List of Top 7 Neocloud Companies in the USA 2026

The top 7 neocloud companies in the USA in 2026 are CoreWeave, Crusoe, Lambda, Fluidstack, Vultr, WhiteFiber and Together AI. CoreWeave leads by a wide margin, with USD 5.13 billion of revenue in FY2025, more than five times the estimated revenue of Crusoe, the largest private US neocloud.

The United States is the center of the neocloud industry. It is home to the biggest GPU cloud providers, their largest customers (Microsoft, Meta, OpenAI, Anthropic and Google) and most of the new AI data center capacity being built. This guide ranks the seven largest US-headquartered neocloud companies by revenue and profiles each one's finances, GPU fleet, data centers, customers and expansion plans.

Why the USA Leads the Neocloud Market

• Customers are here. The biggest buyers of AI compute, including Microsoft, Meta, Google, Oracle, OpenAI and Anthropic, are US companies.

• NVIDIA backs US neoclouds. NVIDIA is an investor in CoreWeave, Lambda and Crusoe, and gives neoclouds early access to new GPU generations.

• Capital markets are deep. US neoclouds raise large GPU-backed loans and private rounds, and the Nasdaq and NYSE are the listing venues of choice, even for foreign neoclouds such as Nebius, IREN and Nscale.

• Power and land are available, at a price. Texas, New York, New Jersey, Ohio and North Carolina have become key build markets, though grid connections are now the main bottleneck.

This list covers US-headquartered companies. Foreign-headquartered neoclouds with large US operations, such as IREN, Nebius and Nscale, are covered under "Other neoclouds to watch" below.

Neocloud Companies in the USA: Top 7 at a Glance

Top 7 Neocloud Companies in the USA (2026)

1. CoreWeave

Founded in 2017 by a group of former commodities traders who first ran GPUs for Ethereum mining, CoreWeave has grown into the heavyweight of the American AI cloud market. The Livingston, New Jersey, company went public on Nasdaq (CRWV) in March 2025 and closed FY2025 with USD 5,131 million in revenue, a 168% jump on the prior year. Momentum has carried into 2026: management expects USD 12.4-13.2 billion in revenue this year and was sitting on roughly USD 104 billion of contracted backlog at the end of June.

That backlog is underpinned by one of the largest GPU estates outside the hyperscalers, with more than 250,000 NVIDIA GPUs spread across 51 data centers and over 1.5 GW of live power. Microsoft is by far its biggest client, contributing about two-thirds of 2025 revenue, with OpenAI and Meta also on the books and NVIDIA acting as both supplier and shareholder. To keep up with demand, CoreWeave is spending USD 35–39 billion on capacity in 2026 and is aiming for 8 GW of active power by 2030, including a planned USD 6 billion campus in Pennsylvania.

2. Crusoe

Denver-based Crusoe began in 2018 by turning flared natural gas into electricity for Bitcoin mining, then pivoted fully to AI after selling its mining arm to NYDIG in early 2025. Today it controls the whole stack, from sourcing power to building data halls to renting out GPUs and inference. As a private company it does not report results, but research firm Sacra estimates 2025 revenue near USD 998 million (other reporting suggested a lower internal projection of about USD 500 million), and investors valued it at USD 30.9 billion in a USD 3.9 billion Series F in September 2026.

Crusoe is best known as the builder behind Stargate's 1.2 GW campus in Abilene, Texas, where Oracle runs OpenAI workloads, and it is now putting up a second 900 MW campus nearby for Microsoft. The company says it has more than 1 GW live and over 6 GW contracted across Texas, Missouri, Wyoming and Alberta, backed by more than USD 140 billion in signed contract value. Meta, Google and trading firm Jane Street, which reportedly signed a USD 13 billion deal, are also among its customers, and its truck-mounted Spark modules let it deploy compute wherever power is available.

3. Lambda

Brothers Stephen and Michael Balaban started Lambda in San Francisco in 2012, building deep learning workstations long before the generative AI boom. In 2025 the company shut down its hardware line to bet entirely on GPU cloud, and the shift is paying off: The Information reported more than USD 520 million of revenue in the year to September 2025, and Bloomberg's sources expect 2026 revenue to top USD 1.5 billion. Lambda was last valued at USD 5.9 billion in its November 2025 Series E and has lined up banks for an IPO.

Lambda now operates 15 US data centers stocked with NVIDIA H100, H200, B200 and GB300 systems, and NVIDIA itself leases back around 18,000 of its GPUs. Its customer list has moved firmly upmarket: Microsoft signed a multi-billion-dollar deployment deal in late 2025, and Anthropic's own IPO filing shows commitments of about USD 41.3 billion to Lambda through 2034. The largest of those deals will be served from a 350 MW site in Nueces County, Texas, developed by Hut 8, as Lambda works toward its goal of 3 GW under management by 2030.

4. Fluidstack

Fluidstack was spun out of the University of Oxford in 2017 by Gary Wu and Jamie Cox as a marketplace for idle GPUs, and it relocated its headquarters to New York at the end of 2025 to chase US demand. The private company's last filed accounts show 2024 revenue of USD 66.2 million, and an investor memo projects roughly USD 200 million for 2025 rising to USD 660 million in 2026. Investors have priced in that growth, valuing Fluidstack at USD 18 billion in an April 2026 round led by Jane Street.

What sets Fluidstack apart is its role as Anthropic's data center partner under a USD 50 billion US build-out, making it the first known company outside Google to run Google TPU data centers at scale. Its pipeline includes 378 MW at TeraWulf's Lake Mariner campus in New York and a 168 MW joint venture in Abernathy, Texas, with Google guaranteeing debt on several projects. The investor memo puts its 2026 footprint at up to 1.3 GW across more than ten US sites, with an ambition to exceed 17 GW by 2030.

5. Vultr

Vultr, founded in 2014 by David Aninowsky and run today by CEO J.J. Kardwell, spent a decade as a self-funded, developer-friendly cloud before turning its attention to AI. The West Palm Beach, Florida, company does not publish financials; its most recent recorded revenue is about USD 150 million, logged when it raised USD 333 million at a USD 3.5 billion valuation in December 2024, and it was reported in 2026 to be seeking at least USD 1 billion in fresh capital.

Vultr's AI strategy leans on AMD rather than NVIDIA. AMD Ventures co-led its last round, and its flagship cluster in Springfield, Ohio, packs 24,000 AMD Instinct MI355X GPUs into a 50 MW facility, with a further 10,000-plus MI355X GPUs in Chennai, India. It also sells NVIDIA capacity, placed a USD 1.2 billion order with HPE for AMD Helios racks in September 2026, and spans 34 cloud regions in 17 countries, with a target of 1 GW of capacity by 2029.

6. WhiteFiber

WhiteFiber was carved out of Nasdaq-listed Bitcoin miner Bit Digital and floated on Nasdaq (WYFI) in August 2025, giving investors a pure play on GPU cloud and AI-ready colocation. The New York company, led by CEO Sam Tabar, booked audited FY2025 revenue of USD 79.2 million, up 66%, with GPU cloud services accounting for USD 68.8 million, and it grew revenue another 31% in the first quarter of 2026.

Rather than building from scratch, WhiteFiber retrofits existing buildings to bring capacity online faster. Its biggest US asset is the NC-1 campus in Madison, North Carolina, where Nscale has signed a 10-year, 40 MW colocation contract worth about USD 865 million; it also operates two sites in Montreal, one serving chipmaker Cerebras. With roughly 5,000 NVIDIA H100 and H200 GPUs under contract and about 76 MW targeted by the end of 2026, WhiteFiber says it is evaluating a pipeline of around 1.5 GW.

7. Together AI

Together AI was founded in San Francisco in 2022 by CEO Vipul Ved Prakash, CTO Ce Zhang and AI researchers Chris Ré, Percy Liang and Tri Dao, the creator of FlashAttention. It blends two businesses: a token-priced API for running open-source models and dedicated GPU clusters for training and fine-tuning. The company has not released annual revenue, but its annualized run-rate was reported at about USD 1 billion in early 2026, it says annual bookings passed USD 1.15 billion, and it raised USD 800 million at a USD 8.3 billion valuation in July 2026.

Originally a reseller of capacity from other neoclouds, Together AI now runs its own data centers in Maryland and Memphis and reports 200 MW of secured power, with NVIDIA Blackwell clusters coming online across North America. Its platform is used by fast-growing AI companies such as Cursor, Cognition and Decagon. Expansion is increasingly global: a 250 MW data center with HUMAIN in Saudi Arabia is set to roughly triple its capacity, and a 10,000-GPU B300 cluster is planned with L&T in Chennai, India.

Where US Neocloud Companies Are Building

Neocloud capacity in the US is concentrated where power and land are available quickly. Texas leads by a wide margin.

Power is now the main constraint. In Texas, an ERCOT pause on energizing new large loads during an audit has put parts of the 2026–2027 build schedule at risk.

Other Neoclouds to Watch in the USA

US-headquartered, run-rate figures only

• RunPod (Mount Laurel, New Jersey): a developer-focused GPU cloud that said it passed USD 325 million in ARR in October 2026, up from USD 120 million in January, with more than 1.2 million developers.

• Lightning AI (merged with Voltage Park in January 2026): combines PyTorch Lightning software with Voltage Park's 35,000+ NVIDIA GPUs across six US data centers; the combined company reported more than USD 500 million in ARR.

Foreign-headquartered, with large US operations

• IREN (Sydney): FY2026 revenue of USD 707.0 million, with its AI cloud built in Childress and Sweetwater, Texas, and a USD 9.7 billion Microsoft contract.

• Nebius (Amsterdam): FY2025 revenue of USD 529.8 million; serves Microsoft from Vineland, New Jersey, and runs a Kansas City cluster.

• Nscale (London): FY2025 revenue of USD 33.0 million; building in Ward County, Texas, and West Virginia, with a USD 45 billion Anthropic contract for the West Virginia campus.

US Neocloud Trends in 2026

• Frontier labs are signing record contracts. Anthropic's filings show about USD 41.3 billion committed to Lambda, on top of its USD 50 billion US data center program with Fluidstack.

• The IPO pipeline is filling. After CoreWeave and WhiteFiber, Lambda and Crusoe are preparing listings, and private valuations have climbed sharply: Crusoe to USD 30.9 billion, Fluidstack to USD 18 billion.

• Bitcoin miners are becoming AI landlords and clouds. Hut 8, TeraWulf and Cipher Mining now host neocloud capacity, and IREN is converting its own mining sites.

• Alternatives to NVIDIA are growing. Vultr is built around AMD Instinct GPUs, and Fluidstack is building Google TPU data centers for Anthropic.

• Power, not chips, is the bottleneck. Grid interconnection queues and Texas's ERCOT energization pause now set the pace of new capacity.

Conclusion

CoreWeave is the largest neocloud company in the USA in 2026. Crusoe and Lambda lead the private challengers, Fluidstack has become Anthropic's main data center builder, and Vultr, WhiteFiber and Together AI round out the top seven. With record contracts signed and IPOs on the way, the question for 2026–2027 is how quickly these companies can get power connected and turn their backlogs into running GPU capacity, much of it in Texas.