According to a report from TrendForce, Nvidia may have to delay the volume increase of its Blackwell machines.

Due to overheating, power consumption and necessity, Nvidia may have to delay the volume increase of next-generation AI servers based on the B200 and GB200 platforms interconnection to optimize. “TrendForce therefore predicts that mass production and peak deliveries are unlikely to occur between the second and third quarters of 2025,” the TrendForce report said. Nvidia has not yet responded to the claims made in this report.

Increasing the delay volume

Blackwell has proven to be a difficult task for Nvidia for some time. The Blackwell B200 accelerators have already been announced as the most powerful of their kind that will train large AI models. Refined versions of the Nvidia B200 processors recently entered mass production, although this doesn’t mean the rise of Blackwell-based servers will increase immediately, according to TrendForce.

According to the report, mass production and peak deliveries of Blackwell machines will occur sometime in mid-2025, half a year later than planned. According to the market research institute, this has to do with overheating, power consumption and necessity interconnection to optimize. A few months ago it turned out that an Nvidia NVL 72 rack based on the GB200 platform with 72 B200 GPUs would consume around 120 kW of power. Nvidia has not yet responded to the report’s findings.