By Hamza in Quantum Weekly — Feb 9, 2026

Quantum #98

Issue #98 of the weekly HPC newsletter by HMx Labs. Intel gives us Z-Angle RAM, AMD makes more money on GPUs than CPUs, we get Xeon 6 disk optimised VMs from AWS

I’ve been down with the lurgies this weekend and in periods of lucidity I managed to watch reruns of Silicon Valley and vibe code an ATS to help us track our hiring instead of the spreadsheet we’d been scraping by with so far.

Of course it’s not going to put any SaaS companies out of business. Ever. It is more than likely a security nightmare so there’s no way on earth it will ever be exposed to the internet either (strictly intranet only). But still it’s about a million times better than trying to use Excel and OneDrive.

And herein lies the dilemma. This is completely at odds with what I’d want to achieve with our little AI first coded HPC scheduler experiment, HAL. Understanding the difference between the two is key and it seems not many analysts (read as influencers/ commentator) are not making that distinction. That’s not to say one is correct, and one is not. Clearly both have their place. Any analysis of future jobs or AI demand though really needs to understand this distinction.

I think this probably deserves a fuller write up of its own but its serves as a useful backdrop to the news cycle which is full of both claims of cloud providers being at capacity (and H100 prices rising) and GPU utilisation rates of 30% and warehouses stacked full of GPUs still in their boxes.

The HPC specific news is a little thin on the ground this week but Z-Angle memory from Intel sounds interesting and I do wonder what AMD making more on GPUs than CPUs now says too. Oh and we got the disk optimised versions of AWS’ latest Xeon 6 powered C8/R8/M8 family.

In The News

Updates from the big three clouds on all things HPC.

Z-Angle memory from Intel, well at least we’re not talking about RAM prices. HPC Wire also have some more details

In itself this probably isn’t that interesting… but is it a sign of things to come? Will compute in future be accelerator dominated (at least in terms of cost)