CoreWeave sold 37.5 million shares in its $1.5 billion offering. CoreWeave shares closed 42.1% higher at $52.69 after a lackluster debut, rising well above the $40 per share IPO price. I find the company’s $104 billion revenue backlog the most compelling aspect of the offering. This backlog excludes more than $25 billion of new commitments added in early Q3. CoreWeave’s revenue jump to $2.575 billion in the second quarter of 2026 represents a 112% increase year-over-year, yet the company still recorded a net loss of $626 million for that same period. This growth follows a 700% revenue jump in 2024 to $1.9 billion. The company posted a net loss of $863 million for 2024. Microsoft accounts for 62% of the 2024 revenue. Nvidia anchored the IPO by purchasing $250 million in stock. You can see the scale in the 4.2 GW of contracted power across 51 data centers. The company maintains 32 data centers with over 250,000 Nvidia GPUs and expects more than 1.85 GW of active power by year end.
Lambda and Crusoe offer different paths to compete with the CoreWeave scale. Lambda offers the market’s lowest B200 on-demand rate at $6.69 per GPU-hour. The company closed a $926 million senior secured term loan B facility in August 2026 to fund GPU deployment. This facility carries a price of SOFR + 3.00% and matures on December 31, 2030. Moody’s assigned the facility a Baa2 rating. I find Lambda’s $1 billion credit facility from May 2026 a solid indicator of its scaling capacity. Michel Combes leads Lambda as CEO. Lambda builds supercomputers for AI training and inference for tens of thousands of customers. Crusoe faces harder headwinds after Bloomberg reported the company lost a Wyoming AI campus in June. The CEO also addressed demand concerns to CNBC as market conditions shifted. Crusoe’s secondary market valuation reached $23.62 billion in June, which sits 136% above its October 2025 Series E valuation. Crusoe reported $276 million in revenue for 2024 and expects revenue to surpass $2 billion in 2026. Crusoe builds a 1.2 GW AI data center campus in Abilene, Texas, using $11.6 billion in debt and equity through a joint venture.
The AI infrastructure race now centers on power control rather than chip availability alone. CoreWeave manages 4.2 GW of contracted power across 51 data centers. Crusoe targets 40 GW in its development pipeline using a strategy of capturing flared gas at oilfields. Crusoe builds a 1.2 GW AI data center campus in Abilene, Texas, for the Stargate project.
| Service | CoreWeave | Lambda | Crusoe |
|---|---|---|---|
| H100 On-demand | $6.16 /GPU/hr | $3.99 /GPU/hr | $3.90 /GPU/hr |
| H200 On-demand | $6.30 /GPU/hr | Not listed | $4.29 /GPU/hr |
| A100 On-demand | $2.70 /GPU/hr | Not listed | $2.00 /GPU/hr |
| B200 On-demand | $8.60 /GPU/hr | $6.69 /GPU/hr | On request |
CoreWeave holds the only Platinum-rated GPU cloud status in SemiAnalysis ClusterMAX 2.0. CoreWeave’s H100 rate of $6.16 per GPU-hour sits about 60% above Nebius and Lambda. I would suggest that CoreWeave maintains a premium position by offering the industry’s first Vera Rubin NVL72 bring-up in Q2 2026. Is the high cost of interest expense a permanent fixture of this capital-intensive model?
