By Hamza in Quantum Weekly — Jun 29, 2026

The SC #118

Issue #118 of The SC – An ISC26 Special with updates not only from the world of supercomputing from last week but also from ISC26

It would be hard to cover what was important in supercomputing and HPC last week without talking about ISC 26. So, this edition of The SC brings blends the news from Europe’s largest supercomputing conference with everything else that happened last week.

The opening keynote talked about something I’ve been harping on about for almost two years at this point. Homogenous supercomputing is dead. Well maybe not quite yet but the future is absolutely heterogenous. Whilst two years ago that meant having to deal with multiple CPU and GPU manufacturers and knowing which one to pick, today that means a range of possible CPU architectures, GPU architectures and other accelerators including quantum and neuromorphic computing. The framing here, of quantum as another accelerator, in itself was quite interesting. And quite correct.

Quantum computing dominated not only the opening keynote but in many regards the conference itself. Where last year storage vendors took the most prominent position on the conference floor and competed for attention, this year (apart from Nvidia of course) that honour went to quantum computing vendors who dominated by number if not by placement.

My conversations (with multiple vendors) were all of course very optimistic and pointed to a bright future for quantum. Unlike AI however, it is far harder to draw the line on where reality ends and the hype starts with quantum computing. My own interests lie with its applicability to financial risk analytics and while many references were made to papers where this has been shown as feasible, I think this is an area that will need some actual work to be done (by us). Watch this space, I guess? If there’s enough interest, I guess we will have some work to do in separating the wheat from the quantum chaff.

Non ISC news about Microsoft’s topological qubits with Marjorana 2 seemed to echo all my conversations at ISC. i.e. they don’t exist. In fairness to Microsoft, I guess those conversation were all with their competitors in this space 😁

The release of the Top500, even with a new machine taking the number one spot this year, hardly featured in my conversations at all. Perhaps this speaks to the decreasing relevance of the Top 500? The Top500 itself is changing too with control passing over to ACM. I wonder if this will improve things.

China’s LineShine took the top spot, displacing El Capitan which had held it since November 2024. The first thing that stuck me about this was that LineShine had even bothered to submit to the Top500. It has been a number of years since China had stopped participating over increasing tensions between it and the USA. The decision to participate now, and with a CPU (ARM with HBM, quite interesting) only system I think speak volumes.

Overall, much like last year, the conference itself didn’t feature any new releases from vendors. It was no longer focused on attempting to be relevant in the wider AI news cycle either and the main floor felt less like a plumbing trade show (though there were still suppliers of liquid cooling solution present) than last year too. AI did still feature heavily but more in its application within HPC communities and the adoption of technology developed for AI into scientific computing.

Which segues nicely into my next point. The keynotes, presentations and talks all feel very much like a community focused on science and academia. This is quite distinct from the highly commercial nature of the main show floor, and I felt that there’s a trick missing to build a better bridge here. To be clear, I don’t mean letting vendors take over the talks. That would be terrible. I do mean having more commercial users of supercomputing presenting and sharing ideas. I know this is always difficult to achieve but that just makes it all the more valuable. Supercomputing is so much bigger than just science and academia, we need to include more of those people in the conversation.

In other non ISC related news, OpenAI and its partnership with Broadcom has a name for their AI inference accelerator, Jalapeno. There seems to be little detail about it beyond this though.

Lastly, for the finance nerds like me, QuantMinds London is now open for booking with reduced prices till early August.