In a seismic shift for artificial intelligence, Dell Technologies has delivered the world’s first AI supercomputer powered by Nvidia’s revolutionary GB300 Blackwell Ultra GPUs to cloud provider CoreWeave, marking a new era in AI infrastructure.
Unveiled today, this liquid-cooled behemoth, built on Dell’s Integrated Racks, packs 72 Blackwell Ultra GPUs, 36 Arm-based 72-core Grace CPUs, and 36 BlueField DPUs, setting a blistering pace for large language model (LLM) training, inference, and advanced reasoning. CoreWeave hailed the system as a “game-changer” for AI cloud performance, cementing its position at the forefront of the AI revolution.
AI Supercomputers That Shatter Limits
At the heart of this technological marvel are Nvidia’s GB300 superchips, engineered to dominate AI workloads. Industry analysts report these chips accelerate training and inference tasks by orders of magnitude compared to legacy architectures, slashing processing times from weeks to hours.
CoreWeave’s GB300-NVL72 racks, an upgrade from their existing Grace Blackwell GPUs, now rank among the fastest in the AI cloud sector, promising unprecedented speed for data-hungry AI models.
A High-Stakes AI Supercomputer Race
This breakthrough widens the gap in the global AI race. With Dell and CoreWeave deploying next-generation systems in rapid succession, competitors are left scrambling to match this computational firepower.
As global data is projected to soar to 175 zettabytes by 2025, the demand for such high-performance infrastructure is skyrocketing, positioning Dell and CoreWeave as frontrunners in meeting the world’s AI needs.
Navigating U.S. Export Controls
The timing of this rollout is no coincidence. With U.S. export controls tightening on advanced AI technology to China, companies are doubling down on domestic infrastructure.
Dell’s strategic deployment of GB300 systems in the U.S. signals a bold move toward self-reliant compute power, reducing reliance on global chip supplies and strengthening national AI capabilities in a geopolitically charged landscape.
Government Supercomputers Join the Fray
The private sector isn’t alone in this AI surge. The U.S. Department of Energy is powering up Doudna, a cutting-edge AI-public supercomputer at Berkeley Lab, built on Dell servers and Nvidia’s “Vera Rubin” chips. Designed for scientific breakthroughs, Doudna delivers tenfold the computational power of current systems, unlocking new possibilities in climate modeling, medical research, and beyond.