🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Meta to build $10bn AI datacenter complex in Louisiana, complete with 7x750 MW natural gas turbines. Bummer on the environmental front, but they will buy 1.5 GW of zero-carbon energy to compensate. Press release: buff.ly/4ghz4Gp Energy details: buff.ly/4f1Aow3

Canada just announced a plan to spend up to $2B over five years to build out sovereign AI infra. $700M for private partnerships, up to $1B for AI infrastructure, and $300M to subsidize existing infra. Bold move. US is dragging its feet. buff.ly/4gc0xtq #HPC #AI

This is fascinating given DDN's (and Alex Bouzari's) bold stance against accepting outside investment and being anti-VC (see sramanamitra.com/2011/10/16/how…). Is DDN writing checks that it can't cash on its own (e.g., their deal with xAI), like Cray during the Exascale days?

I work on the biggest GPU supercomputers on the planet, and none use a parallel file system, full stop. As Jeff says, file-based storage offers nothing uniquely useful to training. Files are for people, and no people are touching data when training at scale.

Glenn K. Lockwood Glenn K. Lockwood quadrupled down on this topic here. Take a read, you may learn a thing or 4. blog.glennklockwood.com/2025/02/llm-tr…

This is a fantastically useful read about AI model training/inferencing and object storage vs parallel file systems - blog.glennklockwood.com/2025/02/llm-tr…

UK based #HPC benchmarking role at Microsoft Requires real experience with hands on HPC #benchmarking - porting, compiling, tuning, performance analysis etc. of scientific codes on HPC systems. buff.ly/fKfQz6j

Looking forward to moderating this #GTC discussion between Jeff Denworth and Glenn K. Lockwood on the diminishing role of file systems in #AI and #HPC - Register to attend - events.vastdata.com/nvidia-gtc-bre…

If you are going to be at #GTC this session has to be a priority to attend - 1. Because of the #HPC celebs involved - Nicole Hemsoth Prickett Glenn K. Lockwood Jeff Denworth 2. The important topic for #AI and #supercomputing - the future of #IO, #data and filesystems (or not)

Storage Innovations for AI Workloads panel at #GTC hosted by Rob Davis of NVIDIA with Vast Field CTO Kyle Lamb, among others. Davis: “none of these observations would have been made last year things are changing so fast”

#SymLink: GTC 2025 recap @GlennKLockwood bit.ly/425GuY8

Oracle to deploy a cluster of over 30,000 AMD MI355X as part of a multi-billion dollar #AI contract. Great to see big investments in competition here. Remember when these headlines used to start with a national lab’s name? datacenterdynamics.com/en/news/oracle…

Lots of rumors swirling about hyperscalers pulling back on datacenter leases (e.g., bloomberg.com/news/articles/…). Think it's a coincidence that these rumors coincide w/ NVIDIA announcing 600 kW racks for 2027?

Our good friend Glenn K. Lockwood has graciously let me share his insight into @Nvidia #gtc25 over at the More Than Moore substack. It's a deep look and a great read into certain aspects of the conference that most of us don't realise. morethanmoore.substack.com/p/nvidia-gtc-2…

According to this, Microsoft and NVIDIA are building the largest AI supercomputer in the world in Azure using GB200.

🗣️📢 We're thrilled to work with The U.S. Department of U.S. Department of Energy on the flagship NERSC-10 supercomputer! Powered by our most advanced liquid-cooled server tech & NVIDIA's next-generation accelerators - the "Doudna" system fuses #HPC & #AI to empower complex, integrated research

Differences between designing a #supercomputer for #AI and designing a #supercomputer for traditional #modeling and #simulation Posted in Glenn K. Lockwood's Digital Garden glennklockwood.com/garden/differe… #HPC #Network #Storage

I'm glad people find this useful. It is a lot of work to produce, and every year it seems to get harder to do.

We’ve got a request for information out on where we want to take Livermore Computing and other #HPC centers in the next five years. hpc.llnl.gov/fg-hpcc-rfi Check it out and send us your thoughts.

I may have been at that meeting. Hearty congratulations to all parties. I’m glad to see NERSC taking big swings and no longer settling for Lustre as the answer to every question.