Which AI data center startup is growing the fastest?

Why are AI data center startups growing so fast right now?

AI data center startups are growing at extraordinary speed because companies building AI models need more computing power than their existing infrastructure can provide.

The clearest evidence comes from the size of the contracts being signed. CoreWeave reported approximately $104 billion in revenue backlog in its second-quarter 2026 results. Crusoe disclosed more than $140 billion in total contracted value in September 2026, while Nscale's IPO filing revealed approximately $103 billion in active and contracted commitments. Backlogs that size need cash to build, and we follow who has raised the most, and from whom, in our data centers market report.

These are multiyear commitments rather than annual sales, but their scale is remarkable. Just three specialized providers have disclosed contractual figures totaling more than $340 billion, although the definitions differ and the amounts cannot be treated as equivalent guaranteed revenue.

Microsoft, Meta, OpenAI, and other major AI customers are looking beyond their own data centers to secure enough GPUs and electricity. That has created an opening for specialized providers that can build facilities, secure power connections, and deploy computing equipment quickly.

The unusual part is how far ahead customer commitments have moved compared with operational capacity. Some startups have signed contracts worth tens of billions of dollars while most of the infrastructure needed to fulfill them is still under development.

Which AI data center startup is actually growing the fastest?

Nscale currently leads the major AI data center challengers in disclosed percentage revenue growth, while Nebius is growing much faster in absolute dollar terms and Crusoe has announced one of the largest infrastructure expansion programs.

According to Nscale's September 2026 IPO filing, its revenue increased from $10.4 million to $140.6 million between the first halves of 2025 and 2026. That represents 1,252% growth, meaning the company generated more than thirteen times as much revenue as it did a year earlier.

Nebius reported $981.3 million in first-half group revenue, up 529% from $156 million. Its core AI cloud business grew even faster, expanding 614% to $964.6 million.

CoreWeave, which went public in March 2025, generated $4.65 billion over the same six-month period, an increase of 112%. Its growth rate was much lower, but the additional $2.46 billion it generated was nearly three times Nebius's revenue increase.

Crusoe is harder to rank. It reported more than twentyfold year-over-year growth in Crusoe Cloud bookings in September, alongside over $140 billion in total contracted value. Without comparable recognized-revenue figures, we cannot establish that its actual sales are growing faster than Nscale's.

Is Nscale's 1,252% revenue growth as impressive as it sounds?

Nscale's revenue growth is genuinely exceptional, but the company's small starting point makes its 1,252% increase easier to achieve than a similar result at Nebius's or CoreWeave's scale.

Nscale generated $140.6 million in the first half of 2026, compared with just $10.4 million one year earlier. That is approximately $130 million in additional revenue.

The longer trajectory supports the acceleration. Nscale's annual revenue increased from $19.1 million in 2024 to $33 million in 2025. Its first-half 2026 sales were already more than four times its entire 2025 revenue.

However, Nebius added approximately $825 million in group revenue over the same first-half comparison, while CoreWeave added approximately $2.46 billion. Both companies generated several times more additional business than Nscale, despite posting lower growth percentages.

Ranking these businesses by a single percentage would be misleading. Nscale leads on speed relative to its starting size. Nebius has achieved something more difficult: maintaining triple-digit growth with a revenue base approaching $2 billion annually, even before considering its accelerating quarterly run rate.

Company	H1 2025 revenue	H1 2026 revenue	Growth	Revenue added
Nscale	$10.4M	$140.6M	+1,252%	$130.2M
Nebius Group	$156.0M	$981.3M	+529%	$825.3M
Nebius AI cloud segment	$135.1M	$964.6M	+614%	$829.5M
CoreWeave	$2.19B	$4.65B	+112%	$2.46B

These figures compare recognized revenue over the same six-month periods. The Nebius AI cloud segment is already included in Nebius Group's total.

Why is Nebius growing so fast right now?

Nebius is growing unusually fast because its AI cloud business is bringing more capacity online while earning more revenue from its existing computing infrastructure.

Its second-quarter 2026 AI cloud revenue reached $574.9 million, compared with $93.7 million one year earlier, an increase of 514%. For the first six months, growth was even stronger at 614%. A 514% jump in one quarter changes the pecking order, and we look at who is ahead and who is falling behind in our report covering data centers.

Nebius's annualized run-rate revenue reached approximately $3 billion at the end of June, compared with $1.9 billion at the end of March. That represents a 58% increase in one quarter. The annualized figure measures the pace of revenue at a particular point rather than actual revenue earned over twelve months.

Another interesting detail is that the AI cloud segment accounted for approximately 98% of Nebius Group's second-quarter revenue. Its smaller businesses, including autonomous-driving technology and education, contributed very little to the overall growth figure.

Nebius also reported higher utilization and improved pricing. The company was getting more revenue from available computing capacity while adding new infrastructure.

That's a particularly strong combination for a business already generating hundreds of millions of dollars each quarter.

Is CoreWeave still the biggest AI cloud company?

CoreWeave remains substantially larger than the fast-growing AI cloud challengers examined here, with $2.58 billion in second-quarter 2026 revenue and approximately $104 billion in reported revenue backlog.

For comparison, Nebius generated $582.3 million in group revenue during that quarter. CoreWeave's quarterly revenue was approximately 4.4 times larger.

CoreWeave's infrastructure is also much further along. According to its August 2026 earnings release, the company reached 1.5 gigawatts of active power capacity after adding nearly 500 megawatts during the quarter. Its total contracted power reached approximately 3.7 gigawatts.

Adding nearly 500 megawatts of active power in three months is a major achievement. It demonstrates how quickly an established operator can expand once construction, electrical infrastructure, and GPU deployment are coordinated.

CoreWeave has also broadened its enterprise customer base. Its quarterly update named Bentley Systems, Caterpillar, Grammarly, and Isomorphic Labs among its customer relationships, alongside expanded work with companies such as Databricks, Cognition, and Runway.

CoreWeave is already publicly traded, so it no longer qualifies as a private startup in the strict sense. Still, it is the most useful benchmark for understanding how large a specialist AI infrastructure business can become.

Is Crusoe growing even faster than Nscale and Nebius?

Crusoe could be expanding its physical AI infrastructure faster than Nscale or Nebius, but its disclosed bookings and contracted capacity cannot establish a revenue-growth lead.

In September 2026, Crusoe announced the initial closing of a $3.9 billion Series F financing round at a $30.9 billion post-money valuation. The company also disclosed more than $140 billion in total contracted value, over 6 gigawatts of gross contracted capacity, and more than twentyfold year-over-year growth in Crusoe Cloud bookings.

The capacity expansion is particularly interesting. Crusoe reported 4.9 gigawatts of contracted capacity in June and more than 6 gigawatts by September. That is an increase of at least 1.1 gigawatts, or more than 22%, in roughly three months.

Crusoe also reported 1 gigawatt of gross capacity delivered and operational, meaning most of its contracted infrastructure was still awaiting deployment.

Its managed inference offering had secured more than $100 million in contracted annual recurring revenue. That gives us another indication of customer demand, although contracted ARR should not be confused with revenue already recognized.

Crusoe is preparing for enormous expansion. But declaring it the fastest-growing AI cloud startup based on contracts alone would be premature. We would need comparable recognized-revenue figures to determine whether it is already outperforming Nscale's percentage growth or Nebius's dollar growth.

Which AI data center startup has the most computing capacity running today?

CoreWeave has the clearest disclosed lead in active power capacity among the major specialists in this comparison, with 1.5 gigawatts reported in its second-quarter results.

Crusoe reported 1 gigawatt of gross operational capacity in September 2026. Nscale disclosed approximately 25,000 active GPUs out of 461,000 active and contracted GPUs, meaning only about 5.4% of its stated GPU total was operational at that point.

These figures reveal something that revenue growth alone cannot show. Nscale has secured a much larger future computing footprint than its active deployments suggest. Crusoe has also contracted several times more capacity than it has operational. Capacity signed but not yet built is the kind of gap we cover among the biggest unsolved problems in our data centers report.

The power numbers need some care because companies use different definitions. CoreWeave reports active power, Crusoe reports gross operational capacity, and Nscale combines active and contracted power in its headline infrastructure figure.

Electricity is increasingly central to this race. GPU purchases alone cannot create usable computing capacity without power connections, cooling equipment, substations, and functioning data centers.

Securing a site with access to electricity can therefore be as important as securing the Nvidia GPUs that eventually go inside it.

Company	Active or operational capacity	Contracted capacity	What the figures show
CoreWeave	1.5 GW active power	Approx. 3.7 GW total contracted power	About 41% of contracted power was active
Crusoe	1 GW gross operational	Over 6 GW gross contracted	Less than 17% of stated contracted capacity was operational
Nebius	No directly comparable active-power total in this comparison	5 GW contracted-power guidance	Large future deployment program
Nscale	Approx. 25,000 active GPUs	Approx. 461,000 active and contracted GPUs	About 5.4% of stated GPUs were active
Nscale power	Not separately comparable	Approx. 1.37 GW active and contracted	Most deployment detail remains tied to future capacity

The reported figures use different definitions and disclosure periods. They are useful for comparing the scale of commitments and operational progress, but do not form a standardized capacity ranking.

Will Nscale really make $103 billion from its AI data center contracts?

Nscale has secured approximately $103.4 billion in active and contracted total contract value, but converting those commitments into revenue will take years and require substantial additional infrastructure.

The company's September IPO filing showed that total contract value increased from approximately $38 billion at the end of 2025 to $103.4 billion by August 2026. That amounts to roughly $65 billion in additional commitments in eight months.

Only $2.6 billion was associated with active capacity, compared with approximately $500 million at the end of the previous year. Around 97.5% of the reported contract value therefore related to commitments beyond its active contracted base.

The contracts generally include take-or-pay provisions, and Nscale disclosed a weighted average contractual service life of approximately 5.7 years.

That structure gives the company much better revenue visibility than an ordinary business selling computing capacity on demand. Customers are making long-term commitments, often tied to infrastructure built specifically for them.

The difficult part will be fulfilling those agreements. Nscale must secure financing, complete construction, install equipment, and bring its facilities online before the majority of those future commitments can become recognized revenue. That's a lot of execution still ahead.

Can Nebius keep growing this fast with Microsoft and Meta as customers?

Nebius has a credible route to several more years of strong growth thanks to major customer contracts, although its extraordinary 500%-plus growth rates will become harder to maintain as the business gets larger.

In March 2026, Nebius announced an agreement under which Meta committed to purchase $12 billion of dedicated computing capacity over five years, with deliveries scheduled to begin in early 2027.

A separate arrangement covered additional available capacity worth up to $15 billion, while allowing Nebius to continue serving other AI cloud customers.

This structure is interesting because Nebius can benefit from a large customer commitment while still expanding its broader cloud business. The company has also secured substantial agreements with Microsoft and other major buyers.

According to its second-quarter shareholder materials, Nebius had more than $40 billion in customer commitments and increased its contracted-power guidance to 5 gigawatts.

Some of that business will only start producing revenue as new facilities open. The Meta agreement, for example, includes deliveries beginning in 2027.

The big question now is how quickly Nebius can turn these customer commitments into working computing capacity without losing the pricing and utilization benefits that helped drive its recent growth.

Is Lambda still one of the fastest-growing AI cloud startups?

Lambda remains one of the largest private AI cloud challengers, although its available revenue figures point to slower growth than Nscale or Nebius.

In April 2026, private-market research firm Sacra estimated that Lambda had reached approximately $760 million in annualized revenue at the end of 2025, compared with $425 million a year earlier. That represents estimated growth of 79%.

Lambda has since arranged substantial financing for new deployments. It announced a $1 billion secured credit facility in May 2026, a $926 million secured term loan in August, and another $1 billion institutional debt financing in October. Lambda lining up billions in debt for new deployments matters when we weigh the likely future winners in our data center market report.

The latest facility supports three committed deployments for two investment-grade customers. That suggests Lambda is increasingly serving large customers that want dedicated infrastructure, alongside its existing AI cloud offering.

These financing transactions provide evidence of expansion, although they cannot be treated as equivalent to additional revenue. Separate facilities may also have different terms, uses, and repayment obligations.

Lambda's estimated annualized revenue already puts it among the larger private specialists. But without more recent, comparable financial disclosures, we cannot place it ahead of Nscale or Nebius in the current growth ranking.

Are AI data center startups taking business away from AWS, Azure, and Google Cloud?

Specialized AI infrastructure providers are winning enormous contracts, but much of their growth is coming from the same technology companies that operate the world's largest cloud platforms.

Microsoft has worked with CoreWeave and Nebius. Meta has committed billions to Nebius. CoreWeave's customer relationships also include AI laboratories and enterprises that need specialized GPU infrastructure.

These companies often need more computing power than they can deploy internally within their preferred timeline. Buying capacity from specialists gives them additional locations, suppliers, and access to available GPUs.

There is little evidence that this amounts to a widespread abandonment of AWS, Microsoft Azure, or Google Cloud by ordinary enterprise customers.

Hyperscalers have become both customers and competitors of specialized AI infrastructure firms. They build their own data centers while purchasing capacity from independent providers.

That arrangement could last as long as AI infrastructure demand grows faster than the largest technology companies can satisfy it internally. It also leaves specialized providers exposed if those customers eventually become less dependent on outside capacity.

Are Nebius and CoreWeave making money from all this growth?

Nebius and CoreWeave are generating substantial adjusted EBITDA from their AI infrastructure businesses, but their reported financial results show how expensive it remains to build and finance these facilities.

Nebius reported $236.2 million in group adjusted EBITDA for the second quarter of 2026, compared with a $21 million adjusted EBITDA loss a year earlier. Its AI cloud segment generated approximately $285.7 million in adjusted EBITDA, with a margin close to 50%.

CoreWeave reported $1.51 billion in adjusted EBITDA for the same quarter, representing a 59% margin.

Those figures look excellent until we include expenses excluded from adjusted EBITDA.

CoreWeave reported $640 million in quarterly net interest expense and a $626 million net loss. Its adjusted operating income was only $128 million.

Nebius recorded a $190.4 million loss from continuing operations during the quarter, despite the large improvement in adjusted EBITDA.

Depreciation is particularly important here because GPUs are expensive assets that lose economic value as newer generations become available.

We can already see that both companies can generate substantial earnings before certain accounting and financing costs. Whether these businesses ultimately produce attractive returns on all the capital invested remains a harder question.

Could AI data center startups run out of money before finishing their projects?

Financing shortages could slow even the fastest-growing AI infrastructure companies because their spending requirements are enormous compared with the revenue they currently generate.

CoreWeave reported approximately $9.4 billion in capital expenditures during the second quarter of 2026, more than three times its quarterly revenue.

Nebius spent approximately $5.7 billion on property, equipment, and intangible assets during the same quarter, compared with $582.3 million in revenue.

Together, the two companies spent more than $15 billion on capital investments in just three months.

Nebius expects customer prepayments to provide significant support, disclosing anticipated prepayments exceeding $9 billion for 2026. It also secured approximately $775 million in asset-backed financing in July.

Nscale faces an especially demanding funding challenge because so much of its business remains tied to future deployments. Its September IPO filing disclosed that binding financing commitments had not yet been obtained for infrastructure required under certain major Anthropic agreements.

Crusoe has also been raising large amounts of capital, including its multibillion-dollar Series F financing.

The spending makes sense if contracted customers pay enough over the life of the equipment. Problems arise if construction runs late, financing gets more expensive, or newer GPUs reduce the commercial value of older installations before their costs have been recovered.

Are AI data center startups becoming too dependent on a few customers?

Customer concentration is a serious weakness for fast-growing AI infrastructure providers, especially when a single contract can represent billions of dollars in future revenue.

CoreWeave's first-half 2026 disclosures showed that approximately 93% of its revenue increase came from existing customers. This reflects strong demand from established relationships, but it also shows that the company's additional revenue depends heavily on customers already using its services.

Nscale's earlier business was even more concentrated. According to its IPO filing, one customer accounted for approximately 93% of revenue in 2024. One customer making up 93% of revenue is a real weak spot, one of the open problems we dig into in our data centers report.

Its subsequent contracts with major AI customers should broaden its commercial relationships over time, although the financial effects will depend on when those deployments begin operating.

Nebius has a similar challenge as it signs exceptionally large agreements with hyperscalers and AI developers. A few customers can account for a substantial part of future infrastructure demand.

Large take-or-pay agreements help protect providers against weak utilization once contracted services are available. But delays, changes in technical requirements, or disagreements over deployment schedules can still affect when revenue arrives.

The biggest danger would be several major customers slowing their expansion plans at the same time, particularly after infrastructure providers have already committed billions to construction and GPU purchases.

Can Nebius, Nscale, or Crusoe catch up with CoreWeave?

Nebius currently has the clearest path toward narrowing CoreWeave's revenue lead, while Nscale and Crusoe still need to turn much more of their contracted infrastructure into operational businesses.

CoreWeave generated approximately $2.58 billion in second-quarter 2026 revenue. Nebius produced $582.3 million, equivalent to roughly 23% of CoreWeave's quarterly total.

That is still a substantial gap. However, Nebius's first-half group revenue grew 529%, compared with CoreWeave's 112%.

Those growth rates cannot continue indefinitely, and Nebius's smaller starting base explains part of the difference. But the company is now adding hundreds of millions of dollars in revenue annually while expanding its capacity and securing multiyear customers.

Nscale is further behind in commercial scale, despite its exceptional percentage growth. Its first-half revenue averaged approximately $70 million per quarter, although that average does not capture its exact second-quarter exit rate.

Crusoe could become a major competitor given the size of its contracted infrastructure footprint. The missing piece is comparable evidence of how much recognized revenue its facilities are producing.

For now, Nebius is the most convincing challenger on actual financial growth. CoreWeave retains a major advantage in operating scale, experience, and established customer relationships.

What could bring the AI data center boom to a halt?

The biggest threats to AI data center growth are construction delays, rising financing costs, cheaper computing services, and customers reducing their planned spending.

The industry's current contracts assume that enormous amounts of computing capacity will be needed over several years. If demand develops more slowly than expected, some providers could end up with facilities or GPUs generating less revenue than originally projected.

Hardware economics could create a separate problem. Nvidia continues developing more powerful GPU systems, potentially making each generation more efficient than the last. That can help customers while placing pressure on the prices providers charge for older equipment.

Construction is another major risk. Delivering a large AI data center requires electrical connections, cooling, servers, networking equipment, permits, and often extensive grid upgrades. A delay in one part of the project can hold up the entire facility.

Companies with long-term take-or-pay agreements and strong financing arrangements are better protected, but even those contracts cannot eliminate every risk.

Some companies will almost certainly keep growing rapidly while others struggle to turn their announcements into working infrastructure. The winners will be those that can deliver capacity on schedule and recover their investments before the economics of their equipment change.

Which AI data center startup is growing the fastest today?

Nscale is currently the fastest-growing major AI data center startup by disclosed percentage revenue growth, while Nebius stands out for combining exceptional growth with a much larger commercial business.

Nscale's first-half revenue increased 1,252%, making it the numerical leader among the major providers with comparable public disclosures. Its smaller revenue base helps explain the extraordinary percentage, and much of its future business still depends on deploying contracted GPUs. A small revenue base makes the percentage look extraordinary; we look at who is growing the fastest in our data center market report.

Nebius grew its core AI cloud business 614% over the same period and reached nearly $1 billion in first-half AI cloud revenue. Its customer contracts and accelerating revenue make it a particularly strong contender for continued expansion.

CoreWeave remains much larger, generating $4.65 billion in first-half revenue, but its percentage growth is slower and it has been publicly traded since 2025.

Crusoe is the company to watch for physical infrastructure expansion. Its announced contracted capacity and customer commitments are enormous, although the available data does not establish a comparable revenue-growth lead.

Our judgment is that Nscale wins the current percentage-growth race, while Nebius has the strongest combination of growth and demonstrated commercial scale among the challengers. Crusoe could eventually overtake both in infrastructure deployed, but that outcome still depends on substantial construction and commissioning.

For now, Nscale is the clearest answer to which AI data center startup is growing the fastest. Nebius is the more convincing candidate for maintaining very rapid growth as the business becomes substantially larger.