Nvidia B200 GPU Rental Prices Drop Over 30%

According to prediction market platform Kalshi, traders are betting that the compute rental price of Nvidia’s flagship B200 GPU is unlikely to recover to its late-May peak before the end of June, reflecting a short-term cooling in AI infrastructure pricing expectations.

Recent market performance also shows diverging sentiment around Nvidia. Over the past month, Nvidia shares have declined by roughly 3%, while year-to-date gains stand at about 14%. In contrast, the VanEck Semiconductor ETF has surged around 15% over the past month and is up approximately 84% year-to-date, highlighting stronger momentum across the broader semiconductor sector.

Pricing data from real-time GPU compute marketplace Ornn indicates that B200 rental costs peaked at $6.11 per hour on May 30, marking a three-month high. However, prices have since declined steadily, falling to around $4.22 per hour by June 21.

Within Kalshi’s prediction contracts, traders currently remain skeptical that B200 compute pricing will exceed its May highs again in the near term. These contracts are tied to second-quarter GPU pricing benchmarks, and outcomes depend on whether Ornn-reported prices surpass defined thresholds before June 30.

The majority of AI compute demand today flows through cloud service providers and emerging GPU cloud startups. As demand for AI infrastructure fluctuates, GPU rental prices can experience significant volatility driven by shifting supply-demand dynamics.

Seoyoung Kim, a finance professor at Santa Clara University, noted that uncertainty remains high across the ecosystem: many enterprises are unsure how much compute they will need over the next year, cloud providers are uncertain about GPU procurement levels, and manufacturers—including Nvidia—face challenges in forecasting production requirements.

Despite short-term price softness, some market signals remain constructive. Earlier this month, Google signed a large-scale compute leasing agreement with SpaceX, committing to approximately $9.2 billion per month from October 2026 through June 2029. The arrangement is expected to utilize around 110,000 Nvidia GPUs, alongside CPUs and storage infrastructure.

Analysts at RBC Capital Markets maintain a positive outlook on Nvidia’s longer-term positioning, stating that the company is “well positioned relative to peers” through 2026 and 2027. They also suggest that large-scale GPU leasing agreements may help ease investor concerns about potential competition from custom AI accelerators (ASICs), at least in the near term.