NVIDIA accelerates artificial intelligence across more than 80 new science systems

At SC25 NVIDIA said more than 80 new scientific systems worldwide now run on its accelerated computing platform, contributing a combined 4,500 exaflops of artificial intelligence performance and including the 300-petaflop Horizon system at TACC.

At the SC25 conference in St. Louis, NVIDIA announced that over 80 new scientific systems powered by its accelerated computing platform have been unveiled globally in the last year, contributing a combined total of 4,500 exaflops of artificial intelligence performance. The company positioned its full-stack platform of GPUs, CPUs, DPUs, NICs, scale-out switches, CUDA-X libraries and NVIDIA AI Enterprise software as the foundation enabling large-scale scientific modeling across fields such as climate, materials, robotics and digital biology.

Among the new systems is the horizon supercomputer at the Texas advanced computing center, a 300-petaflop academic system slated for 2026 that will use NVIDIA GB200 NVL4 and NVIDIA Vera CPU servers with Quantum-X800 InfiniBand networking. Horizon is configured with 4,000 NVIDIA Blackwell GPUs and can deliver up to 80 exaflops of artificial intelligence compute at FP4 precision. TACC plans to run molecular dynamics and AI-enhanced simulations for disease research, astrophysics models tied to James Webb telescope data, atomic-scale materials studies and seismic wave mapping. John Cazes, director of high-performance computing at TACC, said the system will transform how researchers pursue artificial intelligence driven initiatives at scale.

The U.S. department of energy is partnering with NVIDIA to build seven new artificial intelligence supercomputers at Argonne and Los Alamos national laboratories. Argonne will host systems including solstice with 100,000 NVIDIA Blackwell GPUs, capable of roughly 1,000 exaflops of artificial intelligence training compute at scale, and equinox with 10,000 Blackwell GPUs, while other systems will support inference and workforce development. At Los Alamos, mission and vision systems will use the NVIDIA Vera Rubin platform and Quantum-X800 networking to run classified NNSA workloads and open science including foundation models and agentic artificial intelligence, expected in 2027. International deployments include JUPITER in Germany achieving exaflop HPL performance, blue lion, gefion and isambard-artificial intelligence in Europe, RIKEN and fugakunext in Japan, tokyo university of technology and ABCI-Q for quantum research, and major initiatives in south Korea and Taiwan deploying tens of thousands of NVIDIA GPUs across sovereign clouds and factory supercomputers. These systems are presented as tools to accelerate scientific discovery by combining scale, unified software and specialized networking.

70

Impact Score

OpenAI expands cyber model access through trusted program

OpenAI has introduced GPT-5.4-Cyber as a restricted model for cybersecurity professionals, widening access through its Trusted Access for Cyber program. The release highlights both the defensive value and misuse risks of more capable Artificial Intelligence tools in security work.

Chinese tech firms and Li Fei-Fei push world models forward

Chinese tech companies and Li Fei-Fei’s World Labs are accelerating work on world models, a field focused on helping Artificial Intelligence learn from and interact with physical reality. Alibaba’s new Happy Oyster system targets real-time virtual world creation with more continuous user control.

UK launches Sovereign Artificial Intelligence backing for startups

The UK government has unveiled Sovereign Artificial Intelligence, a state-backed initiative aimed at helping domestic startups build, scale and stay in Britain. The first support includes an equity investment in Callosum and supercomputing access for 6 additional companies working across drug discovery, infrastructure and national security.

Intel joins terafab to expand Artificial Intelligence chip manufacturing

Intel has joined Elon Musk’s Terafab initiative, bringing chip manufacturing and packaging capabilities to a project designed to scale compute production for Artificial Intelligence and robotics. The effort reflects a broader shift from securing chip access to controlling how compute is built and allocated.

Contact Us

Got questions? Use the form to contact us.

Contact Form

Clicking next sends a verification code to your email. After verifying, you can enter your message.