AI BriefingOrnn Explains Primary Compute Market Structures: Reserved, On-Demand, and Spot
·2026.11.06 00:59
Key point
The article uses a hypothetical 1,024 NVIDIA B300 cluster to illustrate how reserved, on-demand, and spot pricing models function in the primary compute market.
Details
- Primary compute markets involve three transaction structures: reserved (long-term commitments), on-demand (flexible, higher-cost access), and spot (interruptible, lowest-cost access).
- Reserved contracts, such as a five-year take-or-pay agreement for 640 GPUs at $4 per GPU-hour, provide the revenue certainty needed to secure financing, like a delayed draw term loan at SOFR + 5%.
- On-demand rentals allow customers with variable workloads to pay higher rates for flexibility, while spot rentals offer lower rates in exchange for the provider's right to reclaim capacity with short notice.
- The Ornn Compute Price Index (OCPI) measures only on-demand GPU rental prices in USD per GPU-hour using executed transactions, excluding reserved and spot deals, and underpins monthly cash-settled futures contracts on ICE Futures U.S.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Analysis: AGI Arrival Could Drive H100 Rental Prices to $30-$75 Per HourWhy Compute Costs Could Become 10x More Expensive in the Coming Years