The neocloud market has evolved from a collection of GPU rental startups into a strategic infrastructure sector, now dominated by five distinct companies. CoreWeave, Nebius, Lambda, Crusoe, and Groq each pursue different business models, hardware strategies, and customer segments, making direct comparison essential for any serious buyer. This analysis cuts through the marketing claims to examine what actually matters: published pricing, deployed and contracted power, hardware roadmaps, contract structures, and independent quality ratings. The result is a clear ranking of where each provider excels and where they fall short.
The Five Contenders Defining the GPU Neocloud Market in 2026
CoreWeave (CRWV): The Platinum Standard
CoreWeave reported second quarter 2026 revenue of $2.575 billion, up 112 percent year over year, with a revenue backlog of roughly $104 billion. Active power reached 1.5 GW, and contracted power stands at 4.2 GW across 51 data centers. The company completed the industry’s first bring-up and validation of NVIDIA Vera Rubin NVL72 during the quarter. Net loss was $626 million, driven by $640 million in net interest expense, reflecting the capital-intensive nature of its build-out. Full-year 2026 guidance calls for $12.4 to $13.2 billion in revenue and $35 to $39 billion in capital expenditures. CoreWeave is the only provider to earn a Platinum rating in SemiAnalysis ClusterMAX 2.0, a distinction it has now received twice.
Nebius (NBIS): The Well-Capitalized Challenger
Nebius reported second quarter 2026 group revenue of $582.3 million, up 454 percent year over year. AI cloud revenue alone was $574.9 million, with an annualized run rate of $3.0 billion. The AI cloud segment posted a 49.7 percent adjusted EBITDA margin. Nebius raised its year-end 2026 contracted power target to 5 GW and plans to deploy more than 1 GW per year starting in 2027. It holds $8.0 billion in cash and has a Gold rating in ClusterMAX 2.0. Nebius is the only provider in this group publishing on-demand pricing for the B300 GPU.
Lambda: The Private Powerhouse Targeting an IPO
Lambda remains private, having closed a Series E of more than $1.5 billion in November 2025 and a $1 billion senior secured credit facility in May 2026. Michel Combes, former CEO of SoftBank International and Sprint, became CEO in May 2026. Lambda holds the lowest published B200 on-demand rate at $6.69 per GPU-hour. The company has a multibillion-dollar, multi-year agreement with Microsoft covering tens of thousands of NVIDIA GPUs, including GB300 NVL72 systems. The Information has reported an IPO target of the second half of 2026.
Crusoe: The Energy-Integrated Infrastructure Developer
Crusoe raised a $1.375 billion Series E at a valuation above $10 billion in October 2025, and by July 2026 was in talks to raise roughly $3 billion at a valuation near $30 billion. The company is building the 1.2 GW Abilene, Texas campus for Oracle and OpenAI, and a separate 900 MW Abilene campus for Microsoft. Crusoe is the only provider of the five with AMD Instinct MI300X and MI355X on its published rate card. As of June 2026, contracted AI infrastructure reached 4.9 GW across five US campuses, with a development pipeline above 40 GW.
Groq: The Inference Specialist Going Hybrid
Groq launched GroqCloud in 2024 around its LPU inference chip, but the company transformed in December 2025 when NVIDIA signed a non-exclusive license for Groq’s inference technology in a transaction reported at about $20 billion. Groq stayed independent and raised $650 million in June 2026 and a $350 million Series A in August 2026 at a $3.5 billion valuation. The company now operates 13 data centers and expects to scale from 54 MW to 200+ MW in 2027. On August 12, 2026, Groq joined the NVIDIA Cloud Partner program and said it plans to bring NVIDIA accelerated computing online, though that GPU capacity is not yet a published service.
Published Pricing: Who Offers the Lowest Rates and Why
What is the cheapest GPU neocloud for H100s?
Nebius offers the lowest published H100 on-demand rate at $3.85 per GPU-hour, followed closely by Crusoe at $3.90 and Lambda at $3.99. CoreWeave charges $6.16 per GPU-hour for the same hardware, a premium of roughly 60 percent. For budget-sensitive experiments, Nebius preemptible H100 instances at $2.15 and CoreWeave spot H100 at $2.46 per GPU-hour represent the lowest published rates across all providers.
Blackwell and Beyond: B200 and B300 Pricing
Lambda holds the cheapest published B200 instance at $6.69 per GPU-hour. Nebius offers B200 at $7.15 and is the only provider publishing a B300 on-demand rate at $7.85. CoreWeave lists B200 at $8.60 and GB200 NVL72 at an effective $10.50 per GPU. CoreWeave charges a premium across the board, but SemiAnalysis reports that the company commands roughly 10 to 15 percent more per GPU-hour than Nebius, Crusoe, and Lambda on managed Slurm or Kubernetes clusters. The premium reflects the only Platinum ClusterMAX rating, indicating superior reliability, networking, and operational practices.
The AMD Alternative: Crusoe’s Unique Position
Crusoe is the only provider among the five offering AMD Instinct MI300X on its on-demand card at $3.45 per GPU-hour, with MI355X available via sales. This makes Crusoe the default choice for teams evaluating AMD hardware or running workloads optimized for ROCm. CoreWeave, Nebius, and Lambda remain NVIDIA-only on their public pages.
Deployed and Contracted Power: Measuring True Scale
Power capacity is the most direct proxy for a neocloud’s ability to deliver compute at scale. CoreWeave leads in active power at 1.5 GW, with expectations to exceed 1.85 GW by year end 2026. Nebius targets up to 1 GW connected by year end and 5 GW contracted. Crusoe has 4.9 GW contracted, though most of that is dedicated to specific anchor tenants. Lambda does not disclose active power but operates 15 US data centers. Groq is the smallest at 54 MW, scaling to 200 MW in 2027. For buyers planning frontier-scale training runs requiring 10,000 or more GPUs for extended periods, CoreWeave and Nebius are the only two providers with demonstrated rack-scale operating experience and the financial infrastructure to support multi-year commitments.
Hardware Roadmap: Securing Access to Vera Rubin and Next-Gen Silicon
Access to next-generation hardware is a critical differentiator. CoreWeave completed the industry’s first bring-up and system-level validation of NVIDIA Vera Rubin NVL72 in the second quarter of 2026. Nebius has received its first Vera Rubin NVL72 systems and is validating compute, networking, and orchestration ahead of production availability. Lambda was a GTC 2026 launch partner for NVIDIA’s Vera CPU and Quantum-X800 InfiniBand co-packaged optics. Crusoe announced launch support for the Vera CPU and NVIDIA Nemotron 3 models. Nebius is the only provider with a public B300 on-demand price, while CoreWeave lists GB300 NVL72 and HGX B300 as contact-sales, with a B300 spot rate available. For AMD, Crusoe remains the sole option with MI300X and MI355X.
Anchor Deals and Financial Health: The Ultimate Test of Staying Power
Contract structure and anchor customers reveal a provider’s financial stability and long-term viability. CoreWeave’s $104 billion backlog from Meta, Anthropic, and OpenAI provides extraordinary revenue visibility. Nebius has secured $17.4 billion from Microsoft and a complex $3 billion agreement with Meta that includes a floor on utilization for covered clusters. Lambda holds a multibillion-dollar Microsoft deal and a reported $1.5 billion lease-back arrangement with NVIDIA. Crusoe’s long-term leases with Oracle, OpenAI, and Microsoft are backed by massive project financing, including a $7.1 billion construction loan for the Abilene joint venture. Groq has raised $1 billion in 2026 and is backed by NVIDIA through its technology licensing and cloud partnership. For buyers needing a public balance sheet and audited financial statements, only CoreWeave and Nebius file with the SEC.
Matching Workloads to Providers: A Practical Guide for Buyers
Frontier-scale training requiring 10,000 or more GPUs over multi-year periods belongs to CoreWeave and Nebius. Both have demonstrated NVL72 rack-scale operating experience, Vera Rubin units in hand, and SEC-reported financial scale. Mid-size training and fine-tuning on 16 to 512 GPUs for weeks to months is best served by Lambda’s 1-Click Clusters or Nebius’s on-demand B200 and B300 instances. Crusoe’s H200 at $4.29 is the cheapest Hopper-class option for memory-heavy jobs. Budget-sensitive experiments should target Nebius preemptible H100 instances at $2.15 or CoreWeave spot H100 at $2.46 per GPU-hour. Lambda has no spot tier. AMD GPU evaluation is possible only at Crusoe. Low-latency inference at high token volume is GroqCloud’s distinctive offering, though Nebius Token Factory and Crusoe Managed Inference are GPU-based alternatives with published per-token rates.
The neocloud market is no longer a collection of undifferentiated GPU rental services. CoreWeave has established itself as the premium, reliable choice for the world’s largest AI labs, commanding higher prices because it delivers Platinum-grade reliability. Nebius is the aggressive, well-capitalized challenger offering competitive pricing, strong financials, and the only published B300 access. Lambda, Crusoe, and Groq are carving out specific niches based on price, AMD hardware, and inference architecture. For buyers, the decision comes down to a trade-off between price, reliability, scale, and hardware roadmap. The providers that survive and thrive will be those that can maintain operational excellence while securing the capital and power needed to meet the insatiable demand for AI compute.
