Jensen Huang, CEO of NVIDIA, speaks during a press conference after arriving at Gimpo International Airport in Seoul, South Korea, on June 5, 2026.
Chris Jung | Nurphoto | Getty Images
Shares of Nvidia have been faltering recently — and Kalshi traders predict that what the company can charge for chips is also declining.
While the artificial intelligence chipmaker is up about 12% in 2026, shares have been skidding in the past month, losing roughly 3%. In comparison, the VanEck Semiconductor ETF (SMH) is up 84% this year, and the fund has advanced 15% in the past month.
Nvidia has sat on the sidelines as Wall Street has focused on memory chips and infrastructure in the next steps of the AI buildout. This has been driving gains for companies like Micron Technology and Sandisk, both up nearly 60% in the past month alone.
As shares of Nvidia have been languishing, the price of computing power for the company’s B200 chip has also dropped. Nvidia’s B200 is the company’s flagship graphics processing unit or GPU that helps run data centers at a massive scale.
B200’s compute per hour price climbed to $6.11 on May 30, the highest within the past three months, according to Ornn, which provides live GPU compute prices across major hardware types. Since that day, the dashboard shows B200 compute per hour price declining, sitting at $4.22 as of June 21.
Kalshi traders are now pessimistic that Nvidia’s AI chip’s compute price will surpass May’s high. The contract asks the price of Nvidia B200 compute for the second quarter and will be resolved if the value, appearing on Ornn, is above a certain threshold by June 30.
Most companies rent access for GPUs through cloud providers or neoclouds, another growing ecosystem. However, the cost to rent GPUs can fluctuate as demand for AI infrastructure grows.
“A lot of people don’t know how much computing power they’ll need in the next year, and a lot of suppliers of that computing power right now don’t know how many GPUs and to what capacity they should order,” Seoyoung Kim, a finance professor at Santa Clara University, previously told CNBC. “And the manufacturers, like Nvidia, they don’t know how much they should produce.”
Earlier this month, Google agreed to pay SpaceX $920 million a month to rent AI computing capacity from October 2026 through June 2029. In doing so, Google will use roughly 110,000 Nvidia GPUs, CPUs, memory and other related components. After the deal, RBC Capital Markets was bullish on Nvidia’s performance for the second half of 2026 and 2027, saying the chip giant “looks best positioned among peers.”
“Regardless of the precise rationale, these GPU rental agreements should put to rest lingering concerns about NVDA losing share to [application-specific integrated circuits], at least in the short term,” the analysts wrote.
— CNBC’s Yun Li contributed to this story.
Disclosure: CNBC and Kalshi have a commercial relationship that includes customer acquisition and a minority investment.
