Rental rates for Nvidia B200 GPUs have risen significantly over the past three months, driven by demand that continues to outpace supply. According to data from a GPU rental marketplace, costs for these chips increased by 74% during this period.
Research firm SemiAnalysis reported a 79% surge in on-demand rental rates for the B200, reaching $8.01 per GPU-hour by late September 2026. The Ornn Compute Price Index tracked a 48% increase between late July and late October 2026, with prices rising from $6.58 to $9.76 per GPU-hour.
The market for B200 rentals remains highly fragmented. Data from early November 2026 shows that rates across 10 to 20 cloud providers range from $1.89 to $16.11 per GPU-hour. Major providers like AWS and Google Cloud charge over $14 to $16 per GPU-hour, while typical median rates fall between $6 and $8.50.
Spot pricing offers a lower-cost alternative, typically ranging from $3 to $5 per GPU-hour, though it involves the risk of interrupted service. This volatility impacts startups and mid-sized firms that rely on cloud infrastructure rather than purchasing hardware directly.
Supply constraints for the Blackwell-series chips are expected to persist through 2028. Rising component costs, including memory, continue to exert upward pressure on infrastructure pricing as the industry anticipates sustained demand through 2027.
