Quantum #61

Issue #61 of the weekly HPC newsletter by HMx Labs. More HPC Club merch, AI Supercomputers and Microsoft proves it still can’t name things.

Quantum #61

Just last week I was told by a reader that this newsletter is, and I quote,

“my favourite source of competitor information.”

😁 😁 I love it. So, if you’re not signed up yet, what’s holding you back? Equally if there are things I should be talking about, and I’m not then let me know.

At the time of writing this (on a sunny Sunday morning this time) there are exactly 3 tickets left for HPC Club. If you were planning on coming along and haven’t booked your spot yet, then I wouldn’t hang around.

We have more stickers and swag too.

Talking of signing up and HPC Club, at the moment we operate a mailing list for HPC Club (signed up for on the HPC Club site) which only sends an email once or maybe twice for each event. We also have this newsletter which includes updates on HPC Club but these don’t go to the aforementioned mailing list. Would it make more sense to carve out another option on this newsletter to get email only about HPC Club and ditch the other mailing list? How would people feel about that? Too closely tied to HMx Labs? 

Have a good week all. As ever, the rest of your HPC News is below 👋


In The News

Updates from the big three clouds on all things HPC. Worth reading just for the Phil Karlton references if I do say so myself.

HPC Cloud Updates WE 25 May 2025
Updates to AWS, Azure & GCP in the last week relevant for HPC practitioners. There are only two hard things in Computer Science: cache invalidation and naming things… and Microsoft proves it’s still terrible at one and leaves the other to you.

A little reminder that it’s the output of HPC that matters…

HPC in Action: Academia, Industry, and Government Moving the Needle
High-performance computing advancements can assist companies with large quantities of data in accelerating their digital transformations. Knowing what they have done so far can inspire others to follow their lead […]

400,000 GPUs in OpenAI’s new Texas site? Reading a bit closer that’s across 8 buildings with 50k in each one. It doesn’t say which exact model of GPU but a B200 is theoretically capable of about 40 TFLOPs (FP64)… which means, if my terrible mental arithmetic is correct, we’d be looking at 8 exascale installations. Each one about the size of El Capitan, the number one supercomputer on the Top500. 8 of them 🤯

But then it is OpenAI and you can’t believe a word that comes of Sam A’s mouth so who knows.

Report: $15B OpenAI Data Center in Texas Will House up to 400,000 Blackwells - High-Performance Computing News Analysis | insideHPC
Amid reports of hundreds of billions in AI data center investments in the Middle East, development of AI factories in the U.S. continues apace. The Wall Street Journal reported today that OpenAI has secured more .…

From HMx Labs

My favourite paragraph from chapter 1 of the the intro to supercomputing book I’m working on:

What Makes a Supercomputer?
Extract from Chapter 1 of Intro to Supercomputing and HPC

A rhetorical question about the future of AI and not so rhetorical one about how much detail to include in these posts

Can We Ask the Right Question?
Suppose we do manage to create AGI or even super intelligence. Do we even know what to ask it? Will we understand its reply?

Know someone else who might like to read this newsletter? Forward this on to them or even better, ask them to sign up here: https://cloudhpc.news