Nvidia B200 GPU Rental Rates Climb 74% in Three Months
On-demand prices for the Blackwell-series chip rose sharply as AI demand outpaces available cloud capacity.
Updated on Nov. 9, 2026 in Data Centers

Live Poll
Do you expect the cost of everyday tech and online services to rise in the next year?
On-demand rental rates for Nvidia's B200 GPU rose 74% over three months, according to the extraction, with SemiAnalysis recording a 79% increase. The B200 on-demand price reached $8.01 per GPU-hour by the end of September 2026.
Why it matters
Rising rental costs for Blackwell-series hardware affect the price of accessing AI computing capacity for companies and developers that depend on cloud infrastructure. Supply has remained constrained as demand for these chips has run ahead of the buildout of cloud hardware.
The B200 carries 180 GB of HBM3e memory and belongs to Nvidia's Blackwell series. Across providers in early November 2026, on-demand rates ranged from $1.89 to $16.11 per GPU-hour, with typical medians between $6 and $8.50.
The players
Nvidia
Nvidia is the American semiconductor company that designs the Blackwell-series GPUs, including the B200, used widely for AI training and inference workloads.
SemiAnalysis
SemiAnalysis is an independent semiconductor and AI hardware research firm known for tracking GPU pricing and supply trends.
Google Cloud
Google Cloud is the cloud computing division of Alphabet Inc., offering GPU-accelerated instances to enterprise and developer customers.
AWS
Amazon Web Services is the cloud computing arm of Amazon, and the largest provider of cloud infrastructure worldwide.
The details
Google Cloud and AWS listed B200 rates above $14 to $16 per GPU-hour, while spot rates often fell between $3 and $5 where available. Spot capacity can be reclaimed or interrupted, which accounts for its lower price, and providers need time to acquire and install new hardware.
Timeline
Late July 2026: Ornn index window begins with the B200 rental price at $6.58 per GPU-hour.
End of September 2026: The B200 on-demand rate reaches $8.01 per GPU-hour.
Late October 2026: Ornn index window ends with the B200 rental price at $9.76 per GPU-hour.
Early November 2026: Live provider listings show rates ranging from $1.89 to $16.11 per GPU-hour.
2028: Nvidia's forecasting points to limited B200 availability through this year.
The Tech Race
Blackwell-series chips are central to the current race among cloud providers to supply AI computing capacity, and the B200's rising rental prices show how supply constraints are shaping competition. Nvidia's forecast of limited availability through 2028 suggests the Blackwell generation will remain a bottleneck for providers competing on AI infrastructure.
Developers and businesses renting cloud GPUs for AI work face higher hourly costs, and the spread between on-demand and spot pricing means flexibility on interruptible workloads can reduce spending. Users should compare provider rates, since listings in early November 2026 ranged from $1.89 to $16.11 per GPU-hour.
The takeaway
Compare on-demand and spot rates across providers before committing to Blackwell-class capacity, since prices in the extraction ranged from under $2 to above $16 per GPU-hour. Monitoring indexes like Ornn's can help time workloads as rental prices continue to shift.
Further reading
For broader context on the infrastructure behind these prices, explore our Data Centers coverage.
Live Poll
Do you expect the cost of everyday tech and online services to rise in the next year?



