Dell Technologies has unveiled a new generation of AI servers powered by Nvidia’s Blackwell Ultra chips, targeting the growing demand for enterprise AI adoption. The servers, available in both air-cooled and liquid-cooled variants, can house up to 192 Nvidia Blackwell Ultra chips, with configurations supporting as many as 256 chips per rack. Dell claims these systems can train AI models up to four times faster than previous models, marking a significant leap in performance for large-scale AI workloads.

At the heart of the new lineup are the PowerEdge XE9780 and XE9785 servers, designed to accelerate AI implementation from experimentation to production. These servers feature direct-to-chip liquid cooling, addressing the high power requirements of the Blackwell Ultra GPUs while improving energy efficiency. Dell’s PowerCool technology further enhances efficiency, and the updated network infrastructure includes the PowerSwitch SN5600 and SN2201, delivering up to 800 gigabits per second throughput-crucial for data-intensive AI tasks.

Dell’s strategy extends beyond hardware, with a focus on integrating networking and storage solutions to maintain profitability amid high production costs and fierce competition. The new servers are also future-proofed, supporting Nvidia’s upcoming Vera CPUs, which will replace the current Grace processors, and the Vera Rubin chips, set to follow the Blackwell series. This compatibility ensures enterprises can scale and upgrade their AI infrastructure as new technologies emerge.

In addition to server advancements, Dell introduced the ‘Pro Max Plus’ laptop, tailored for AI development. This device features a neural processing unit, enabling engineers to run large AI models locally without relying on cloud services. By expanding its AI Factory portfolio, Dell aims to make AI more accessible and accelerate enterprise adoption, positioning itself as a leader in the rapidly evolving AI infrastructure landscape.