Which AI infrastructure startup is growing the fastest?

Why are AI infrastructure startups growing so fast right now?

AI infrastructure startups are growing at extraordinary speeds because companies need far more computing power to train AI models and run them for millions of users. The clearest evidence comes from actual sales: CoreWeave generated $4.65 billion in the first half of 2026, Nebius reached $981 million, and Nscale increased its revenue more than thirteenfold.

The industry is also changing. A few years ago, much of the demand came from companies training large AI models. Today, a growing share comes from running those models continuously inside products, coding assistants, customer support systems, and other applications.

That change is creating opportunities at several levels. CoreWeave and Nebius rent powerful GPU computing infrastructure. Nscale builds and operates large computing facilities through long-term customer contracts. Fireworks AI and Together AI help developers run AI models more efficiently.

The growth figures are striking. Nebius reported 454% year-over-year revenue growth in the second quarter of 2026, compared with CoreWeave's 112%. Nscale disclosed a 1,252% increase for the first half of the year.

Yet these companies are at very different stages. Some already generate billions in sales, while others have secured enormous contracts for facilities that are still being built. That difference is central to identifying which startup is really growing the fastest.

What should we measure to find the fastest-growing AI infrastructure startup?

Revenue growth is currently the best way to identify the fastest-growing AI infrastructure startup, although we also need to look at how much additional revenue each company generates. Funding rounds, valuations, and data center announcements can tell us something about future growth, but they are less convincing than actual sales.

Consider two companies. One increases annual revenue from $10 million to $140 million, while another goes from $2 billion to $4.6 billion. The first has grown thirteenfold, but the second has created $2.6 billion in additional business.

Both achievements are impressive for different reasons. The smaller company is growing faster as a percentage, while the larger company is expanding faster in absolute dollars.

There is also a problem with private-company reporting. CoreWeave and Nebius publish financial statements, while Fireworks AI and Together AI mainly disclose annualized revenue run rates. A run rate measures the annual revenue implied by activity at a particular moment, which can differ considerably from revenue actually earned over twelve months.

For our comparison, we prioritize reported year-over-year revenue growth, then examine absolute revenue gains and the amount of business already running. We also separate private startups from publicly traded companies, even though the latter remain useful competitors to study.

Which AI infrastructure startups have the highest revenue growth?

Nscale has the highest documented revenue growth rate among the major AI infrastructure companies we examined, followed by Nebius. CoreWeave is growing more slowly in percentage terms, but its absolute revenue gains are much larger.

Nscale's September 2026 SEC filing reported a 1,252% increase in first-half revenue. Nebius reported 529% growth over the same six-month period, while CoreWeave increased second-quarter revenue by approximately 112%.

Fireworks AI and Together AI deserve attention too. Fireworks announced that its annualized revenue run rate exceeded $1 billion in July 2026. Sacra estimates that Together AI reached a similar level in February.

These figures put five companies at the center of the race, although Fireworks and Together AI have not disclosed the same level of financial detail as the others.

Company	Revenue growth	Reported commercial scale	Data source
Nscale	+1,252% YoY, H1 2026	$140.6M first-half revenue	SEC filing
Nebius	+529% YoY, H1 2026	$981.3M first-half revenue	SEC filing
CoreWeave	+112% YoY, Q2 2026	$2.58B quarterly revenue	SEC filing
Fireworks AI	From reported $280M to over $1B annualized run rate in under 9 months	$1B+ annualized run rate	Company announcement and historical estimates
Together AI	Approximately +375% YoY in 2025	~$1B annualized run rate in February 2026	Sacra estimates
Crusoe	Comparable group revenue growth undisclosed	Managed inference contracted annual revenue above $100M in 2026	Company announcement
Lambda	Comparable recent growth undisclosed	Private GPU cloud provider	Limited public financial data

Is Nscale really the fastest-growing AI infrastructure startup?

Nscale is currently the fastest-growing AI infrastructure startup we can identify using comparable published revenue figures. Its sales increased from $10.4 million in the first half of 2025 to $140.6 million in the first half of 2026, a 1,252% jump.

The acceleration becomes even more striking when we examine the previous years. Nscale generated $19.1 million in 2024 and $33 million in 2025, a 73% increase. Its first six months of 2026 then produced more than four times its entire 2025 revenue.

Something clearly changed in the pace of the business.

Nscale has been signing enormous computing infrastructure contracts with major technology customers. According to its SEC registration statement, active and contracted total contract value climbed from $38 billion at the end of 2025 to $103.4 billion by August 2026. That represents roughly 172% growth in eight months.

However, Nscale remains relatively small in actual revenue. CoreWeave generated approximately 33 times as much revenue during the first half of 2026.

The company's 1,252% growth rate is therefore both exceptional and influenced by a low starting base. We consider Nscale the percentage-growth leader, but its next challenge is to repeat that performance as its revenue becomes much larger.

If you want more recent data on this point, see the latest figures in our AI Infrastructure report ↓.

Is Nebius growing faster than CoreWeave?

Nebius is growing roughly four times faster than CoreWeave in percentage terms, even though CoreWeave is adding substantially more revenue. Their second-quarter financial results make the difference unusually clear.

Nebius generated $582.3 million in the second quarter of 2026, compared with $105.1 million a year earlier. That is 454% growth, representing approximately $477 million in additional quarterly revenue.

CoreWeave increased revenue from $1.21 billion to $2.58 billion over the same period, a 112% increase. Its additional quarterly revenue reached approximately $1.36 billion.

CoreWeave added nearly three times more revenue in absolute dollars while Nebius grew about four times faster as a percentage.

The six-month comparison confirms the pattern. Nebius added approximately $825 million in revenue, whereas CoreWeave added about $2.46 billion.

We are looking at two very different stages of development. CoreWeave already has a multibillion-dollar business and continues expanding quickly. Nebius is still catching up, but the gap is narrowing as its new computing capacity comes online.

Is Nebius the fastest-growing AI cloud company at scale?

Nebius has the strongest claim to being the fastest-growing AI cloud company at substantial revenue scale among those examined. Its first-half revenue increased 529% to $981.3 million, making its growth difficult to explain away as a small startup benefiting from a low starting point.

The quarterly progression is particularly revealing. Revenue increased from $50.9 million in the first quarter of 2025 to $105.1 million in the second quarter. One year later, those figures had reached approximately $399 million and $582 million.

Nebius therefore added roughly $183 million of revenue between the first and second quarters of 2026 alone, equivalent to 46% sequential growth.

There is also evidence that the business is becoming more efficient as it expands. Nebius reported $236.2 million in adjusted EBITDA for the second quarter, compared with an adjusted EBITDA loss of $21 million a year earlier.

That improvement came despite the company remaining unprofitable on a conventional operating basis, with large depreciation expenses from its computing infrastructure.

Nebius stands out because it combines several hundred percent revenue growth with nearly $1 billion in six-month sales. Few emerging AI cloud companies have demonstrated both at the same time.

If you want more recent data on this point, see the latest figures in our AI Infrastructure report ↓.

Why is CoreWeave still so far ahead of its AI cloud competitors?

CoreWeave remains the commercial heavyweight of specialized AI cloud infrastructure because it has already turned demand into billions of dollars in computing services. Its second-quarter 2026 revenue reached $2.58 billion, more than four times Nebius's revenue and vastly above Nscale's current sales.

The company's revenue trajectory shows how quickly it has expanded. Quarterly sales rose from $1.21 billion in the second quarter of 2025 to $1.37 billion in the third, $1.57 billion in the fourth, $2.08 billion in the first quarter of 2026, and $2.58 billion in the second.

That represents more than $1.3 billion of additional quarterly revenue in one year.

CoreWeave has also built considerable physical infrastructure. Its active power capacity reached approximately 1.5 gigawatts by the second quarter of 2026, with around 3.7 gigawatts of total contracted power.

The company has expanded beyond major AI laboratories, announcing relationships with enterprises including Caterpillar, Bentley Systems, and Grammarly.

CoreWeave's 2025 IPO means it no longer qualifies as a private startup. Still, it provides a useful benchmark: any challenger claiming to be the fastest-growing AI infrastructure business needs to explain how its growth compares with a company already adding billions in annual revenue.

Is Fireworks AI growing faster than the big GPU cloud companies?

Fireworks AI is one of the fastest-growing AI inference infrastructure businesses, with its annualized revenue run rate exceeding $1 billion in July 2026. Its reported trajectory is extraordinary, although we cannot directly compare those figures with the recognized revenue reported by GPU cloud providers.

Historical figures compiled by ARR.Club indicate that Fireworks reached approximately $200 million in annualized revenue in July 2025, $280 million in October, and $800 million in May 2026.

The increase from $280 million to more than $1 billion represents at least 257% growth in less than nine months.

Fireworks is benefiting from companies running increasingly large numbers of AI requests. When it announced its $1.505 billion Series D in July 2026, the company said its infrastructure was processing more than 40 trillion tokens per day.

More interestingly, over 95% of those tokens reportedly came from models specialized for particular tasks rather than general-purpose models used without customization.

That suggests Fireworks is finding demand among companies building AI into their own products and workflows, rather than relying exclusively on a small number of frontier model developers.

We cannot establish whether Fireworks is growing faster than Nscale or Nebius because its quarterly recognized revenue remains undisclosed. But among private inference platforms, it is clearly one of the companies worth watching most closely.

Is Together AI keeping up with Fireworks AI?

Together AI is keeping up with Fireworks AI in reported commercial scale, although Fireworks appears to have reached its latest billion-dollar annualized revenue milestone through a particularly sharp acceleration. Both companies are now around $1 billion in annualized revenue.

According to Sacra, Together AI reached approximately $1 billion in annualized revenue in February 2026, up from about $618 million at the end of 2025. Its estimated annual bookings exceeded $1.15 billion by the second quarter.

Together AI operates across two businesses. Developers pay to run models through its APIs, while other customers rent GPU computing capacity for training, fine-tuning, and inference.

Sacra estimates that APIs account for roughly 30-40% of Together AI's revenue, leaving GPU rentals as the larger business.

Fireworks places greater emphasis on specialized model serving and inference optimization. Its rapid growth suggests that developers increasingly want infrastructure tailored to their models rather than generic access to GPUs.

Together AI is also moving toward operating more of its own infrastructure, including through a major Saudi Arabian partnership announced in August 2026 involving 250 megawatts of power and 120,000 chips.

The difference between these companies is becoming more interesting than their headline revenue figures. Fireworks is expanding through specialized inference services, while Together AI combines inference software with a much larger GPU rental operation.

Which AI infrastructure company has the most business already contracted?

CoreWeave and Nscale currently stand out for the enormous size of their disclosed future customer commitments, with each reporting contract portfolios above $100 billion. However, the two companies are at very different stages of turning those commitments into revenue.

CoreWeave reported approximately $104 billion in revenue backlog at the end of June 2026, excluding more than $25 billion in additional customer commitments announced shortly afterward.

Nscale disclosed $103.4 billion in active and contracted total contract value at the end of August, up from $38 billion eight months earlier.

One detail in Nscale's filing deserves particular attention. Of its $103.4 billion total, approximately $2.6 billion was classified as active contract value. That means active contracts represented only about 2.5% of the combined figure.

A large proportion of Nscale's potential business still depends on future infrastructure delivery.

Nscale's average contract duration was approximately 5.7 years. Dividing its combined contract value by that duration gives roughly $18.1 billion per year, but that calculation should never be mistaken for a revenue forecast. Different contracts begin at different times, and some depend on construction and financing conditions.

CoreWeave has a much stronger record of converting contracts into sales. Its first-half revenue reached $4.65 billion, against Nscale's $140.6 million.

Their contract measures also differ, preventing a perfectly direct comparison. Still, Nscale's rapid increase in commitments suggests it could become a much larger competitor if its facilities are delivered successfully.

Could Nscale overtake CoreWeave with its $103 billion in contracts?

Nscale could eventually challenge CoreWeave's revenue leadership, but the company's existing sales are still far too small to support a prediction that it will overtake its rival soon. Nscale generated $140.6 million in first-half revenue, compared with CoreWeave's $4.65 billion.

Closing a 33-fold revenue gap would require an enormous amount of infrastructure to become operational.

According to Nscale's filings, the company had approximately 461,000 GPUs active or contracted by August 2026. Its customer agreements could support massive revenue once the underlying facilities are ready, but the distinction between contracted and operational GPUs remains important.

Financing is another obstacle. Reuters reported that Nscale recorded a $1.02 billion net loss during the first half of 2026. Part of that loss came from accounting adjustments, but the company still faces substantial construction, equipment, and financing requirements.

The Financial Times has also reported on conditions attached to major customer agreements, including the risk of losing contracts if financing or delivery requirements are missed.

CoreWeave has already proved it can deliver computing services at multibillion-dollar scale. Nscale has proved that large customers are willing to make enormous commitments to its platform.

Those commitments are growing much faster than Nscale's current revenue base. The next few reporting periods will tell us how quickly that gap can close.

If you want more recent data on this point, see the latest figures in our AI Infrastructure report ↓.

Are AI infrastructure startups too dependent on a few giant customers?

Nscale remains heavily dependent on a small number of major customers, even as its revenue grows rapidly. Its SEC filings show that its largest customer accounted for 52% of first-half 2026 revenue, down from 73% in 2025.

That improvement is meaningful. Nscale has started broadening its customer base, whereas one customer accounted for substantially all its revenue in 2024.

Still, a business generating more than half its sales from a single customer faces considerable concentration risk. A delayed deployment, renegotiated agreement, or cancelled expansion could significantly affect revenue growth.

CoreWeave also relies heavily on major AI and technology companies, although its newer enterprise relationships suggest a gradual expansion into different customer groups. Nebius has secured large computing agreements while developing services for a broader range of AI developers and businesses.

The economics of this market explain the concentration. Only a relatively small group of customers can commit billions of dollars to computing infrastructure.

But we should distinguish between growth generated by one enormous new customer and growth supported by hundreds of expanding accounts. Both create revenue, although the second is generally less vulnerable to a single commercial relationship.

Nscale's customer concentration has improved considerably. Whether that continues as its largest contracts begin generating revenue will be worth monitoring.

Are the fastest-growing AI infrastructure startups actually making money?

The fastest-growing AI infrastructure companies are still struggling to turn rapid revenue growth into conventional net profits. Nscale, Nebius, and CoreWeave have all reported substantial losses, although their adjusted operating results differ considerably.

Nscale recorded a $1.02 billion net loss in the first half of 2026 against $140.6 million in revenue. Significant valuation-related accounting charges contributed to the loss, so it should not be treated entirely as operational cash spending.

CoreWeave generated approximately $1.51 billion in adjusted EBITDA during the second quarter, but also reported a $567 million adjusted net loss. Large depreciation and financing costs remain central to its business.

Nebius reported $236.2 million in adjusted EBITDA on $582.3 million in quarterly revenue. That represents an adjusted EBITDA margin of roughly 41%, although the company also reported a $190.4 million net loss from continuing operations.

These results reflect the enormous cost of building AI infrastructure. Companies often need to purchase computing equipment and secure power long before they earn the revenue expected from those investments.

Private competitors disclose much less information. Fireworks AI and Together AI have reported or estimated annualized revenue around $1 billion, but there is insufficient public information to compare their net profitability reliably.

Company	Recent revenue	Earnings indicator	What remains unclear
Nscale	$140.6M H1 2026	$1.02B net loss	Future costs as contracted projects become operational
Nebius	$582.3M Q2 2026	$236.2M adjusted EBITDA	Path to sustained net profitability
CoreWeave	$2.58B Q2 2026	$1.51B adjusted EBITDA	Impact of depreciation and financing costs
Fireworks AI	$1B+ annualized run rate	Undisclosed	Actual quarterly profits and cash flow
Together AI	~$1B annualized run rate	Undisclosed	Profitability across APIs and GPU rentals
Crusoe	Comparable group revenue undisclosed	Undisclosed	Economics across its infrastructure businesses
Lambda	Recent comparable revenue undisclosed	Undisclosed	Margins and cash requirements

Can Nscale keep growing revenue by more than 1,000%?

Nscale probably cannot maintain annual revenue growth above 1,000% for many years, but its existing contracts leave considerable room for another period of exceptional expansion. The next challenge is whether the company can increase revenue by billions rather than hundreds of millions.

Its first-half revenue increased from $10.4 million in 2025 to $140.6 million in 2026. Repeating that same thirteenfold expansion would require approximately $1.9 billion in first-half revenue during the following year.

That is an enormous increase, but Nscale already has customer commitments that could eventually support a business of that size.

The difficulty lies in delivering the infrastructure quickly enough.

Nscale's reported active or contracted GPU count increased from approximately 284,000 in April 2026 to 461,000 by August. That represents roughly 62% growth in four months, although the figures include equipment that has not necessarily been deployed.

A company can sign new computing contracts much faster than it can build data centers, connect electricity, install cooling systems, and deploy GPUs.

Nscale has already demonstrated that demand for its infrastructure exists. Now we need several consecutive reporting periods showing that actual sales are catching up with its contracts.

Even a drop from 1,252% to 200% annual growth would still represent very fast expansion once revenue reaches a much larger base.

Could Nebius eventually become bigger than CoreWeave?

Nebius could become a serious rival to CoreWeave in total revenue, but CoreWeave remains considerably ahead today. Nebius is investing aggressively in new infrastructure, and its current growth rate suggests the competitive gap could narrow further.

Nebius reported approximately $8 billion in cash at the end of June 2026 and raised its year-end contracted-power guidance to more than five gigawatts.

CoreWeave reported approximately 3.7 gigawatts of contracted power and 1.5 gigawatts of active power around the same period.

The figures suggest Nebius is building a potentially larger future power pipeline, although contracted power does not indicate how much computing capacity is already operating.

Nebius also reported four major second-quarter deals averaging more than $1 billion each. These agreements should provide a substantial foundation for future expansion.

CoreWeave, meanwhile, benefits from an established operating network, existing customer relationships, and the experience of delivering computing services at a far larger scale.

What makes Nebius particularly interesting is the combination of its financial resources, rapid sales growth, and increasing infrastructure commitments. It has the ingredients to become one of the largest specialized AI cloud providers.

Overtaking CoreWeave will require more than maintaining a higher growth percentage. Nebius needs to keep adding revenue faster in absolute dollars, which CoreWeave currently does by a wide margin.

If you want more recent data on this point, see the latest figures in our AI Infrastructure report ↓.

Could inference startups like Fireworks AI grow bigger than GPU cloud providers?

Fireworks AI and Together AI could eventually build businesses as large as major GPU cloud providers, especially if demand for running AI models continues increasing. But their current revenue remains below CoreWeave's, and it is too early to say that inference platforms will become the larger business model.

These companies serve related but different needs.

GPU cloud providers mainly sell computing capacity. Inference platforms make it easier to deploy AI models, manage requests, optimize speed, and reduce the cost of generating responses.

Fireworks's reported processing volume of more than 40 trillion tokens per day shows just how large this activity has become. Together AI's estimated API revenue share of 30-40% also indicates substantial demand for services beyond raw GPU rentals.

The distinction may become less clear over time.

Together AI is investing in its own computing infrastructure, while Nscale has expanded into inference services and acquired Anyscale to strengthen its software capabilities. CoreWeave has also been developing more software around its computing platform.

Infrastructure companies are moving toward the same customers from different directions.

That makes inference one of the most interesting parts of the market to follow. Still, CoreWeave's billions in quarterly revenue show that supplying computing capacity remains an enormous business.

So which AI infrastructure startup is growing the fastest?

Nscale is currently the fastest-growing AI infrastructure startup based on publicly documented percentage revenue growth among the major companies examined. Its 1,252% first-half revenue increase puts it well ahead of the other companies with comparable disclosed figures.

Nebius deserves a separate distinction. Its revenue increased 529% to $981.3 million over the same six-month period, making it the strongest example of exceptional growth at a substantial revenue scale.

CoreWeave remains the leader in absolute revenue expansion, adding approximately $2.46 billion in first-half revenue compared with the previous year. It is also far larger than the other companies examined, although its public listing means it is no longer a private startup.

Fireworks AI stands out among private inference specialists, having surpassed $1 billion in annualized revenue after a particularly rapid expansion. Together AI has reached a similar estimated scale, but both companies disclose too little standardized financial information for us to rank their actual growth against Nscale confidently.

Our conclusion depends on the measure, but the evidence is clear enough to name the leaders. Nscale wins on percentage growth, Nebius on rapid growth at meaningful scale, and CoreWeave on additional dollars of revenue.

Nscale is our answer to the title. Its next challenge will be turning an enormous portfolio of future customer commitments into operating infrastructure and sustained revenue. That is where the race between the fastest-growing AI infrastructure companies will become much more revealing.

If you want more recent data on this point, see the latest figures in our AI Infrastructure report ↓.