US National Labs Seek New Chip Suppliers for Supercomputers

In a nondescript building on Kirtland Air Force Base in the high desert of New Mexico, liquid-cooled supercomputers hum through some of the most complex math problems the U.S. government seeks to solve: simulating how hypersonic nuclear weapons move through the atmosphere, or what happens if one nuclear warhead detonates near another.

For over a decade, the chips handling this secretive and demanding work came from mainstream semiconductor firms like NVIDIA (NVDA) or Advanced Micro Devices (AMD). But with those companies increasingly designing chips for artificial intelligence and facing supply shortages, managers at Sandia National Laboratories—which operates the machines at Kirtland and is one of three U.S. labs tasked with maintaining the nation's nuclear arsenal—are increasingly unsure how they will find computing power for their high-precision scientific work.

Details

Sandia relies on specialized chips capable of complex simulations requiring extreme accuracy, tasks distinct from training AI models that NVIDIA and AMD now prioritize. Surging demand for AI chips has led to supply shortages and higher prices, making it harder for the lab to procure needed quantities.

Context

This comes as the chip industry undergoes a major shift toward AI. NVIDIA has posted record revenues in recent quarters driven by AI chip sales, while scientific and government applications compete for the same silicon.

What It Means for Investors

This trend highlights growing demand for high-performance computing (HPC) chips beyond AI, potentially opening opportunities for niche or emerging players. It also underscores the risk of NVIDIA and AMD focusing on AI at the expense of traditional customers like government labs.