Lately the AI news keeps throwing out names like CoreWeave, Nebius, and neocloud. They are a relatively new kind of cloud company with a very simple core business: buy a huge pile of GPUs and rent them out to people who need to run AI.
This piece spells out the neocloud. First what it is and where it sits in the AI supply chain, then how it differs from a hyperscaler like Amazon’s AWS, and finally its business model and risks. This is one entry in the “AI Industry Watch” series; to see how the money flows along the whole supply chain, pair it with the AI Concept-Stock Money Map, and to see the web of chipmakers and clouds investing in one another behind it, pair it with the AI Circular-Financing Map.
Core-Data Snapshot
The few numbers below help you grasp a neocloud’s scale and its “high-growth, still-loss-making” character.
| Company | Snapshot | Nature |
|---|---|---|
| CoreWeave (Nasdaq: CRWV) | Q1 2026 revenue about US$2.08 billion; order book about US$99.4 billion; quarterly net loss about US$740 million | Official results |
| CoreWeave full year | 2025 revenue about US$5.1 billion; net loss about US$1.17 billion | Official results |
| Nebius (Nasdaq: NBIS) | Q1 2026 revenue about US$400 million, up about 684% year over year | Official results |
| Positioning | GPU-as-a-service, wedged between GPUs/data centers and the AI companies | Industry positioning |
What a Neocloud Is
A neocloud (some call it a GPU cloud or an AI compute cloud) is a new kind of cloud provider that rents out GPU compute. Traditional clouds sell everything; a neocloud does almost one thing only: it builds large volumes of the latest AI GPUs into clusters and rents them to customers that need to train or run AI models.
Here’s an analogy: an ordinary car-rental firm has every kind of car, while a neocloud is like a rental shop that deals only in sports cars — few models, but if you want the newest, fastest sports car, and you want to rent a whole fleet at once, it’s the quickest place to find one.
Why did it emerge? Because over the past two years the latest AI GPUs have been in severe short supply. AI companies want to get large volumes of GPUs “fast, in bulk, right now,” and the traditional hyperscalers can’t always free up that capacity on the spot, which opened a door for the neoclouds that had stocked up on GPUs early.
Where It Sits in the Supply Chain
A neocloud is wedged in the middle. It buys GPUs from NVIDIA upstream, finds data centers to house the machines, hooks up the power and cooling, and then rents the compute to downstream model companies like OpenAI or to large enterprises.
It doesn’t design chips and it doesn’t build models; what it earns is the service margin on “turning individual GPUs into on-demand compute you can rent.” Its position is a bit like the compute landlord of the AI era.
How It Differs from AWS, Azure, and Google Cloud
Hyperscalers (AWS, Azure, Google Cloud) have it all: compute, storage, databases, a full suite of enterprise services and an ecosystem, with AI as just one slice. A neocloud focuses on GPU compute, and its selling points are fast supply, the ability to deliver dedicated large clusters, and sometimes more flexible pricing.
The hyperscalers’ advantages lie on the other side: complete services, a mature ecosystem, deep pockets, and high reliability. Interestingly, even hyperscalers and big model companies will sometimes come back to rent compute from a neocloud to plug a gap when they can’t free up enough capacity of their own. So the two are more complementary and “coopetitive” than locked in a fight to the death.
Business Model and Risks
A neocloud is a very heavy business: spend big on GPUs up front, build or rent data centers, then slowly earn the money back through long-term contracts. That gives it several pronounced traits and risks.
- Asset-heavy, debt-funded: buying GPUs takes a lot of capital, often funded by borrowing, so the interest burden is heavy. CoreWeave’s interest expense alone topped US$500 million in the first quarter of 2026.
- High growth, still loss-making: CoreWeave’s full-year 2025 net loss was about US$1.17 billion and its Q1 2026 net loss about US$740 million; Nebius is also still in its investment phase. But their order books are thick (CoreWeave’s is close to US$100 billion), meaning a lot of future revenue has already been signed.
- Customer concentration: a large share of revenue comes from a handful of big customers, so a change at one customer has a big impact.
- GPUs depreciate and grow obsolete: what gets rented out is a depreciating asset, and once a new generation of GPUs arrives, the competitiveness of the old ones slips.
- Deep ties to NVIDIA: NVIDIA is both a supplier and a shareholder in CoreWeave, and that layer of the relationship gets filed under the much-discussed theme of “circular financing.”
How to Look at These Companies
CoreWeave and Nebius are both U.S.-listed companies (tickers CRWV and NBIS). This piece only explains their business model and their role in the supply chain; it makes no judgment on the share price, gives no buy or sell advice, and sets no price targets.
Their reported figures are growing fast, but CoreWeave clearly carries the risks of high debt, customer concentration, and ongoing losses, Nebius is still in its investment phase, and the GPUs both rent out all depreciate. Understanding how this business “makes money, loses money, and gets stuck” is more practical than chasing the share price. And when the market pairs a given Taiwanese supplier to the neocloud supply chain, that is mostly analyst speculation, not a company announcement.
Key Takeaways for This Piece
A neocloud is a new kind of cloud provider that rents out GPU compute, the headliners being CoreWeave and Nebius. It is wedged between the chipmakers and the AI companies, and what it sells is the service of “turning GPUs into on-demand compute you can rent.”
This business is heavy: it borrows to buy GPUs and leans on a thick order book to underpin future revenue, growing fast but still loss-making, with concentrated customers and depreciating assets, and carrying a deep relationship with NVIDIA where it both buys from and is invested in by the chipmaker. Understanding the structure is more useful than chasing the share price.
To see who along the whole supply chain makes money and who burns it, read the AI Concept-Stock Money Map; to see the circular financing behind it, read the AI Circular-Financing Map; to see whether this counts as a bubble, read Is AI a Bubble; to see the physical hardware supply chain, head back to The AI Hardware Supply Chain, End to End.