US National Labs Tap Startups as AI Reshapes Chip Industry
In an unremarkable structure at Kirtland Air Force Base in the arid highlands of New Mexico, liquid-cooled supercomputers buzz and whir through intricate mathematical challenges the U.S. government aims to address: modeling the trajectory of hypersonic nuclear weapons in the Earth's atmosphere or predicting the effects of a nuclear warhead detonating in proximity to another.
For over ten years, the chips managing this covert and intensive task were sourced from well-known semiconductor companies like Nvidia or Advanced Micro Devices.
However, as those companies are progressively creating their chips for artificial intelligence and encountering supply shortages, the managers overseeing the systems at Sandia National Laboratories, which runs the machines at Kirtland and is one of three U.S. labs responsible for developing and maintaining the nation's nuclear weapons arsenal, are growing more uncertain about how they will secure computing power.
"The pressure we’re experiencing currently comes from the computing sector and the supply chain," stated Steve Monk, manager of Sandia's high-performance computing team, describing the difficulty in obtaining chips that fulfill his requirements. Considering the future, it feels somewhat anxious regarding our capacity to fulfill the mission."
Also Read: How This Techie Turned Visa Struggles into Startup Success
The lab's situation illustrates how the competition for superior AI chips is unintentionally allowing smaller companies, like NextSilicon, an Israeli startup testing its chips in a Sandia program, to enter markets previously held by major firms.
Also Read: 5 Key Leadership Appointments across Global Firms in March 2026
It also highlights Sandia's significant role in nurturing and influencing new computing technologies while having collaborated with Nvidia during its ascent in supercomputing and continuing to work with Nvidia on novel memory technology.
Also Read: How US–China Tariffs Are Disrupting Textile Trade
A significant worry for Sandia officials is the concept of double-precision floating point computation, a technical phrase for the ability to calculate extremely large and extremely small numbers without sacrificing accuracy due to rounding errors. For years, Nvidia and AMD competed to dominate in enhancing that type of computing, securing supercomputing contracts with universities and government laboratories.