AI-HPC Convergence
Articles (19)
Brookhaven's GridFM 2.0 Award Tests AI Against the Interconnection Bottleneck
DOE is funding a grid model designed to evaluate 1 billion scenarios a day. But queue delays arise as much from process, staffing, speculative requests, and construction as from solver time.
EuroHPC's Largest AI Factory Supercomputer Contract Has No Published Absolute Performance Figure
Bull's €387.8 million LUMI-AI contract names AMD MI430X and EPYC hardware but gives no accelerator count or exaflops rating. CSC told SCN the count waits on AMD's final MI430X specifications.
Lux Comes Online in October With No Path Into the TOP500 Top 10. The Next US Shot at No. 1 Arrives in 2028.
ORNL's first Genesis Mission supercomputer misses the top ten even at perfect Linpack efficiency. No US system has published the FP64 to challenge China's LineShine.
Makau's Peak Sits Below the TOP500 Entry Threshold. Its Center's Stated Mission Is Evaluation, Not Capability.
MHPCC cut Makau's ribbon in August, months after the machine began carrying pioneer workloads. A system 13 times its size arrives at the same center this year.
Schrödinger's AI Co-Scientist Is Already Its Supercomputer's Main User
A month after launch, 80% of jobs on Schrödinger's internal supercomputer arrive through Bunsen. Schrödinger expects customer demand to follow.
Genesis-Science-1 Pairs DOE Science With Compute Secured by Arcee AI
Arcee AI leads model development while DOE labs contribute scientific materials and evaluation. The agreement, rights and release criteria remain undisclosed.
Argonne Moves ChemGraph Leaderboard to Monthly Runs as It Studies Score Variance
Across 71 public runs, GPT-4o-labelled scores ranged from 62.5% to 97.5%. Argonne says the files cannot always distinguish model behavior from system failures.
DOE's Genesis Mission Is a Federation Bet Built Over New Supercomputers
The first AmSC funding call centers on integration, even as Genesis expands DOE’s AI compute footprint.
Why Pharma Is Building Its Own AI Supercomputers
Biology models now train on proprietary experiments and feed results back to the lab. That closed loop strengthens the case for owning the machine.
Why Switzerland Put a 100-Petabyte NASA Replica Beside It’s Alps Supercomputer
ETH Zurich’s year-long transfer was not a model launch. It was a decision to keep a continuously updated Earth-observation archive close to Swiss compute, and less dependent on upstream US data services.
ARPA-H Is Funding In Silico Drug Testing. Its Compute Footprint Is Still Opaque.
FDA is expanding non-animal methods while ARPA-H funds human drug-safety models. Public materials omit runtime, hardware, and deployment cost.
The Mission Supercomputer Predates Genesis. The AI Framing Around It Does Too.
Mission and its AI framing predate Genesis; its public requirements remain centered on stockpile simulation and application performance.
Why Simulation Supercomputers Are Being Pitched as AI Infrastructure
Simulation still fills the world's research supercomputers. Funding and vendor roadmaps have tilted to AI, and some practitioners feel pressure to pitch their machines that way.
Time to First Token Is a Real Metric. It Isn't the One That Defines the Era.
Time to first token is the AI industry's favorite candidate for era-defining metric. It is a real whole-stack signal, not a stand-in for business success.
The FP64 Debate Didn't Produce a Winner. It Produced a Joint Byline.
Dongarra, Hoefler, and Matsuoka now ask whether supercomputing needs GPUs at all, and the CPU-only machine atop the Top500 is their test vehicle.
Spectra Clears Sandia's Supercomputer Acceptance. The Fall Mission-Code Gate Is the Real Test.
NextSilicon's Maverick-2 dataflow accelerator has met Sandia's Vanguard system-acceptance requirements on HPCG, LAMMPS, and SPARTA. The harder question comes this fall, when Sandia decides whether to move Spectra toward more demanding, mission-like ASC supercomputing workloads.
MRC Gives Open Ethernet Its First 75,000-GPU Production Proof Point
The 50-author MRC paper gives Ethernet its first multi-vendor, open-spec, production-trace answer to the one argument InfiniBand had left at frontier-training scale.
The training stack is starting to optimize itself
Anthropic’s 2.9× to 51.9× training-optimization curve signals that AI training infrastructure is becoming machine-optimizable, raising rebound demand and control-plane risks for HPC operators.
-800x600.png%3Fprefix%3Dmedia&w=3840&q=75)
Japan's Next Flagship Machine Abandons the Top500 Chase
FugakuNEXT pairs Fujitsu MONAKA-X CPUs with NVIDIA GPUs, ending Japan's all-Arm sovereign architecture and betting on throughput over benchmarks.