Quantum #97

You probably missed it in a GPU obsessed world, but new AMD 5^th gen (Turin) CPU VMs are GA and Intel Xeon 6 VMs are now in public preview on Azure and Google released the Axion (ARM) powered N4A last week. For the 80% of us that still need to run CPU based HPC workloads this is still relevant even if it isn’t exciting.

The University of Cambridge got a cool 36 million to build out the UK’s next supercomputer. However, this a little more than a tenth of what Isambard AI cost (£225million) so we won’t be seeing the next UK AI Factory or Exascale machine me thinks.

We dropped another couple of articles in our mini-series on how to improve efficiency and utilisation in large scale HPC and AI clusters, focusing this time on the concepts of workload fragmentation and compute elasticity.

For a more comprehensive look at supercomputing in AI though, Jordi Torres from the Barcelona Supercomputing Centre published his extensive book

Home

Foundations, Architectures, and Scaling Deep Learning Workloads

Jordi TorresJordi Torres

Lastly, HPC Wire published a piece about how more HPC is more prevalent on cloud. Part of me wonders if this is really a greater adoption of cloud for HPC in the forms it has traditionally been used for, or is this just an increase in the size of the market with AI and the new workload being primarily cloud focused (because that’s where most of the GPUs are)… hmm…

In The News

Updates from the big three clouds on all things HPC.

HPC Cloud Updates WE 01 Feb 2026

Updates to AWS, Azure & GCP in the last week relevant for HPC practitioners. We get Intel, AMD and ARM powered VMs from Azure and Google week.

Flux by HMx LabsHamza

Next Platform is back on form with a really interesting analysis of the relationship between Microsoft and OpenAI. Great quote from the article:  Big Bill is a lot more dependent on OpenAI than OpenAI will be on Big Bill. Which reminds us of the old saying: “If I owe you $250, that is my problem. But if I owe you $250 billion, that is your problem.”

Microsoft Is More Dependent On OpenAI Than The Converse

Everyone is jumpy about how much capital expenses Microsoft has on the books in 2025 and what it expects to spend on datacenters and their hardware in

The Next PlatformTimothy Prickett Morgan

DAWN at Cambridge gets a £36m injection from UK Gov to scale up

UK Invests £36M in AMD-Powered Supercomputer at the Univ. of Cambridge

The UK government announced it has injected £36 million to increase the power of one of the UK’s leading supercomputing centres sixfold. This includes backing a new National Computational Resource supercomputer .... - Read more from Inside HPC & AI News.

Inside HPC & AI News | High-Performance Computing & Artificial Intelligencestaff

HPC on the cloud is becoming more prevalent. Though I do wonder how much of the HPC is AI workloads and how much is other more traditional HPC workloads.

https://www.hpcwire.com/2026/01/27/clouds-are-on-your-hpc-horizon/

From HMx Labs

The next two in the series on improving utilisation. This time focusing on relative job sizes and capacity elasticity

Improving Utilisation in HPC and AI – Part 3

Part 3 in the mini series on improving utilisation rates in HPC and AI. The first (but not the last) on dealing with fragmentation impacting utilisation rates.

Flux by HMx LabsHamza

Improving Utilisation in HPC & AI – Part 4

Part 4 of mini series on improving utilisation (aka reducing unit cost) looks at adjust the supply of compute to deal with demand variability

Flux by HMx LabsHamza

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