AI News Feed
Market watch
LG UltraFine Evo 32U990A-S Hits Record-Low $1,099 on Amazon, TechRadar ReportsNewegg Labor Day Deal: 32GB Kingston Fury Beast DDR4-3200 Kit Nearly Half Price at $179.99 with CodeGalaxy Z Fold 8 inner screen corners reported with excessive giveHuawei unveils Mate XT 2 tri-fold with privacy-screen edition as Apple prepares folding iPhoneWall Street Analysts Recommend Nvidia, Uber and Marvell for Long-Term GrowthLenovo, AMD and Huawei Put Local Large-Model AI at Center of New Devices
LG UltraFine Evo 32U990A-S Hits Record-Low $1,099 on Amazon, TechRadar ReportsNewegg Labor Day Deal: 32GB Kingston Fury Beast DDR4-3200 Kit Nearly Half Price at $179.99 with CodeGalaxy Z Fold 8 inner screen corners reported with excessive giveHuawei unveils Mate XT 2 tri-fold with privacy-screen edition as Apple prepares folding iPhoneWall Street Analysts Recommend Nvidia, Uber and Marvell for Long-Term GrowthLenovo, AMD and Huawei Put Local Large-Model AI at Center of New Devices
AI Chips & Compute

CoreWeave Tops GPU Neocloud Ranking, Lambda Offers Cheapest B200: MarkTechPost

A MarkTechPost ranking of five GPU neoclouds finds CoreWeave the only Platinum-rated provider, Lambda with the lowest B200 price, Nebius the sole B300 list, and Groq pivoting to inference after licensing LPU to Nvidia.

AI News Feed · 2026-08-22 10:20

The five companies have very different business models, the report said. CoreWeave and Nebius are public companies that publish quarterly results; Lambda and Crusoe are private, with Lambda reportedly heading toward an IPO in the second half of 2026. Groq, which licensed its LPU technology to Nvidia in December 2025, has refocused entirely on inference-cloud infrastructure. In published pricing, Nebius lists the cheapest H100 on-demand rate at $3.85 per GPU-hour, Lambda has the lowest B200 at $6.69, and Crusoe the lowest H200 at $4.29.

CoreWeave commands a premium of 10-15% per GPU-hour over Nebius, Crusoe, and Lambda on managed clusters, according to the report. The company reported second-quarter revenue of $2.575 billion, up 112% year over year, with a net loss of $626 million. It has 1.5 GW of active power and over 4.2 GW contracted across 51 data centers. Anchor deals with Meta, Anthropic, and OpenAI contribute to a disclosed backlog of approximately $104 billion. CoreWeave is also the first provider to bring up Nvidia's Vera Rubin NVL72 platform, and it plans $35-39 billion in capital expenditures for 2026.

Nebius, rated at the top of the Gold tier, reported Q2 revenue of $582 million, up 454% year over year, and AI cloud annual recurring revenue of $3.0 billion. It has announced anchor deals including Microsoft at $17.4 billion and Meta at $3 billion with the potential to reach about $27 billion. Nebius recently priced $5.0 billion in convertible notes on August 19. The company has received Vera Rubin NVL72 units and is currently validating them. It offers committed-use discounts of up to 35% and free network egress.

Lambda, which is rated Silver, offers self-serve on-demand instances with no spot tier. Its posted B200 rate of $6.69 per GPU-hour is the lowest among the five for that chip. The company has raised $1.5 billion in a Series E, secured a $1 billion credit facility, and completed a $350 million pre-IPO round, according to the report. Crusoe, rated Gold, is the only provider with AMD MI300X/MI355X on its rate card and lists the lowest H200 price at $4.29. Crusoe has 4.9 GW of contracted power and a pipeline of over 40 GW, with its Abilene site energized. It is involved in a 1.2 GW project with Oracle and OpenAI and a 900 MW campus for Microsoft.

Groq, which has operated GroqCloud on its LPU architecture since 2024 and counts more than 6 million developers on the platform, was certified as an Nvidia Cloud Partner on August 12 and plans to add Nvidia GPU capacity alongside its LPU inference cloud. It operates 13 data centers and expects to grow from 54 MW to more than 200 MW by 2027. The company raised $650 million in June and a $350 million Series A at a $3.5 billion valuation on August 17, with Nvidia participating. Groq was not rated in the ClusterMAX 2.0 assessment, which evaluated inference API at the time.

Contract terms also differ. CoreWeave offers reserved instances with discounts up to 60% and free egress; Nebius discounts up to 35% for committed use; Lambda and Crusoe use sales-quoted pricing; Groq uses per-token pricing for its cloud. CoreWeave's premium pricing is attributed to its Platinum rating in SemiAnalysis ClusterMAX 2.0, which it achieved twice. The report says that on Q3 short-term deals at Nebius, capacity clears at over $40 million per megawatt, compared with roughly $12 million on its 2026 base.