On May 18, 2026, Sandia National Laboratories highlighted its reliance on new chip manufacturers like NextSilicon to secure computing power for high-precision scientific work, particularly amid a shift in the semiconductor industry towards AI-focused designs that may not meet their needs. As traditional suppliers, such as Nvidia and AMD, increasingly pivot to AI chips, Sandia faces challenges in obtaining processors capable of double-precision calculations essential for simulating nuclear weapon dynamics and other complex physics problems. Sandia’s ongoing collaboration with newcomers reflects a broader trend where national labs are considering startups as viable alternatives to larger chip vendors, ensuring continued access to critical technology necessary for fulfilling their national security mission.

Nvidia: Nvidia is a leading semiconductor company that designs GPUs and accelerated computing platforms widely used in artificial intelligence, high‑performance computing, and data centers. In this story, Nvidia is both a longtime Sandia partner for supercomputing and a source of concern as its chip roadmap leans toward AI, raising questions about double‑precision performance for scientific simulations even as it continues to signal support for HPC users.
Steve Monk: Steve Monk is the manager of the high‑performance computing team at Sandia National Laboratories, responsible for procuring and operating supercomputing systems that support the lab’s scientific and national security missions. In the article, he explains Sandia’s mounting pressure around supply chains and the need for chips that deliver precise double‑precision performance rather than being optimized primarily for AI.
Ian Cutress: Ian Cutress is the chief analyst at More Than Moore, a consultancy focused on semiconductor technology and industry trends. In the piece, he provides expert commentary on how Nvidia’s upcoming Rubin architecture appears to de‑emphasize double‑precision performance, heightening concern within parts of the HPC community.
James Laros: James Laros is a senior scientist at Sandia National Laboratories who leads programs focused on evaluating and advancing novel computing architectures for mission‑critical workloads. In the article, he describes Sandia’s engagement with smaller chip vendors like NextSilicon as a way to maintain diverse supply options and ensure future access to suitable supercomputing technology.
NextSilicon: NextSilicon is an Israeli semiconductor startup developing specialized accelerators that use a data‑flow–oriented architecture and dynamic reconfiguration to optimize complex HPC workloads. In this news, NextSilicon’s chips have been integrated into a test supercomputer at Sandia, passed key benchmarking milestones, and are being considered for more demanding nuclear security simulations.
Daniel Ernst: Daniel Ernst is the senior director of supercomputing products at Nvidia, overseeing strategy and product direction for Nvidia’s HPC‑oriented solutions. In this news, he represents Nvidia’s position that its chips remain designed to balance scientific computing needs with AI workloads, responding to concerns about changing performance priorities.
Advanced Micro Devices: Advanced Micro Devices (AMD) is a major chipmaker that develops CPUs, GPUs, and data center accelerators for general computing, gaming, and high‑performance computing. In the article, AMD appears as one of Sandia’s traditional HPC suppliers and is highlighted for offering a variant of its accelerators targeted at scientific workloads that require strong double‑precision performance.
Sandia National Laboratories: Sandia National Laboratories is a U.S. Department of Energy national lab that supports nuclear deterrence, national security, and advanced science through high‑performance computing, engineering, and experimental research. In this news, Sandia is evaluating new chip architectures such as those from NextSilicon and collaborating with established vendors like Nvidia and AMD to secure future supercomputing capacity for nuclear security and other precision scientific workloads.

HPC_vs_AI_shift: Recent industry coverage highlights growing tension between AI‑optimized accelerators and traditional high‑precision HPC needs, with some national labs warning that AI‑first chip designs may not suit physics and climate simulations.
Startup_HPC_role: Analysts and lab officials have noted that newer chip startups are increasingly being invited into national lab testbeds as potential alternatives or complements to the largest GPU vendors for future exascale‑class systems.
Energy_efficiency_trend: Across DOE‑aligned facilities, energy efficiency has become a central requirement for new supercomputers, driving interest in architectures and cooling approaches that reduce power use for double‑precision scientific workloads.