US DoE and AMD Launch $1B AI Supercomputing Initiative to Accelerate Nuclear, Fusion, and Quantum Research

The U.S. Department of Energy (DOE) has partnered with AMD to build two of the world’s fastest AI supercomputers, aimed at advancing nuclear power, fusion energy, and quantum technologies.

On October 27, AMD and the U.S. DOE announced two next-generation systems at Oak Ridge National Laboratory (ORNL) designed to expand America’s leadership in artificial intelligence (AI) and high-performance computing (HPC), the Lux AI supercomputer and the Discovery supercomputer.

• Strategic Partnership Overview
  • Investment: $1 billion collaboration between the U.S. DOE and AMD
  • Purpose: Construct two cutting-edge AI supercomputers for scientific and national security challenges
  • Partners: AMD (hardware), Hewlett Packard Enterprise and Oracle (infrastructure and software)
• Scientific and Technological Focus
  • Nuclear & Fusion: Accelerate simulations for next-gen reactors and fusion containment
  • Quantum Tech: Support quantum algorithm development and hybrid computing experiments
  • Medical Breakthroughs: Drug discovery and cancer modeling using massive datasets
  • National Security: Defense simulations, threat detection, and strategic planning
• AI Capabilities
  • Performance: Among the fastest AI systems globally, real-time data processing
  • Use Cases: Scientific discovery, energy dominance, and national security operations
• Geopolitical Implications
  • Reinforces U.S. leadership in AI and high-performance computing
  • Supports President Trump’s push to repatriate advanced tech infrastructure

Lux and Discovery are two next-generation AI supercomputers being developed by the U.S. Department of Energy (DOE) at Oak Ridge National Laboratory (ORNL), with deployment planned for 2026 and 2028 respectively. Lux will serve as the nation’s first AI Factory for science, while Discovery will be a second-generation exascale system designed to unify AI and high-performance computing.

• Lux AI Supercomputer (Deployment: Early 2026)

• • Purpose: First U.S. AI Factory supercomputer for science, energy, and national security
  • Architecture: AMD Instinct MI355X GPUs, AMD EPYC CPUs, AMD Pensando networking
  • Platform: Multi-tenant, cloud-like system for flexible AI training and inference
  • Use Cases: Fusion energy, quantum computing, cancer research, cybersecurity
  • Partners: ORNL, AMD, Oracle Cloud Infrastructure, HPE

• Discovery Supercomputer (Deployment: 2028)

• • Purpose: Second-generation exascale successor to ORNL’s Frontier system
  • Architecture: HPE Cray Supercomputing GX5000 and K3000 storage systems
  • Performance: Up to 10X productivity boost over current systems
  • Use Cases: Precision medicine, nuclear energy, aerospace, climate modeling
  • Strategic Role: Supports secure, federated AI infrastructure and sovereign AI leadership
• Combined Impact
  • Investment: $1 billion joint public-private funding
  • Strategy: Aligned with U.S. AI Action Plan for scientific and competitive leadership
  • Location: Oak Ridge National Laboratory (ORNL), Tennessee

Nvidia Unveils Desktop Sized Personal AI Supercomputer Called Digits

Nvidia recently unveiled Project DIGITS, a personal AI supercomputer designed to bring cutting-edge AI computing to researchers, data scientists, and students.

Project DIGITS features the new NVIDIA GB10 Grace Blackwell Superchip, offering a petaflop of AI computing performance for prototyping, fine-tuning and running large AI models.

With Project DIGITS, users can develop and run inference on models using their own desktop system, then seamlessly deploy the models on accelerated cloud or data center infrastructure.

Here are some key details:

Key Features:

• GB10 Grace Blackwell Superchip: At the heart of Project DIGITS is the GB10 Superchip, delivering up to 1 petaflop of AI performance.
• Unified Memory: 128GB of unified memory and up to 4TB of NVMe storage for handling large datasets.
• Power Efficiency: Developed in collaboration with MediaTek, the superchip offers industry-leading efficiency.
• Scalability: Two DIGITS systems can be connected to handle models up to 405 billion parameters.
• Software Stack: Preloaded with NVIDIA's AI software stack, including tools like NVIDIA NeMo for fine-tuning models and NVIDIA RAPIDS for accelerating data science tasks.

Availability:

Price: Starting at $3,000, Project DIGITS will be available from Nvidia and its partners starting in May 2025.

Usage: Designed for prototyping, fine-tuning, and running AI models locally, with seamless deployment to cloud or data center environments.

This innovation aims to make powerful AI computing accessible to a broader audience, empowering developers to engage and shape the age of AI.

Intel Says Its Aurora Supercomputer is the Fastest AI System in the World

On May 13, 2024, at International Supercomputing Conference (ISC) 2024, Intel, Argonne National Laboratory and Hewlett Packard Enterprise (HPE) announced that the Aurora supercomputer has broken the exascale barrier.

The Aurora Supercomputer has made a significant breakthrough by surpassing the exascale barrier, achieving over a quintillion calculations per second, which is an incredible milestone in computing. Announced at the ISC High Performance 2024 conference, Aurora, developed by Argonne National Laboratory in collaboration with Intel and Hewlett Packard Enterprise, has registered 1.012 exaflops.

This achievement not only places Aurora as one of the world's fastest supercomputers but also highlights its prowess in artificial intelligence, achieving 10.6 AI exaflops on the HPL-MxP benchmark. The supercomputer's capabilities are expected to transform scientific research, enabling breakthroughs in various fields such as climate science, energy storage, and fusion energy.

The emphasis on an open ecosystem in high-performance computing (HPC) and AI is crucial as it allows for greater collaboration and innovation, fostering advancements that can benefit a wide range of scientific and technological domains.

 

Aurora supercomputer has broken the exascale barrier and leads as the highest ranked supercomputer for high performance computing and artificial intelligence convergence. (Credit: Argonne National Laboratory)

The Aurora supercomputer is an expansive system with 166 racks, 10,624 compute blades, 21,248 Intel® Xeon® CPU Max Series processors and 63,744 Intel® Data Center GPU Max Series units, making it one of the world's largest GPU clusters.

Aurora also includes the largest open, Ethernet-based supercomputing interconnect on a single system of 84,992 HPE slingshot fabric endpoints.

“The Aurora supercomputer surpassing exascale will allow it to pave the road to tomorrow’s discoveries. From understanding climate patterns to unraveling the mysteries of the universe, supercomputers serve as a compass guiding us toward solving truly difficult scientific challenges that may improve humanity", said Ogi Brkic, Intel vice president and general manager of Data Center AI Solutions.

Aurora's hardware components were completed last year, and it's not yet fully operational. However, it has already shown a 2X performance advantage over AMD's MI250X GPUs, and a 20% gain over NVIDIA's H100 GPUs in shared benchmarks.

Once fully operational, Aurora could potentially take the first spot from Frontier on Top500's flagship HLP ranking, as there's a contractual target number that is faster than Frontier.