Nvidia B300 GPU-hour Index Hits All-Time High as Neoclouds Pivot to Inference
rickasaurus · x · 2026-08-07
The Nvidia B300 GPU-hour index has reached an all-time high. As profit margins on model training grow thinner, emerging cloud providers (neoclouds) are collectively shifting their focus from training to inference.
Despite the high cost of B300s, providers are generating strong ROI by lowering the marginal cost per token generated through batch inference. Meanwhile, industry observers are noting the lack of a B300 futures market to hedge against these surging compute costs.
More from Venture
- Stop Selling AI Tools: How to Land 7-Figure Enterprise Deals by Selling Competitive Edge — aryanXmahajan · 2026-08-07
- Stripe in exclusive talks to buy OpenRouter for ~$10B — steph_palazzolo · 2026-08-07
- Sequoia Calculates $2.5B Max Tolerable Loss for Anthropic Bet — nmasc_ · 2026-08-07
- SpaceFlow's AI Employees Manage $400M in Supply Chain Spend — ycombinator · 2026-08-07
- Path Robotics Lands $600M Deal to Deploy Physical AI in Shipbuilding — Rewkang · 2026-08-07
- YC S26 Startup Understudy Cuts Anthropic Bills by 80% via Model Distillation — ycombinator · 2026-08-07