01 Industry overview
The neocloud industry at a glance
The neocloud industry exceeded $25 billion in revenue in 2025 and is forecast to approach $400 billion by 2031, driven by specialized GPU infrastructure for AI training, inference and enterprise workloads.
- $25B+[21]
- Global revenue, 2025
- 58%[21]
- Forecast CAGR to 2031
- $5.1B[11]
- CoreWeave revenue, 2025
- 850MW+[11]
- CoreWeave active power, 2025
- Market size
- Global neocloud revenue exceeded $25 billion in 2025, according to Synergy Research Group. CoreWeave, the largest publicly reported specialist provider, generated $5.131 billion in revenue in 2025.[21][11]
- Forecast & CAGR
- Synergy Research Group forecasts the neocloud market to approach $400 billion by 2031, implying a 58% compound annual growth rate. A separate William Blair estimate places 2031 revenue at $200 billion to $300 billion, with a base case of approximately $250 billion.[21][15]
- Market share & regions
- The available sources do not provide a consistent global revenue breakdown by region. CoreWeave reported more than 850MW of active power across 43 data centers globally at the end of 2025, while its disclosed footprint and major customer base remain heavily linked to North American AI infrastructure demand.[11]
- Segmentation
- Neocloud services are segmented into GPU-as-a-service, AI model training, inference, agentic AI, enterprise AI platforms and specialized high-performance computing. William Blair identifies GPU-first cloud specialists such as CoreWeave, Crusoe, Lambda, IREN, Nebius and Nscale, alongside inference-focused providers such as Together AI, OpenRouter, Fireworks AI and Baseten.[11][15]
- Key players & concentration
- CoreWeave reported $5.131 billion of 2025 revenue and $66.8 billion of revenue backlog at December 31, 2025. Customer concentration is high: Microsoft accounted for approximately 71% of CoreWeave revenue in the three months ended June 30, 2025, although the company serves AI labs, hyperscalers and enterprises.[11]
- Pricing
- Available market evidence places GPU rental pricing at roughly $1.70 to $3.90 per GPU-hour for bare-metal and spot offerings, $6 to $8.60 for premium branded providers, and $8 to $20 for software and orchestration layers. Pricing varies materially by GPU generation, reservation term, utilization, networking and support.[22]
- Trends, drivers & challenges
- Demand is being driven by AI model training, inference, agentic workloads, enterprise adoption, hyperscaler outsourcing and sovereign compute. The main constraints are power availability, grid access, data-center construction, GPU financing and utilization risk; CoreWeave reported $2.929 billion of technology and infrastructure expense in 2025, equal to 57% of revenue.[15][11][19]
- Consumer behavior & competition
- The principal buyers are AI laboratories, hyperscalers, enterprises and developers rather than traditional household consumers. Competition is based on GPU availability, deployment speed, performance, reliability, networking, geographic placement and price; CoreWeave reported that approximately 85% of its 2025 revenue increase came from expansion within its existing customer base.[11][15]
- Regulation
- Regulatory pressure is increasingly focused on the electricity, land and environmental impacts of AI data centers. In July 2026, New York imposed a one-year pause on new large data-center construction while it assessed energy and climate risks; broader U.S. and local policy debates also address grid access, emissions, water use and permitting.[19][23]
- Latest news
- In November 2026, NVIDIA disclosed $32 billion in agreements with neocloud providers that combine hardware sales with participation in cloud-rental revenue above guaranteed minimums. The structure highlights the industry’s dependence on NVIDIA financing, supply commitments and accelerator economics.[24]
02 Ranking
Top 10 Companies in the Neocloud Industry
The leading companies in the neocloud industry, ranked on market impact, performance, momentum, innovation and brand popularity.
Updated November 20, 2026 · Next review December 20, 2026
Crusoe stands out for AI infrastructure across training, fine-tuning and inference because its Crusoe Cloud workload coverage is explicit, while Fluidstack is the pick for dedicated AI data centers and GPU cloud since it deploys scaled AI infrastructure for frontier labs, governments and enterprises.
Ranked by Overall score · How we score ↓
Editor's top 3 picks
Where to start, by need.
#1 Overall · Leader in GPU cloud infrastructure
Crusoe
US
Ranked #1 · 9.4 · GPU cloud infrastructure
Why we picked it: It sells AI infrastructure across training, fine-tuning, and inference workloads, with revenue growth of 81.6%.
#2 overall
Fluidstack
US
Ranked #2 · 9.3 · GPU cloud infrastructure
Why we picked it: It focuses on dedicated AI data centers and GPU cloud delivery for frontier labs, governments, and enterprises, with 139.5% revenue growth.
#3 overall
Nscale
GB
Ranked #3 · 9.1 · GPU cloud infrastructure
Why we picked it: It builds vertically integrated AI infrastructure and private AI cloud environments, with 72.8% revenue growth.
Pick 1 is the #1 company overall. No other segment places in the top 5, so picks 2 and 3 follow rank.
Full ranking
| # | Company | Overall ↓ | Impact | Performance | Momentum | Innovation | Company snapshot |
|---|---|---|---|---|---|---|---|
| 1 | Crusoe GPU cloud infrastructure · US | 9.4 | 8.8 | 9.9 | 9.9 | 9.1 | |
Company snapshot · Crusoe · Data as of November 20, 2026 Crusoe Cloud provides AI infrastructure for training, fine-tuning and inference workloads.[25]
| |||||||
| 2 | Fluidstack GPU cloud infrastructure · US | 9.3 | 8.7 | 9.9 | 9.9 | 9.1 | |
Company snapshot · Fluidstack · Data as of November 20, 2026 Fluidstack deploys scaled AI infrastructure for frontier labs, governments and enterprises.[26]
| |||||||
| 3 | Nscale GPU cloud infrastructure · GB | 9.1 | 8.8 | 9.9 | 9.4 | 9.1 | |
Company snapshot · Nscale · Data as of November 20, 2026 Nscale builds vertically integrated AI infrastructure and private AI cloud environments.[27]
| |||||||
| 4 | CoreWeave GPU cloud infrastructure · US | 9.1 | 9.9 | 8.0 | 9.9 | 7.0 | |
Company snapshot · CoreWeave · Data as of November 20, 2026 CoreWeave operates the CoreWeave Cloud Platform for AI and high-performance workloads.[28]
| |||||||
| 5 | Vultr GPU cloud infrastructure · US | 9.0 | 8.8 | 8.9 | 9.2 | 9.1 | |
Company snapshot · Vultr · Data as of November 20, 2026 Vultr Cloud GPU provides on-demand NVIDIA and AMD GPU infrastructure for AI workloads.[29]
| |||||||
| 6 | IREN Large-scale AI cloud and data centers · AU | 9.0 | 8.8 | 8.0 | 9.9 | 9.9 | |
Company snapshot · IREN · Data as of November 20, 2026 IREN Cloud delivers large-scale GPU clusters for AI training and inference.[30]
| |||||||
| 7 | Runpod Developer-focused GPU cloud and serverless GPUs · US | 8.9 | 8.8 | 8.9 | 9.2 | 8.2 | |
Company snapshot · Runpod · Data as of November 20, 2026 Runpod offers GPU instances, serverless GPUs and clusters for AI development and production.[31]
| |||||||
| 8 | Applied Digital AI and HPC data centers and cloud services · US | 8.5 | 8.5 | 8.0 | 9.4 | 8.2 | |
Company snapshot · Applied Digital · Data as of November 20, 2026 Applied Digital develops and operates digital infrastructure and cloud services for AI and HPC.[32]
| |||||||
| 9 | Gcore Global GPU cloud, inference and edge AI · LU | 8.5 | 7.6 | 9.9 | 9.4 | 8.2 | |
Company snapshot · Gcore · Data as of November 20, 2026 Gcore offers GPU Cloud, AI inference and AI Cloud Stack services.[33]
| |||||||
| 10 | Civo Developer cloud and GPU instances · GB | 7.1 | 7.0 | 7.0 | 7.0 | 7.3 | |
Company snapshot · Civo · Data as of November 20, 2026 Civo offers NVIDIA GPU instances for AI and machine-learning training and inference.[34]
| |||||||
How we score
- Impact (30%): scale and weight in the market (revenue, market share, employees, market position).
- Performance (20%): recent revenue growth and operating margin; an analyst assessment where a company does not publish them.
- Momentum (20%): growth trajectory and future potential.
- Innovation (15%): R&D intensity and how technologically advanced the company is.
Popularity (brand search demand, 15%) counts toward Overall but is not published as a column. Overall leaves out what is not known.
Scale Scores are shown on a 7–10 scale with one decimal; bars start at 7.0.
Cadence Recalculated when new figures are published; next review December 20, 2026.
Independence. Rankings are editorially independent. Scores, ranks, picks and the verdict follow the method above; no company can pay to change them.
Cite this ranking
APA
Samuel Norberg. (2026, November 20). Top 10 companies in the neocloud industry. In Neocloud Industry Statistics. Gitnux. https://gitnux.org/neocloud-industry-statistics/#top-companies
Worldwide neocloud and Tier-2 GPU rental market size
Worldwide neocloud and Tier-2 GPU rental market size grew from $1.2B in 2023 to $12.0B in 2025.
03 Trends & forecasts
Top 10 Neocloud Industry Trends & Forecasts 2027
What is changing in the neocloud industry, with the evidence behind each trend and a forecast for 2027.
Forecasts were generated by an AI model from the cited evidence. They are estimates, not guarantees.
01Neocloud revenue accelerates toward $400 billion
High confidenceSynergy Research Group reported that neocloud revenue exceeded $25 billion in 2025, including $9 billion in fourth-quarter revenue, up 223% year over year. It forecasts the market to approach $400 billion by 2031.[21]
Forecast 2027: Revenue should continue expanding rapidly in 2027, but the market is unlikely to maintain the exceptional 2025 growth rate indefinitely. A plausible outcome is another year of high-double-digit growth as inference, enterprise AI and hyperscaler outsourcing broaden demand.
02Inference overtakes training as the growth engine
High confidenceJLL identifies 2027 as the point at which AI inference workloads are expected to overtake training workloads. Gartner separately forecasts inference-driven AI infrastructure spending to grow 55% in 2026 and reach a $66 billion market in 2027.[35][3]
Forecast 2027: In 2027, neocloud providers should allocate more capacity to low-latency inference, model serving and retrieval workloads rather than exclusively to large training clusters. This should increase demand for geographically distributed sites, optimized networking and predictable availability.
03Agentic workloads raise CPU intensity
Medium confidenceIntel said in its 2026 first-quarter earnings discussion that agentic AI could shift data-center deployment ratios from roughly one CPU for every four to eight GPUs toward approximately one CPU per GPU. The change reflects the greater orchestration, tool use and monitoring requirements of agentic systems.[24]
Forecast 2027: By 2027, neocloud cluster design should become less GPU-only, with materially greater spending on CPUs, memory, storage and interconnects per GPU. Providers that price complete agentic workloads rather than isolated GPU hours should gain an advantage.
04Capacity expands through hyperscaler outsourcing
High confidenceWilliam Blair estimates that global neocloud revenue could reach $200 billion to $300 billion by 2031, supported partly by hyperscaler outsourcing. CoreWeave reported $66.8 billion of revenue backlog at December 31, 2025, with weighted-average contract length increasing from four years to five years.[15][36]
Forecast 2027: Long-term capacity contracts should remain central in 2027 as hyperscalers and AI laboratories seek guaranteed access without building all infrastructure themselves. This will improve revenue visibility but may increase customer concentration and reduce providers’ flexibility during demand changes.
05Powered capacity becomes the main bottleneck
High confidenceCoreWeave operated 43 data centers with more than 850 MW of active power at the end of 2025. New York’s July 14, 2026 executive order described unprecedented data-center demand and imposed a one-year pause on new hyperscale data-center permits while energy and environmental standards were developed.[11][37]
Forecast 2027: In 2027, access to substations, transmission capacity and secured power contracts should determine expansion more often than GPU procurement alone. Providers with permitted sites and interconnection positions will likely command strategic premiums and deploy capacity faster than new entrants.
06Regulation shifts costs toward data centers
High confidenceNew York’s 2026 moratorium was explicitly linked to concerns about transmission and infrastructure costs, energy affordability, environmental impacts and local communities. A Congressional Research Service review noted that states were considering moratoria and policies requiring large data centers to bear the grid costs they create.[37][38]
Forecast 2027: By 2027, permitting, utility tariffs, financial guarantees and environmental reviews should become standard components of neocloud site economics. Projects with weak community benefits or uncertain grid-cost allocation are likely to face longer schedules and higher required returns.
07GPU prices remain segmented by service level
Medium confidenceA November 2026 market index placed the median publicly listed H100 SXM on-demand rate at $3.49 per GPU-hour across 35 providers, compared with a $9.90 median for AWS, Azure, Google Cloud and Oracle. The same index placed specialist GPU-cloud pricing at a median of $3.25 per hour.[39]
Forecast 2027: In 2027, basic GPU rental prices should drift lower as newer accelerators enter the market, while premium pricing persists for guaranteed capacity, high-bandwidth networking, managed software and scarce next-generation systems. The spread between commodity and full-service offerings should therefore remain wide.
08Next-generation accelerators reset pricing
High confidenceNVIDIA reported $107.6 billion of data-center revenue in its fiscal third quarter of 2027, up 110% year over year, while its Vera Rubin platform generated $20 billion in the quarter according to reporting on the company’s results. This indicates rapid migration toward newer accelerator generations alongside continuing demand for existing systems.[24]
Forecast 2027: During 2027, neoclouds should use a mixed fleet of older, heavily utilized GPUs and newer systems targeted at high-value training and inference. New hardware will support premium rates initially, while older generations will increasingly compete on price and availability.
09NVIDIA financing deepens supplier dependence
High confidenceNVIDIA disclosed $32 billion of neocloud agreements in the quarter ended October 25, 2026. Under the agreements, NVIDIA receives revenue from infrastructure sales and can participate in revenue generated by neoclouds from third-party customers when specified conditions are met.[11]
Forecast 2027: In 2027, equipment financing, supply commitments and revenue-sharing structures should remain important to neocloud expansion. They will accelerate deployment for well-capitalized providers but increase exposure to utilization shortfalls, refinancing risk and changes in NVIDIA’s credit or allocation policies.
10Capital intensity keeps utilization critical
High confidenceCoreWeave reported $2.0 billion, or 205%, growth in technology and infrastructure expense in 2025 and used approximately $10.3 billion in investing cash flow. The company said it expected to increase investment in technology and infrastructure beyond 2025.[11]
Forecast 2027: In 2027, providers should prioritize contracted utilization, workload scheduling and cluster reconfiguration to protect returns on expensive accelerators. Providers unable to maintain high utilization may respond with more spot capacity, shorter commitments or asset sales, increasing competitive pressure.
04 Statistics
Market Intelligence
Market Intelligence quantifies the 2026 neocloud GPU-cloud market at US$25.199 billion and tracks how adoption and operator lists expanded through 2026 updates.
Market size and revenue in the neocloud industry
- US$25.199 billion was the estimated 2026 global neocloud GPU-cloud market, covering GPU cloud services sold by specialized providers.[1]
- US$25 billion was the estimated 2025 global neocloud revenue, covering specialized AI-cloud providers, according to Synergy Research Group as cited in a 2026 Centuria Capital presentation.[2]
Growth forecasts for the neocloud industry
- US$267 billion is Gartner’s projected 2030 global AI-cloud market, the broader market from which neocloud providers are forecast to capture revenue; the category includes AI-focused cloud providers.[3]
- Nearly US$250 billion is ABI Research’s projected 2030 global revenue for GPU-as-a-service offerings from neocloud providers, up from US$42 billion in 2025.[4]
- 43% is ABI Research’s projected compound annual growth rate for global neocloud GPU-as-a-service revenue from 2025 to 2030.[4]
- 20% is Gartner’s projected neocloud-provider share of the US$267 billion global AI-cloud market in 2030.[3]
Product counts and provider structure in the neocloud industry
- More than 400 GPU-cloud operators were tracked worldwide in American Compute’s broader directory, which includes neoclouds alongside sovereign AI clouds, AI factories, former bitcoin miners and serverless GPU platforms.[5]
- 70 named GPU-cloud operators were listed in American Compute’s 2026 neocloud list, updated October 17, 2026.[5]
- 46 providers were classified as neoclouds in The Neocloud Guide’s 97-provider directory, updated October 14, 2026.[6]
Demand shifts in the neocloud industry
- Neocloud adoption among U.S. software-spending businesses on Ramp rose from 2.0% to 3.3% over the two years preceding September 24, 2026.[7]
- Modal’s neocloud adoption rate increased by an average 1.0 percentage point per month in November 2026, making it the fastest-growing vendor in Ramp’s category data.[8]
- Neocloud software usage among Ramp-tracked businesses reached 2.0% in November 2026, up 0.2 percentage points from October 2026.[8]
05 Statistics
Strategy & Growth
Strategy & Growth summarizes the deal stream, including four notable neocloud acquisitions identified for 2025 and one merger announced in February 2026.
M&A and consolidation in the neocloud industry
- 4 notable neocloud acquisitions were identified for 2025, including Weights & Biases, OpenPipe and Monolith AI by CoreWeave and Saagie by Scaleway.[9]
- 1 merger was announced in February 2026 between Radiant, a Brookfield-backed AI-infrastructure company, and Ori Industries, combining their AI-infrastructure platforms.[10]
- $1.4 billion was the aggregate consideration for CoreWeave’s acquisition of AI developer-platform provider Weights & Biases, which closed on May 5, 2025, in the neocloud industry.[11]
06 Statistics
Pricing & Packaging
Pricing & Packaging anchors published GPU-hour and inference rates, including $3.99 per GPU-hour for an NVIDIA H100 SXM instance on Lambda in November 2026.
Price levels in the neocloud industry
- $3.99 per GPU-hour was Lambda’s published price for an NVIDIA H100 SXM instance on its AI cloud pricing page, observed in November 2026; the price is for self-service GPU computing.[12]
- $3.85 per GPU-hour was Nebius’s published on-demand price for an NVIDIA HGX H100 instance in its AI Cloud pricing table, effective October 1, 2026; the service is neocloud GPU computing in the global market.[13]
- $6.16 per GPU-hour was CoreWeave’s published inference price for an NVIDIA HGX H100 GPU, versus $49.24 per hour for the eight-GPU system, on its North America pricing page in November 2026.[14]
07 Statistics
Finance
Finance adds margin benchmarks for public neocloud companies and records the scale of AI-neocloud startup funding activity through November 20, 2026.
Margins and unit economics in the neocloud industry
- 15.6% median adjusted gross margin was projected for public neocloud companies in 2026, with depreciation and amortization included.[15]
- 40.1% median EBITDA margin was projected for public neocloud companies in 2027, compared with 59.7% median EBITDA margin in 2027 for the same model after the stated company projections.[15]
- 62% EBITDA margin is shown for a single U.S. B200 GPU unit under a five-year annualized model for the dominant bare-metal neocloud model, before GPU and server depreciation.[16]
- 55%–65% gross margins are typical for global bare-metal-as-a-service neoclouds before depreciation, depending on utilization and pricing, according to McKinsey’s 2025 category analysis.[16]
Funding and investment in the neocloud industry
- 142 neocloud startup financings at 65 companies raised at least $37.2 billion globally from early 2024 through November 20, 2026, according to New Market Pitch’s tracker.[17]
- 187 AI-neocloud companies had recorded funding activity globally by October 31, 2024, according to PitchBook’s category dataset.[18]
- 779 funding deals involving AI neoclouds had been recorded globally by October 31, 2024, according to PitchBook.[18]
- Follow-on financings represented 108 deals, or 76% of recorded global neocloud funding events, and 90% of disclosed capital in New Market Pitch’s tracker through November 2026.[17]
08 Statistics
Sustainability
Sustainability provides a parent-category proxy: 82 terawatt-hours of electricity consumption for 403 of the largest U.S. data centers between May 2024 and April 2025.
Energy use, emissions and disclosure in the neocloud industry
- 82 terawatt-hours of electricity were consumed by 403 of the largest U.S. data centers between May 2024 and April 2025; this is a parent-category proxy for neocloud computing infrastructure.[19]
- 28.3 million tons of carbon dioxide were estimated from electricity use by 403 of the largest U.S. data centers between May 2024 and April 2025; this is a parent-category proxy for neocloud emissions.[19]
- 89% of 200 global digital companies reported Scope 1 emissions and 81% reported Scope 2 emissions for 2024; this parent-category disclosure rate provides context for neocloud providers.[20]
Key questions about the neocloud industry
What is the neocloud market size in 2026?
How fast is the neocloud market growing toward 2031?
Which companies lead the neocloud industry?
What technology shifts are reshaping demand?
How is the market structured among providers?
What constraints and regulation affect capacity planning?
Cite This Report
This report is designed to be cited. We maintain stable URLs and versioned verification dates. Copy the format appropriate for your publication below.
Samuel Norberg. (2026, November 20). Neocloud Industry Statistics. Gitnux. https://gitnux.org/neocloud-industry-statistics
Samuel Norberg. "Neocloud Industry Statistics." Gitnux, 20 Nov 2026, https://gitnux.org/neocloud-industry-statistics.
Samuel Norberg. 2026. "Neocloud Industry Statistics." Gitnux. https://gitnux.org/neocloud-industry-statistics.
Sources & references
39 sources cited across this report
- verishkainfo.com
company-announcements.afr.com
gartner.com
abiresearch.com
- amcompute.com
sonnetx.github.io
apollo.com
ramp.com
s.veneneo.workers.dev
- radiant.co
sec.gov
lambda.ai
nebius.com
wf.coreweave.com
williamblair.com
mckinsey.com
newmarketpitch.com
files.pitchbook.com
apnews.com
itu.int
srgresearch.com
- plugin-capital.com
axios.com
tomshardware.com
crusoe.ai
fluidstack.io
nscale.com
s205.q4cdn.com
blogs.vultr.com
iren.com
runpod.io
ir.applieddigital.com
gcore.com
civo.com
jll.com
d18rn0p25nwr6d.cloudfront.net
governor.ny.gov
everycrsreport.com
- gpueconomy.com