Nvidia has announced its latest data center GPU, the Blackwell Ultra.

The chip is built on the Blackwell architecture introduced last year, with Nvidia shifting to an annual GPU release cadence since the AI boom.

Blackwell Ultra comes in the Nvidia GB300 NVL72 rack-scale solution and the Nvidia HGX B300 NVL16 system.

The GB300 NVL72 connects 72 Blackwell Ultra GPUs and 36 Arm Neoverse-based Nvidia Grace CPUs in a single liquid-cooled rack, which the company claims offers 1.5x more AI performance than the Nvidia GB200 NVL72.

The Ultra combines two reticle-sized GPUs, has 15 petaflops of FP4 performance, and 288GB HBM3e.

Ultra products are expected to be delivered from the second half of 2025, although it should be noted that Blackwell GPUs were hit by delays and missed their launch.

Cisco, Dell Technologies, Hewlett Packard Enterprise, Lenovo, and Supermicro are expected to deliver a wide range of servers based on Blackwell Ultra. Servers are also expected from Aivres, ASRock Rack, ASUS, Eviden, Foxconn, Gigabyte, Inventec, Pegatron, Quanta Cloud Technology (QCT), Wistron, and Wiwynn.

Hyperscalers and neoclouds are also planning to offer Blackwell Ultra-powered instances, including Amazon Web Services, Google Cloud, Microsoft Azure, and Oracle Cloud Infrastructure, along with CoreWeave, Crusoe, Lambda, Nebius, Nscale, Yotta, and YTL.

Nvidia itself will offer access to the GB300 NVL72 on its DGX Cloud.