Every dollar they bring in gets plowed back into expanding
In the infrastructure layer beneath the AI revolution, a familiar drama is unfolding: upstarts racing to claim territory before the giants consolidate their grip. CoreWeave and Nebius have emerged as specialized cloud providers built entirely around artificial intelligence workloads, growing at rates that dwarf even the impressive expansion of AWS, Azure, and Google Cloud. They are following a well-worn path — build fast, scale hard, worry about profits later — in a market where the rewards for those who achieve dominance have proven to be extraordinary. The question, as it so often is, is whether speed and focus can outrun incumbency and capital.
- CoreWeave and Nebius are posting triple- and quadruple-digit revenue growth, numbers so large they read as errors — 112% and 684% year over year respectively — signaling that demand for AI-specific computing infrastructure is outpacing what the established giants can absorb alone.
- The hyperscalers are not standing still: AWS, Azure, and Google Cloud are growing at 28–63% annually, generating fortress-level profits, and Google has even begun selling its own proprietary AI chips to outside customers, blurring the line between cloud provider and hardware competitor.
- Neither neocloud is profitable, and neither is expected to be through 2027, as every dollar earned is immediately reinvested into infrastructure — a high-wire act that leaves them exposed if AI demand softens or incumbents respond aggressively.
- Major customers like Microsoft and Meta are already buying from CoreWeave and Nebius even while operating their own data centers, suggesting the appetite for flexible, on-demand AI compute is real and growing — but also that the neoclouds remain dependent on the goodwill of companies that are simultaneously their rivals.
- Analysts are projecting continued explosive growth, but Wall Street's optimism is a forecast, not a floor — and the gap between a compelling growth story and a sustainable business remains the central risk investors must sit with.
The cloud computing industry has become the essential infrastructure of artificial intelligence, and the three companies that built it first — Amazon Web Services, Microsoft Azure, and Google Cloud — are reaping extraordinary rewards. AWS alone generates 59 percent of Amazon's total operating income. Azure is growing at 40 percent annually. Google Cloud is expanding at 63 percent and has begun selling its own AI chips to outside customers. These are not struggling businesses. They are among the most profitable enterprises in the history of technology.
And yet, a new category of competitor is gaining ground at a pace that makes even those numbers look modest. CoreWeave and Nebius are neoclouds — companies built from scratch to serve the AI economy, with no legacy businesses and no diversified revenue streams. They do one thing: provide computing infrastructure for artificial intelligence workloads. In the first quarter of this year, CoreWeave grew 112 percent year over year. Nebius grew 684 percent. Analysts expect these trajectories to continue, forecasting CoreWeave at 147 percent growth in 2026 and Nebius at 551 percent. Microsoft and Meta are already customers, drawn by the ability to access additional computing capacity without the years-long commitment of building their own facilities.
The risk is structural and significant. Neither company is profitable, and neither is expected to be through 2027. Every dollar of revenue is being reinvested into infrastructure expansion — the classic growth-at-all-costs playbook that rewards patience and punishes slowdowns. If demand for AI computing softens, or if the hyperscalers respond with pricing pressure or accelerated capacity, CoreWeave and Nebius could find themselves holding enormous infrastructure investments with no clear path to the margins they're chasing.
The template they're following is real: AWS and Google Cloud proved that cloud infrastructure, at scale, becomes one of the most profitable businesses imaginable. The neoclouds are betting they can replicate that outcome faster and with sharper focus. Whether they can execute that bet while the incumbents are watching — and competing — is the question that will define the next chapter of the AI infrastructure story.
The cloud computing business has become the engine room of artificial intelligence. Companies building AI systems need somewhere to run them, and most don't have the capital or expertise to construct their own data centers. So they rent processing power from the established players: Amazon Web Services, Microsoft Azure, and Google Cloud. These three have built fortress-like positions in the market, and they're printing money from it. AWS alone accounts for 59 percent of Amazon's operating income. Azure is growing at 40 percent annually. Google Cloud is expanding even faster, at 63 percent, though that figure includes sales of its proprietary Tensor Processing Units to outside customers.
But a new breed of competitor is emerging, and they're growing at a pace that makes the hyperscalers look sluggish by comparison. CoreWeave and Nebius are what the industry calls neoclouds—cloud specialists built from the ground up to serve the AI economy. They don't have the legacy businesses or the diversified revenue streams of Amazon, Microsoft, and Alphabet. They do one thing: provide computing infrastructure for artificial intelligence workloads. And they're doing it with stunning velocity.
In the first quarter of this year, CoreWeave's revenue grew 112 percent year over year. Nebius grew 684 percent. Those numbers are not typos. Wall Street is betting these growth trajectories will continue. Analysts expect CoreWeave to expand 147 percent in 2026 and 97 percent in 2027. Nebius is forecast to grow 551 percent this year and 224 percent next year. Major technology companies—Microsoft and Meta among them—are already customers, even though they operate their own data centers. The appeal is clear: they can access additional computing capacity without the years-long commitment and capital expenditure of building it themselves.
The catch is that neither CoreWeave nor Nebius is profitable. They're not even close. Every dollar they bring in gets plowed back into expanding their infrastructure. This is the classic growth-at-all-costs playbook, and it carries real risk. If the demand for AI computing capacity slows, or if the established cloud providers respond aggressively to the threat, these companies could find themselves with massive infrastructure investments and no clear path to profitability. The bet is that they'll grow fast enough and long enough to eventually achieve the kind of margins that AWS and Google Cloud enjoy—but that's a bet, not a guarantee.
What makes this moment interesting is that cloud computing itself has proven to be a standalone, highly profitable business. AWS generates more operating income for Amazon than the entire retail operation. Azure is doing the same for Microsoft. Google Cloud, despite being the smallest of the three, is the fastest-growing and is now selling its own chips to competitors. That success has created a template: build the infrastructure, attract customers, achieve scale, and the profits follow. CoreWeave and Nebius are betting they can follow that same path, just faster and more focused. Whether they can execute that bet while the incumbents are watching closely is the question investors will be wrestling with for the next few years.
Citações Notáveis
Cloud computing can be a viable standalone business— Analysis of AWS, Azure, and Google Cloud profitability
A Conversa do Hearth Outra perspectiva sobre a história
Why would a company like Microsoft or Meta use CoreWeave or Nebius when they could just expand their own data centers?
Because building a data center takes years and billions of dollars. If you need computing capacity next quarter, you can't build your way there. You rent it. These neoclouds can provision capacity much faster than the hyperscalers can, and they're specialized in AI, so they understand the workloads better.
But AWS and Azure are already specialized in AI. Why would they lose customers to smaller competitors?
They're not losing customers—not yet. But the neoclouds are growing faster because they're starting from zero. They're not carrying legacy infrastructure or supporting older workloads. They're built for this moment. It's like comparing a startup to an incumbent. The startup can move faster.
If these companies are growing at 500 percent a year but making no profit, aren't they just burning cash?
They're investing cash. There's a difference. The question is whether that investment will eventually generate returns. AWS took years to become profitable, but once it did, it became one of the most valuable business units in the world. CoreWeave and Nebius are betting they can follow that same arc.
What happens if they can't?
Then you've got companies with massive infrastructure investments and no way to pay for them. That's the risk. But if they do achieve profitability, the upside could be enormous. That's why Wall Street is bullish on them despite the lack of profits today.