AI demand is hot and verified again! CoreWeave (CRWV.US) and Nebius (NBIS.US) computing power services both increased prices

The Zhitong Finance App learned that Nvidia-supported AI cloud infrastructure provider CoreWeave (CRWV.US) said that since June 30, 2026, the company has continued to sign new computing power capacity contracts at higher prices, including short-term agreements signed in the third quarter. According to reports, in the third quarter of 2026, CoreWeave has signed a contract with the customer for a period of about 3 to 6 months, and the price is about 40 million US dollars per megawatt. The price is calculated based on the ratio of annualized revenue to the electricity required to operate the relevant cluster. CoreWeave also reiterated that the company added more than $25 billion in net customer commitments at the beginning of the third quarter, and these commitments did not count towards the revenue backlog as of June 30. As of August 11, the total power capacity contracted by the company increased to about 4.2 gigawatts, compared to about 3.7 gigawatts on June 30.

It is worth mentioning that the news that CoreWeave signed a new computing power capacity contract at a higher price in the third quarter echoes the news that its new cloud partner NeBius (NBIS.US) raised the price of computing power services, and together confirms the market judgment that demand for AI computing power is still strong.

However, as of press release, CoreWeave's US stocks fell 0.33% on Thursday, underperforming Nebius and iREN (IREN.US) — up more than 9% and 5%, respectively. This is probably related to the news that the company plans to sell up to 35 million Class A common shares.

According to a notice issued by Nebius to customers, Nebius has raised the prices of a number of on-demand computing resources since October 1. Among them, the latest Nvidia GPU increased the most. Under the new price, the H100 rose from $3.85 per GPU hour to $4.50 per GPU hour, an increase of about 17%; the H200 rose from $4.50 to $5.40, an increase of 20%; the B200 rose from $7.15 to $8.50, an increase of nearly 19%; the Nvidia B300 definitely saw the biggest increase, from $7.85 to $9.50 per GPU hour, an increase of about 21%.

This price increase also covers the older Hopper architecture GPUs and the newer Blackwell series, indicating that market demand is not limited to a certain generation of chips, but is widely distributed over various types of AI computing power. This is Nebius' second round of price increases in months. In May of this year, the company announced an average price increase of 29% for on-demand capacity and a 51% price increase for preemptible capacity. If you take the B300 as an example, the price is calculated from about 6.10 US dollars/hour before 5 months, and the cumulative increase has reached about 56% so far.

According to reports, Nebius has tried manual price adjustments before, but management said that due to strong demand, price adjustments are still insufficient to balance supply and demand. The company then conducted an auction test on the scarce Blackwell computing power, and the actual price paid by the customer was 15% to 20% higher than the previous maximum price to ensure access. Daniel Newman, CEO of Futurum Group, posted on X: “Demand for AI computing power is still exploding. Nebius's 20% price increase is just another example of strong demand.”

Furthermore, the news that CoreWeave signed a new computing power capacity contract at a higher price in the third quarter may also boost investors' optimistic expectations about the company's third-quarter results. The company's second-quarter revenue and profit both exceeded Wall Street expectations, and the backlog of orders broke through the $100 billion mark — up 246% year over year to about US$104 billion, and continued to rise from US$99.4 billion at the end of the previous quarter. Based on strong demand momentum, CoreWeave also raised its full-year outlook. As one of the few new cloud computing vendors listed on the market, CoreWeave's performance is seen as an important weather vane for overall AI computing power demand, and its performance exceeding expectations confirms that the AI infrastructure investment boom continues to heat up.