Is Nebius overtaking CoreWeave?

Why is Nebius suddenly challenging CoreWeave?

Nebius is becoming a serious challenger to CoreWeave in the AI cloud market, and the gap between the two companies is closing much faster than it was a year ago.

Both companies rent powerful Nvidia GPU infrastructure to AI laboratories, technology companies, and enterprises. CoreWeave built its lead by supplying massive computing clusters to customers such as Microsoft and OpenAI. Nebius, a much smaller competitor, has recently started securing contracts of similar strategic importance. Landing customers like these is hard to fake, which is why we separate real customers from promises in our AI infrastructure report.

The turning point came when Nebius signed multibillion-dollar agreements with Microsoft and Meta. Its expanded Meta agreement, announced in March 2026, included $12 billion in dedicated capacity and up to $15 billion in additional purchases over five years. Nvidia also invested $2 billion in Nebius, matching an investment it had announced in CoreWeave earlier that year.

CoreWeave remains considerably larger, with $2.575 billion in second-quarter revenue against Nebius's $582 million. But its debt burden has become a growing concern. Reuters reported in mid-October that CoreWeave had approximately $50 billion in debt, while the cost of insuring that debt against default had risen sharply.

Nebius now has the customers, hardware partnerships, and expansion plans to compete with CoreWeave directly. The question is how quickly those advantages can turn into revenue, and whether CoreWeave's enormous head start will prove difficult to overcome.

What would it take for Nebius to overtake CoreWeave?

Nebius would need to close CoreWeave's revenue and operating-capacity gap while proving it can keep expanding without taking on unsustainable financing costs.

The two companies compete for much the same business, but their headline figures measure different things. CoreWeave already generates billions of dollars in quarterly revenue. Nebius is growing faster and has ambitious plans to bring much more computing capacity online.

One complication is how AI cloud companies report capacity. Contracted power represents electricity access reserved for current or future data centers. Active power represents infrastructure already operating. A company can announce several gigawatts of contracted power while still needing years to build the facilities and install GPUs.

Customer commitments create a similar distinction. A $10 billion contract spread over several years says something important about future demand, but the actual revenue arrives as computing services are delivered.

For our comparison, the strongest evidence comes from revenue already generated, operating infrastructure, customer contracts, and financing costs. Future capacity matters too, especially when the difference between two competitors is changing quickly.

Nebius needs to demonstrate that it can turn its huge pipeline into actual operating scale. Until then, CoreWeave's lead is more substantial than the growth percentages alone suggest.

How much bigger is CoreWeave than Nebius today?

CoreWeave is still about 4.4 times larger than Nebius in quarterly revenue, although that multiple has collapsed from 11.5 in just one year.

According to their second-quarter 2026 financial results, CoreWeave generated $2.575 billion in revenue, compared with $582.3 million for Nebius. A year earlier, the figures were $1.212 billion and $105.1 million, respectively.

That gives us a useful way to measure the competitive shift. Nebius went from generating approximately 8% of the two companies' combined quarterly revenue to more than 18%. It has more than doubled its relative share in twelve months.

CoreWeave still dominates in absolute dollars. Its revenue increased by $1.363 billion year over year, versus $477 million for Nebius. In other words, CoreWeave added almost three times as much quarterly revenue despite Nebius growing much faster.

The six-month numbers confirm the scale difference. CoreWeave generated $4.653 billion in first-half revenue, compared with $981.3 million for Nebius.

Revenue comparison	CoreWeave	Nebius
Q2 2025 revenue	$1,212M	$105.1M
Q2 2026 revenue	$2,575M	$582.3M
Q2 year-over-year growth	112%	454%
Absolute Q2 revenue increase	$1,363M	$477.2M
H1 2026 revenue	$4,653M	$981.3M
Q2 annualized revenue	$10.30B	$2.33B
Share of combined Q2 2025 revenue	92.0%	8.0%
Share of combined Q2 2026 revenue	81.5%	18.5%

Is Nebius growing faster than CoreWeave?

Nebius is growing more than four times as fast as CoreWeave in percentage terms, making it the clear winner on revenue growth right now.

Nebius reported 454% year-over-year revenue growth in the second quarter, against 112% for CoreWeave. The difference is substantial even allowing for Nebius's much smaller starting point. Growth of 454% against 112% is a wide gap, so we compare who is ahead and who is falling behind in our AI infrastructure market report.

Nebius also reported that its annualized revenue run rate reached $3 billion at the end of June, compared with $1.9 billion at the end of March. That represents a 58% increase in just three months. Its recognized quarterly revenue rose approximately 46% over the same period.

The gap between those two growth rates is interesting. Nebius was bringing revenue-generating capacity online especially quickly toward the end of the quarter, which could support higher recognized revenue in subsequent periods.

CoreWeave is experiencing a different phase of growth. Its quarterly revenue has already reached an annualized pace exceeding $10 billion, while its year-over-year growth remains above 100%. Maintaining that rate at such a large scale is impressive.

Still, Nebius has been gaining ground consistently enough that the comparison is changing. The real test will come as its revenue base grows beyond $1 billion per quarter, when sustaining triple-digit growth becomes considerably harder.

Could Nebius catch CoreWeave in revenue within a few years?

Nebius could catch CoreWeave in revenue within a few years, but the latest results are insufficient to predict when that might happen.

The strongest evidence comes from the changing revenue multiple. CoreWeave generated 11.5 times Nebius's quarterly revenue in the second quarter of 2025. One year later, the multiple had fallen to 4.4.

At the same time, the absolute revenue gap actually widened, from approximately $1.11 billion to $1.99 billion per quarter. Nebius has become much bigger relative to CoreWeave, yet CoreWeave continues adding more dollars of business.

For Nebius to close that $1.99 billion quarterly gap, it would need another period of extraordinary growth, a meaningful slowdown at CoreWeave, or some combination of the two. Its contracted customer demand provides a potential route, especially with major Meta deployments scheduled to begin in early 2027.

We should also consider what happens when hardware deliveries accelerate for both companies. CoreWeave has its own large expansion pipeline and long-term customer contracts, so projecting Nebius's growth while holding CoreWeave's revenue constant would produce a misleading crossover date.

A revenue crossover is now a credible possibility. The evidence supports rapid convergence, although it doesn't yet support a specific timetable. And frankly, anyone offering an exact crossover year from these numbers is getting ahead of the evidence.

Does CoreWeave still have more AI cloud contracts than Nebius?

CoreWeave still has a much larger disclosed contract portfolio than Nebius, giving it a substantial advantage in future revenue visibility.

CoreWeave reported approximately $104 billion in revenue backlog at the end of June 2026. Its second-quarter earnings announcement also identified more than $25 billion in net new customer commitments added early in the third quarter.

Taken together, those disclosures point to at least $129 billion in backlog and subsequent additions before later revenue recognition or contract adjustments.

Nebius reported approximately $40 billion in customer commitments in its second-quarter materials. This is an enormous commercial position for a company that generated less than $1 billion of revenue during the first half of the year.

Comparing the headline figures suggests CoreWeave has roughly 3.2 times as much committed future business. However, the companies do not necessarily apply identical definitions to backlog and customer commitments, so we should avoid treating that ratio as an accounting-standard comparison.

Another useful observation emerges when we compare these commitments with current revenue. Nebius's $40 billion represents more than seventeen times its second-quarter annualized revenue, while CoreWeave's $104 billion backlog represents roughly ten times its equivalent annualized revenue.

Those ratios are indicative rather than perfectly comparable, but they show how much of Nebius's future growth is already tied to contracts that have yet to become revenue.

Is Nebius taking big AI customers away from CoreWeave?

Nebius is winning the same kinds of customers as CoreWeave, although there is little public evidence of major customers abandoning CoreWeave to move exclusively to Nebius.

Nebius's recent customer wins include Microsoft, Meta, Reflection, and Cohere. Its second-quarter shareholder letter also mentioned a major quantitative trading company and another US-based AI laboratory. Four landmark AI cloud contracts signed during the quarter averaged more than $1 billion in total contract value.

CoreWeave's customer base includes Microsoft, Meta, OpenAI, Databricks, Cognition, Runway, Bentley Systems, Caterpillar, and Grammarly. Its quarterly disclosures also identified expanded relationships with Hudson River Trading, Periodic Labs, and Rescale.

The overlap between Microsoft and Meta is especially revealing. Both technology companies have signed substantial agreements with multiple AI infrastructure providers, suggesting that large buyers increasingly want access to several independent sources of computing power.

Microsoft is a good example. According to CoreWeave's regulatory filings, Microsoft represented approximately 67% of its 2025 revenue, compared with 62% in 2024. Nebius subsequently gained a major Microsoft infrastructure contract and reported meeting all scheduled capacity deliveries through the second quarter of 2026.

Nebius clearly has access to the same level of buyer as CoreWeave. But without customer-level spending data showing contract reductions or migrations, we cannot claim that its growth is directly replacing CoreWeave's existing revenue.

Could Nebius's $27 billion Meta deal change the competition?

Nebius's agreement with Meta could dramatically expand its AI cloud business, although only $12 billion of the $27 billion headline opportunity represents dedicated capacity orders.

The March 2026 agreement covers $12 billion in dedicated capacity over five years, with initial deployments expected in early 2027. Meta can also purchase otherwise-unsold capacity from specified clusters under the agreement, potentially adding up to $15 billion. Capacity that only starts arriving in early 2027 leaves open questions; we go through the biggest unsolved problems in our report covering AI infrastructure.

The additional $15 billion is an upper limit tied to qualifying capacity that Nebius intends to sell to other customers first. Adding both figures together as guaranteed incremental revenue would overstate the commercial commitment.

The arrangement nevertheless gives Nebius a useful advantage. It can market new computing capacity to other AI companies while having a large technology customer positioned to absorb qualifying capacity that remains unsold.

Nebius also plans to deploy early Nvidia Vera Rubin systems for the agreement. That could help it become a major provider of next-generation AI infrastructure, provided the hardware arrives and operates as expected.

It's a very attractive contract structure, particularly for a company building capacity so quickly. But the real impact will become clearer when those facilities enter service. Until then, Meta's agreement tells us much more about future demand than about how much computing capacity Nebius can deliver today.

Does Nebius have more data-center capacity than CoreWeave?

Nebius is targeting more contracted power than CoreWeave's latest reported figure, but CoreWeave remains ahead in publicly demonstrated active infrastructure.

CoreWeave reported approximately 3.7 gigawatts of contracted power and 1.5 gigawatts of active power at the end of the second quarter. It added nearly 500 megawatts of active capacity during that quarter alone.

Nebius raised its year-end contracted-power target to 5 gigawatts, having previously guided to more than 4 gigawatts. It also expects to deploy more than 1 gigawatt annually beginning in 2027.

Those plans give Nebius a potentially enormous future footprint. However, we are comparing CoreWeave's reported operating capacity with a Nebius target that includes infrastructure still under development.

A megawatt reserved through a power agreement requires substantial additional investment before it can generate AI cloud revenue. Construction, grid connections, cooling equipment, GPU deliveries, and networking infrastructure can all affect the schedule.

Nebius has expanded across the United States and Europe, including sites in New Jersey, Missouri, Oklahoma, Alabama, Minnesota, Pennsylvania, Finland, Iceland, France, Spain, and the United Kingdom. But it has not disclosed a directly comparable active-power figure that would let us establish whether it has caught CoreWeave operationally.

Infrastructure measure	CoreWeave	Nebius
Contracted power	Approx. 3.7 GW at Q2 end	5 GW year-end target
Active power	1.5 GW at Q2 end	No directly comparable disclosure
Recent operating expansion	Nearly 500 MW added in Q2	Additional US and European capacity commissioned
Longer-term ambition	More than 5 GW by 2030 through Nvidia partnership	More than 5 GW by 2030 through Nvidia partnership
Major cloud customers	Microsoft, Meta, OpenAI	Microsoft, Meta
Nvidia Vera Rubin preparations	NVL72 platform validation completed	Early NVL72 systems received for validation
Planned future rollout	Continuing multi-GW expansion	More than 1 GW annually from 2027
Biggest execution challenge	Financing additional infrastructure	Turning contracted power into active capacity

Is Nvidia giving Nebius the same support as CoreWeave?

Nvidia is now backing Nebius with roughly the same strategic commitment it has made to CoreWeave, which is a major change from the earlier competitive picture.

In January 2026, Nvidia announced a $2 billion investment in CoreWeave at $87.20 per share. The partnership included plans to accelerate the deployment of more than 5 gigawatts of AI infrastructure by 2030.

Six weeks later, Nvidia announced a separate $2 billion investment in Nebius, also supporting the deployment of more than 5 gigawatts of Nvidia systems by the end of 2030.

These announcements are unusually similar. Both involve the same investment amount, the same broad infrastructure target, and cooperation around Nvidia's advanced computing technology.

Nvidia has obvious commercial reasons to maintain several large AI cloud partners. Each successful provider represents another customer for its GPUs, CPUs, networking equipment, and software.

Still, Nebius receiving this level of backing changes the competitive balance. CoreWeave previously stood out as Nvidia's best-known specialist cloud partner. Nebius now has direct financial support and a comparable expansion ambition from the company supplying the market's most important AI accelerators.

The partnership gives Nebius credibility and access to technical cooperation, although equal investments do not necessarily imply equal hardware allocations, commercial terms, or future purchasing volumes.

Is Nebius getting better prices for its GPUs than CoreWeave?

Nebius is securing increasingly attractive contracts for its GPU capacity, but we cannot yet establish whether its pricing is better than CoreWeave's for comparable infrastructure.

In its second-quarter shareholder letter, Nebius reported that four major AI cloud deals averaged more than $20 million in annual contract value per megawatt. Its earlier baseline was approximately $12 million per megawatt.

That represents an increase of at least 67% in contracted annual value per unit of power. The company also reported that pricing for older-generation GPUs rose by more than 30% quarter over quarter. Prices jumping like this are why we weigh which business models are winning in our AI infrastructure market report.

Early third-quarter short-duration deals presented even higher pricing opportunities, exceeding $40 million in annualized contract value per megawatt.

These figures suggest that Nebius is currently benefiting from very strong demand for immediately available compute. However, the highest short-duration prices should not be confused with the economics of long-term contracts.

At $20 million per megawatt annually, a fully contracted 100-megawatt facility would theoretically represent $2 billion in annual contract value. That is a large revenue opportunity, but delivering the capacity requires expensive hardware, electricity, cooling, networking, and financing.

CoreWeave does not provide a sufficiently similar breakdown of its latest contracts to support a direct comparison. Its existing customers also operate across different GPU generations and contract durations.

Nebius's improvement is clear. Whether it can consistently make more money from an equivalent megawatt than CoreWeave will depend on costs and actual utilization, not simply headline contract prices.

Is Nebius financing its growth better than CoreWeave?

Nebius has developed a promising financing model built around customer prepayments and GPU-backed loans, while CoreWeave is increasingly paying the price for its much larger debt burden.

According to Nebius's second-quarter shareholder letter, approximately 70% of its new deals included customer prepayments covering 50% to 60% of the related capital expenditure.

This arrangement helps Nebius pay for infrastructure before the full contract revenue arrives. Instead of financing every GPU entirely through its own cash or borrowing, it receives substantial funding from customers who want guaranteed capacity.

In July, Nebius also secured approximately $775 million through a senior secured financing facility backed by GPU infrastructure and contracted cash flows. The loan carried a rate of SOFR plus 2.50 percentage points, with maturity in October 2030.

Nebius has also turned to convertible financing on a substantial scale. It completed approximately $4.3 billion in convertible-note financing in March and an additional $5.75 billion offering on August 24, 2026. That second transaction came after the company's second-quarter results and materially strengthened its capacity to fund expansion, while adding future repayment and potential dilution considerations.

CoreWeave has raised far more capital overall. Its second-quarter disclosures included a $3.1 billion infrastructure-backed term loan and more than $10 billion raised through unsecured and convertible debt instruments.

The difference is becoming visible in borrowing costs. Reuters reported on October 15 that CoreWeave's debt had reached approximately $50 billion and that the cost of insuring it against default had risen to roughly 1,000 basis points.

Nebius's newer financing arrangements look attractive by comparison, especially when customers cover a large portion of upfront costs. But it would be premature to describe Nebius as lightly financed or insulated from credit risk. Building several gigawatts of AI infrastructure remains an extremely capital-intensive exercise, whatever the financing structure.

Could CoreWeave's debt allow Nebius to overtake it?

CoreWeave's growing debt burden gives Nebius a genuine competitive opening, especially if borrowing becomes more expensive or investors become less willing to finance large AI infrastructure projects.

CoreWeave reported $640 million in net interest expense during the second quarter of 2026, up from $267 million a year earlier. That is a 140% increase, compared with 112% revenue growth.

The company generated $128 million in adjusted operating income during the quarter. Its net interest expense was therefore five times that amount, contributing to a $626 million net loss.

CoreWeave's adjusted operating margin also fell from 16% to 5% year over year, despite the substantial revenue increase. Financing and infrastructure costs are consuming much of the operating benefit from its growing customer base.

The credit market has become particularly concerned. Reuters reported that credit-default-swap pricing implied a nearly 63% five-year default probability under standard market assumptions, compared with around 30% a year earlier. Such probabilities reflect market pricing and model assumptions rather than a reliable prediction of actual default.

CoreWeave does have important protections. Its long-term contracts provide future revenue visibility, its infrastructure-backed loans are supported by assets, and management can refinance obligations as conditions allow.

But the arithmetic is becoming uncomfortable. Interest expense is rising faster than revenue, margins are narrowing, and creditors are demanding much more protection. If Nebius can keep its financing costs under control while fulfilling large customer contracts, CoreWeave's balance sheet could become its biggest competitive disadvantage.

Is Nebius more profitable than CoreWeave?

Nebius has a healthy reported margin in its AI cloud operations, but CoreWeave currently reports a higher adjusted EBITDA margin, and both companies remain unprofitable after major costs.

Nebius reported a 50% adjusted EBITDA margin for its AI cloud segment in the second quarter. CoreWeave reported a 59% consolidated adjusted EBITDA margin, equivalent to approximately $1.51 billion.

The comparison requires care because the companies report these figures at different levels. Nebius's number relates to its AI cloud operations, while CoreWeave's covers its consolidated business.

CoreWeave's strong EBITDA figure also hides the burden of depreciation and interest. It reported $1.393 billion in depreciation and amortization, $640 million in net interest expense, and a $626 million net loss during the quarter.

Its reported operating loss was $49 million. Adjusted operating income reached $128 million, equivalent to just 5% of revenue.

Nebius also reported an operating loss at the group level, reflecting substantial infrastructure costs and its ongoing expansion.

Both companies can generate attractive operating returns from rented computing capacity while still struggling to produce net profits after paying for hardware and financing.

At present, Nebius's cloud business is showing encouraging operating economics. We cannot establish that Nebius generates better overall returns than CoreWeave, particularly because it has not yet operated at the same revenue scale.

Does Nebius have better AI cloud technology than CoreWeave?

Nebius has built a serious AI cloud platform, but CoreWeave currently has the stronger record of running a much larger business with demanding AI and enterprise customers.

Nebius has expanded beyond basic GPU rentals through Aether 3.6, Nebius Token Factory, and new deployment tools. It also introduced Nebius Echo, an AI agent designed to help developers manage workloads.

The company has invested heavily in inference, where AI models respond to user requests after training. Its production inference workloads more than tripled during the second quarter. The Eigen AI and Clarifai teams also joined Nebius, bringing expertise in inference optimization and model deployment. Inference workloads tripling in one quarter shows where demand is moving; we cover which technologies are gaining traction in our AI infrastructure report.

CoreWeave has developed similar capabilities through its SUNK orchestration platform, CoreWeave Interconnect, SUNK Anywhere, and LOTA Cross-Cloud. It has also reported strong results in MLPerf testing using Nvidia-based infrastructure.

These technologies help customers run AI models reliably, manage expensive computing resources, and move workloads between environments.

One particularly important question is whether either provider can earn meaningful software revenue beyond GPU infrastructure rentals. Better orchestration and inference tools can attract customers and improve hardware utilization, but neither company has shown that proprietary software is already a dominant independent source of profit.

We also lack comprehensive independent benchmarks comparing Nebius and CoreWeave on identical GPU configurations, workloads, availability requirements, and total costs.

Nebius is competitive enough technologically to win major buyers. CoreWeave has proved it can operate at a much greater scale. For now, that execution record is a more useful differentiator than either company's product announcements.

Could cheaper AI computing hurt Nebius and CoreWeave?

Cheaper AI computing could hurt both Nebius and CoreWeave, especially if GPU rental prices fall faster than the companies can recover their hardware investments.

The immediate evidence actually points toward strong pricing. Nebius reported rising rates for older-generation GPUs and significantly higher annual contract values per megawatt in recent deals.

That suggests customers are still willing to pay substantial premiums for available computing capacity, even as data-center operators continue installing new hardware.

However, GPUs become less attractive as newer generations deliver more computing performance for the same power consumption. A cluster built around older Nvidia chips may need to offer lower prices once customers can access better hardware elsewhere.

CoreWeave's long-term customer agreements provide some protection against sudden repricing. Nebius benefits from a mixture of large multiyear commitments and shorter-duration contracts that can capture higher prices when demand is strong.

Those shorter contracts can become a disadvantage if market conditions reverse. Operators still need to cover depreciation, electricity, maintenance, and debt payments even when rental prices decline.

This risk becomes especially important as Nvidia moves from Blackwell toward Vera Rubin systems. Both Nebius and CoreWeave are preparing to deploy newer hardware, which creates additional investment requirements alongside opportunities to charge premium prices.

For now, both companies are operating in a market with strong reported demand. The concern is what happens to the economics of today's expensive GPU installations when the next two or three generations of hardware arrive.

Is Nebius overtaking CoreWeave?

Nebius is rapidly catching CoreWeave, but CoreWeave is still ahead. Nebius has the stronger percentage growth, increasingly important customer relationships, and ambitious infrastructure plans. It has yet to match CoreWeave's operating scale.

The change over one year is remarkable. CoreWeave's quarterly revenue advantage fell from 11.5 times to 4.4 times, while Nebius grew 454% year over year. That is genuine competitive progress, especially considering CoreWeave's own revenue more than doubled over the same period.

Nebius has also secured major contracts with Microsoft and Meta, received $2 billion from Nvidia, and established an ambitious 5-gigawatt contracted-power target. Its newer customer agreements show that it can compete for infrastructure projects previously associated with much larger providers.

CoreWeave still has the stronger operating position. Its $2.575 billion quarterly revenue, substantial backlog, and 1.5 gigawatts of active power demonstrate a scale that Nebius has not yet reached.

The financial picture is less favorable to CoreWeave. Its approximately $50 billion debt burden, rising interest costs, and widening losses create risks that Nebius may be able to exploit. But Nebius will face similar pressures as it finances its own expansion, even with its recent convertible-note proceeds and customer prepayments.

The biggest unanswered question is how quickly Nebius can turn its contracted power and multibillion-dollar deals into operating GPU infrastructure. That conversion will determine whether its current growth advantage lasts.

CoreWeave remains the AI cloud leader in this comparison today. Nebius has become a credible contender to catch it, and the next wave of data-center deployments could make the race considerably closer.