Nvidia, Dell to supply next US gov't supercomputer by 2026

The US Department of Energy will launch a new supercomputer, Doudna, in 2026 using Nvidia and Dell technology.

Doudna will use Nvidia’s Vera Rubin chips integrated into Dell’s liquid-cooled servers to support 11,000 researchers across various scientific fields, including chemistry, physics, and biology

The system is named after Nobel laureate Jennifer Doudna, who highlighted the growing link between biology and computing.

The Energy Department uses supercomputers for research and to help manage the US nuclear weapons stockpile.

Nvidia CEO Jensen Huang said the importance of scientific computing amid concerns over exporting AI chips to China.

🔗 Source: Reuters

🧠 Food for thought

1️⃣ The DOE’s enduring dual mission in supercomputing

The Department of Energy has maintained a 70+ year leadership role in supercomputing, dating back to the Manhattan Project when the ENIAC computer supported early nuclear weapon feasibility studies 1.

This newest “Doudna” supercomputer continues the DOE’s consistent pattern of developing dual-purpose computing infrastructure that advances both scientific discovery and maintains the U.S. nuclear arsenal, as explicitly mentioned in the announcement.

The historical evolution from the Cray-1’s 160 megaFLOPS in 1976 to today’s exascale systems represents a million-fold increase in computational power, enabling increasingly sophisticated nuclear simulations and scientific research 1.

The naming of the system after Jennifer Doudna, a Nobel laureate in gene-editing, rather than a weapons scientist, reflects the DOE’s strategic emphasis on its scientific mission despite the system’s nuclear security applications.

2️⃣ Supercomputing as a geopolitical chess piece

The announcement of “Doudna” comes amid an intensifying global supercomputing race, with the U.S. currently leading with 173 systems in the TOP500 rankings and a combined performance of 6.48 exaflops 2.

Nations increasingly view sovereign AI infrastructure as essential for technological independence, with countries like Denmark launching supercomputers (Gefion with 1,528 NVIDIA H100 GPUs) specifically to develop technologies aligned with their national interests and values 3.

The DOE’s partnership with Nvidia on this system is particularly notable given the ongoing tensions over U.S. export controls that have limited Nvidia’s ability to sell advanced AI chips to China, with CEO Jensen Huang reporting significant financial impacts 4.