Quantum #110

Issue #110 of the weekly HPC newsletter by HMx Labs. A new Exascale supercomputer, silicon still running short and yet we still haven’t figured out how to do heterogenous compute.

Quantum #110

We have a new Exascale class supercomputer on the way. Not that Exascale really means very much anymore given all the hyperscalers have systems that are easily in this range but neither talk about them in these terms nor submit Top500 results for them. Ha, and I guess neither will LineShine 😆. In any case what makes LineShine interesting is that it’s not chock-full GPUs or even APUs. It’s an honest to goodness CPU based supercomputer of old. With ARM CPUs! Oh, and HBM too so that’s kinda different and cool.

 Talking of CPUs, even Next Platform is talking up CPU shortages. Honestly, I still don’t buy the agentic AI story here. I’ve yet to hear a single actual use case and even this story talks about it as though we need more CPUs in the same racks and chassis as the GPU. Why? This makes no sense. You just wouldn’t design any kind of HPC system where you are interspersing your HPC with slow disk and network bound actions as implied by the word “agentic”. Financials risk system excepted of course 🤣. Seriously though, it makes no sense. The slow IO bound side of the system would just offload the AI compute to a cluster than can target high utilisation level rather than be constantly juggling compute and IO bound operations. 

And since we’re back onto accelerated compute, Uncle Jensen’s not going to be happy that Deepseek has decided to optimise for Huawei’s Ascend 950. What this highlights for me more than anything else though, is quite how far away we still are from a world where we can use heterogenous compute. It’s just about possible in the CPU only space, but for accelerated compute it seems like a developer’s nightmare. Arguably, no one other than Nvidia’s ecosystem has enough traction, with the number of possible alternatives and the lack of anything even resembling a standard I don’t see this improving any time soon. And the only person that’s good for is Uncle Jensen. 😢

Oh, if you’re looking for your HPC release notes this week… they’re coming soon in a new format. For an explanation why and a hot take on Kubernetes see the end of the newsletter.


In The News

AI-Driven CPU Shortage Saves Intel’s Financial Cookies
If you have a few pallets of datacenter CPUs sitting in a barn somewhere, and they have a reasonable…

https://www.reuters.com/world/china/big-chinese-tech-firms-scramble-secure-huawei-ai-chips-after-deepseek-v4-launch-2026-04-29/

https://www.hpcwire.com/2026/04/28/china-unveils-2-exaflop-all-cpu-lineshine-supercomputer/


From HMx Labs

Should we really be running HPC workloads on Kubernetes?

Kubernetes’ Place in a HPC World
Does using Kubernetes to run HPC workloads really make sense? I have an opinion that I don’t usually share online for fear of being ridiculed. But yesterday I shared it with a stranger and lived. Plus, its Friday so even if I’m totally wrong on this everyone will

A better way to try and give you all the release notes from HPC vendors. Coming very soon!

HPC Release Notes Automation
Automating the weekly HPC release notes process which means it can get more comprehensive!

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