make a local operator hard to replace.
Nebius designs its own server racks and runs a current-generation NVIDIA fleet rather than a grab-bag of leased boxes:
This is the same shopping list as the rest of the AI buildout — which is the point: a neocloud is a concentrated bet on the whole upstream chain.
The reason a capital-hungry neocloud can build gigawatts is contracted demand. Nebius has signed anchor deals reported in the tens of billions, led by a multi-year capacity agreement with Meta ($META) — disclosed in our deal graph at up to $27B (roughly $12B tied to the NVIDIA Vera Rubin platform plus ~$15B of flexible capacity) — and a separate large infrastructure agreement with Microsoft ($MSFT) reported around $17.4B. On top of the hyperscaler anchors sit enterprise and AI-lab customers including Shopify ($SHOP), Cloudflare ($NET), Prosus ($PRX.AS), Palantir ($PLTR) as a stack partner, and labs like Mistral, plus a long tail of generative-AI startups training on Nebius B200 clusters.
That backlog is the bull case: take-or-pay-style contracts give revenue visibility against the capex, which is what separates a fundable neocloud from a speculative one.
Open $NBIS on Macroplane and the graph shows the dual structure clearly — a heavy upstream of chip, server, network, power, and colocation suppliers feeding in, and a downstream of hyperscalers, enterprises, and AI labs pulling capacity out:
Nebius is the European-rooted, vertically-integrated neocloud — it owns the data centers rather than only leasing them, which is the structural difference from an asset-light reseller. The cleaner way to hold the theme is the layer:
Yes — Nebius ($NBIS) is one of the defining "neoclouds": specialized AI cloud providers that build and operate their own NVIDIA GPU data centers and rent that accelerated compute to AI labs and enterprises, rather than running general-purpose cloud like AWS or Azure. (New to the category? Start with the neocloud stocks guide.) Its closest pure-play peer is CoreWeave ($CRWV), and both sit in the AI Cloud / Neoclouds basket.
Its origins are unusual. Nebius is the renamed remnant of Yandex N.V. — the Dutch holding company of the Russian search group — which sold its Russian operations for roughly $5.4B in July 2024, renamed itself Nebius Group N.V., and resumed trading on Nasdaq under the ticker $NBIS on 21 October 2024 (Nasdaq had halted the shares since early 2022). What it kept was a cash-rich balance sheet, a seasoned cloud-engineering team, and an early NVIDIA GPU footprint — which it has since pointed entirely at the AI-compute build-out.
Why it is actually a neocloud and not just a GPU reseller: Nebius designs its own server racks and builds and operates its own data centers across Europe and the US (Finland, New Jersey, the UK, France, Iceland) instead of leasing capacity, and it funds that build-out with multi-year anchor contracts reported in the tens of billions with Meta ($META) and Microsoft ($MSFT). That vertical integration — owning the buildings, the racks, and the power deals — is the structural difference between a neocloud and an asset-light reseller. See the full Nebius ($NBIS) investment thesis for the footprint, backlog, and risks.
Nebius ($NBIS) is a specialized AI cloud ("neocloud"). It builds and operates NVIDIA GPU data centers and rents that accelerated compute to AI labs and enterprises, alongside smaller businesses in data labeling (Toloka) and autonomous driving (Avride). It is Nasdaq-listed and emerged from the restructuring of the former Yandex group.
Operating and announced sites span Finland (Mäntsälä and a 310 MW Lappeenranta factory), the US (Kansas City, Vineland NJ, and a gigawatt-scale Independence MO site), the UK (Surrey, via Ark Data Centres), France (Paris and a 240 MW Lille factory), and Iceland (Keflavík). See the map above.
Primarily NVIDIA — H100/H200, HGX B200, GB200 Grace Blackwell, and Blackwell Ultra, with Vera Rubin capacity planned. It evaluates AMD MI300X as a secondary source and runs Intel Xeon host CPUs in Supermicro and Gigabyte racks.
Nebius's growth case rests on a multi-year contracted backlog — long-term capacity deals that lock in future revenue before the data centers are even built. The two anchor contracts publicly reported are with Meta and Microsoft, together in the tens of billions of dollars, and management has pointed to a backlog that runs years out. The bull case is that this backlog de-risks the capex ramp; the bear case is that converting it still requires building gigawatts of power-hungry capacity on time and on budget. Always check the latest quarterly filing for the current contracted figure.
Several catalysts stacked up through 2026: inclusion in the Nasdaq-100 (which forces index buying), reported revenue growth near 684% year over year as the GPU-cloud business scaled, new capacity deals including the UK Ark Data Centres expansion, and visible interest from AI-focused funds. The shares ran roughly 30% in a week and 40%+ over a month heading into mid-June 2026. Momentum cuts both ways, though: the same rich multiple that rewards good news amplifies any disappointment.
That depends on your view of neocloud economics and your tolerance for volatility. The bull case: a contracted backlog (anchor deals with Meta and Microsoft reported in the tens of billions), hypergrowth revenue, and fresh Nasdaq-100 inclusion. The bear case: it is pre-profit on a GAAP basis, trades on a rich forward revenue multiple, funds its build-out with debt and equity raises (dilution), and competes with $CRWV and the hyperscalers for both GPUs and customers — some analysts have called it too expensive even after the UK deal. Wall Street's average rating is a "moderate buy." This is not financial advice — review the latest filings and size for the volatility.
Both are pure-play AI clouds. Nebius is more vertically integrated and European-rooted with its own builds across the US and EU; CoreWeave ($CRWV) is the larger US pure-play — the CoreWeave deep dive covers its backlog and the debt behind it. Track both in the AI Cloud / Neoclouds basket.
Yes — Nebius ($NBIS) is one of the defining "neoclouds": specialized AI cloud providers that build and operate their own NVIDIA GPU data centers and rent that accelerated compute to AI labs and enterprises, rather than running general-purpose cloud like AWS or Azure. (New to the category? Start with the neocloud stocks guide.) Its closest pure-play peer is CoreWeave ($CRWV), and both sit in the AI Cloud / Neoclouds basket. Its origins are unusual. Nebius is the renamed remnant of Yandex N.V. — the Dutch holding company of the Russian search group — which sold its Russian operations for roughly $5.4B in July 2024, renamed itself Nebius Group N.V., and resumed trading on Nasdaq under the ticker $NBIS on 21 October 2024 (Nasdaq had halted the shares since early 2022). What it kept was a cash-rich balance sheet, a seasoned cloud-engineering team, and an early NVIDIA GPU footprint — which it has since pointed entirely at the AI-compute build-out. Why it is actually a neocloud and not just a GPU reseller: Nebius designs its own server racks and builds and operates its own data centers across Europe and the US (Finland, New Jersey, the UK, France, Iceland) instead of leasing capacity, and it funds that build-out with multi-year anchor contracts reported in the tens of billions with Meta ($META) and Microsoft ($MSFT). That vertical integration — owning the buildings, the racks, and the power deals — is the structural difference between a neocloud and an asset-light reseller. See the full Nebius ($NBIS) investment thesis for the footprint, backlog, and risks.
Nebius ($NBIS) is a specialized AI cloud ("neocloud"). It builds and operates NVIDIA GPU data centers and rents that accelerated compute to AI labs and enterprises, alongside smaller businesses in data labeling (Toloka) and autonomous driving (Avride). It is Nasdaq-listed and emerged from the restructuring of the former Yandex group.
Operating and announced sites span Finland (Mäntsälä and a 310 MW Lappeenranta factory), the US (Kansas City, Vineland NJ, and a gigawatt-scale Independence MO site), the UK (Surrey, via Ark Data Centres), France (Paris and a 240 MW Lille factory), and Iceland (Keflavík). See the map above.
Primarily NVIDIA — H100/H200, HGX B200, GB200 Grace Blackwell, and Blackwell Ultra, with Vera Rubin capacity planned. It evaluates AMD MI300X as a secondary source and runs Intel Xeon host CPUs in Supermicro and Gigabyte racks.
Nebius's growth case rests on a multi-year contracted backlog — long-term capacity deals that lock in future revenue before the data centers are even built. The two anchor contracts publicly reported are with Meta and Microsoft, together in the tens of billions of dollars, and management has pointed to a backlog that runs years out. The bull case is that this backlog de-risks the capex ramp; the bear case is that converting it still requires building gigawatts of power-hungry capacity on time and on budget. Always check the latest quarterly filing for the current contracted figure.
Several catalysts stacked up through 2026: inclusion in the Nasdaq-100 (which forces index buying), reported revenue growth near 684% year over year as the GPU-cloud business scaled, new capacity deals including the UK Ark Data Centres expansion, and visible interest from AI-focused funds. The shares ran roughly 30% in a week and 40%+ over a month heading into mid-June 2026. Momentum cuts both ways, though: the same rich multiple that rewards good news amplifies any disappointment.
That depends on your view of neocloud economics and your tolerance for volatility. The bull case: a contracted backlog (anchor deals with Meta and Microsoft reported in the tens of billions), hypergrowth revenue, and fresh Nasdaq-100 inclusion. The bear case: it is pre-profit on a GAAP basis, trades on a rich forward revenue multiple, funds its build-out with debt and equity raises (dilution), and competes with $CRWV and the hyperscalers for both GPUs and customers — some analysts have called it too expensive even after the UK deal. Wall Street's average rating is a "moderate buy." This is not financial advice — review the latest filings and size for the volatility.
Both are pure-play AI clouds. Nebius is more vertically integrated and European-rooted with its own builds across the US and EU; CoreWeave ($CRWV) is the larger US pure-play — the CoreWeave deep dive covers its backlog and the debt behind it. Track both in the AI Cloud / Neoclouds basket.