Nvidia has announced its next-generation AI 'superchips', the Blackwell Ultra GB300 and Vera Rubin, which are expected to further solidify the company's dominance in the AI computing market. The Blackwell Ultra GB300, set to ship in the second half of this year, and the Vera Rubin, slated for release in the second half of next year, promise significant performance boosts and expanded capabilities in AI computing.

The Blackwell Ultra GB300 is an enhanced version of the original Blackwell chip, with a single Ultra chip offering the same 20 petaflops of AI performance as its predecessor, but with 288GB of HBM3e memory, up from 192GB. A Blackwell Ultra DGX GB300 "Superpod" cluster will offer the same 288 CPUs, 576 GPUs, and 11.5 exaflops of FP4 computing as the Blackwell version, but with 300TB of memory, up from 240TB.

Nvidia is positioning the Blackwell Ultra GB300 as a significant upgrade over its 2022 chip, the H100, which originally built Nvidia's AI fortunes. The company claims that the Blackwell Ultra GB300 offers 1.5x the FP4 inference and can dramatically speed up "AI reasoning," with the NVL72 cluster capable of running an interactive copy of DeepSeek-R1 671B that can provide answers in just ten seconds, compared to the H100's 1.5 minutes.

In a surprising move, Nvidia will also offer a single Blackwell Ultra chip for sale, with the DGX Station desktop computer featuring a single GB300 Blackwell Ultra on board, 784GB of unified system memory, built-in 800Gbps Nvidia networking, and the promised 20 petaflops of AI performance. Asus, Dell, and HP will join Boxx, Lambda, and Supermicro in selling versions of the desktop.

Nvidia will also offer a single rack called the GB300 NVL72, which offers 1.1 exaflops of FP4, 20TB of HBM memory, 40TB of "fast memory," 130TB/sec of NVLink bandwidth, and 14.4 TB/sec networking.

The Vera Rubin and Rubin Ultra, set to arrive in 2026 and 2027, respectively, promise even more significant performance boosts. The Rubin chip will feature 50 petaflops of FP4, up from 20 petaflops in Blackwell, while the Rubin Ultra will feature a chip that effectively contains two Rubin GPUs connected together, with twice the performance at 100 petaflops of FP4, and nearly quadruple the memory at 1TB.

A full NVL576 rack of Rubin Ultra claims to offer 15 exaflops of FP4 inference and 5 exaflops of FP8 training, for what Nvidia says is 14x the performance of the Blackwell Ultra rack it's shipping this year.

Nvidia's announcement comes as the company continues to ride the wave of the AI revolution, with its data center business generating massive profits. The company has already shipped $11 billion worth of Blackwell revenue, with the top four buyers alone purchasing 1.8 million Blackwell chips so far in 2025.

Nvidia's CEO, Jensen Huang, emphasized the need for increased computing power to keep up with demand, stating that the industry needs "100 times more than we thought we needed this time last year." The company is positioning its new chips as essential to the future of computing, and is trying to argue that companies will need more and more computing power, not less, as AI technology continues to advance.

The announcement is a significant development in the AI computing market, and is likely to have major implications for the industry as a whole. With its latest 'superchips', Nvidia is solidifying its position as a leader in AI computing, and is poised to continue driving innovation and growth in the field.