@HPC Podcast Archives - OrionX.net - HPC News Bytes – 20260209

Episode Date: February 9, 2026

- Sovereign AI: what is it, and does anyone have it? - Bullish on Eviden: Europe’s top system company restores old name - Intel to build server GPUs of its own - MIT Technology Review AI Predictions [audio mp3="https://orionx.net/wp-content/uploads/2026/02/HPCNB_20260209.mp3"][/audio] The post HPC News Bytes – 20260209 appeared first on OrionX.net.

Welcome to HPC Newsbytes, a weekly show about important news in the world of supercomputing, AI, and other advanced technologies. Hi, everyone. Welcome to HBC Newsbytes. I'm Doug Black of Inside HBC, and with me as Shaheen Khan of OrionX.net. There was an interesting thought piece recently in the Wall Street Journal called The New Bipolar World of AI. The article sought to define what AI sovereignty is and who has it, And according to the authors, only two countries do have it, the U.S. and China. Quote, sovereignty here does not mean access to powerful tools or building applications on top of them, the author said. It means the ability to design, train, operate, secure, and deploy foundational AI systems capable of highly advanced functions and national defense and other sensitive areas of the state.

And here's their central point, without external permission or dependence, unquote. They place their observations within a recent global context of increasing competition, hostility, and economic separation between the U.S. and China, pointing out two countertrends. Quote, even as globalization reversed, connectivity didn't. Data, content, and influence continued to move across borders. The result is a world that is simultaneously de-globalizing and hyper-connected. In this strange new environment, modern AI emerged, and by AI, they're really referring to ChatGPT, November 2020. Adoption of ChatGPT has, quote, occurred at an unprecedented pace. Within months, billions of people had access to these powerful cognitive tools. Control over the underlying systems, however, concentrated into the hands of a few firms in the

U.S. with China as the only rival ecosystem. I thought it was an interesting piece, but I also know or expect that you don't completely agree with it. Well, AI buildout has become mandatory. Fast deployment matters. Building it yourself matters. There are risks everywhere. But supply chains are complex, so it's impossible to bring it all home. Unless you are a big country, it's hard to even bring much of it home.

And even if you can, is that the best investment? Maybe it's better to catch the next wave. We see a lot of that in quantum computing and biotech, for example. The major geopolitical spheres are all working on the above. China and Europe stand out, while several others like Japan and India are big and growing technology forces, and other countries are starting with bigger and more focused AI systems within their borders. So fast-moving technology is fueling fast-changing geopolitics, the equation is different for each country as they try to get it right.

The questions for all countries are, what do you buy? What do you build? How do you see the supply chain changing? And where do you invest in the future? And there are two supply chains, the physical and the digital, like the article says, hyperconnected but fragmented, atoms and bits, so to say. And they cross each other and work in parallel and are both really hard. Countries can invest in local technology and entrepreneurship. be doing that anyway. They can favor local vendors regardless of whose technology they sell or where it is manufactured just to favor the local player. Or they can come up with ways to achieve performance with less reliance on foreign technologies, which all of them will try to do. Many are

passing laws to enforce such policies. It can get harder with software, but the open source model can provide an entry ticket. As our conversation with Paul Block of DDN covered last week, a lot that starts with the data, where it must reside and how digital rights are managed and enforced. This is a good thread to catch up with the rebranding and renaming of France-based supercomputer company Eviden as Bull, a venerable name in European HPC for decades. Bull went through many mergers and acquisitions and name changes since its founding in 1931. In 2014, the French IT company Atos took a controlling stake. And less than three years ago, Atos announced that its supercomputing unit would be called Eviden, which has established itself as the leading European

HBC Systems Company, standing up among other systems, the continent's first exoscale supercomputer called Jupiter at the ULIC Supercomputing Center in Germany. The Bull acquisition was then essentially undone when Eviden was carved out again last year. and a deal with the French government, which was interested in securing strategic national assets. The new bull will include supercomputing, AI, and cybersecurity. You know, there was a time when European computer vendors had their own architecture, and there were big players in the market. ICL in the UK, Siemens-Nicksdorf in Germany, Olivetti in Italy, and Bull in France.

And there's a longer history before that. The bull name refers to punch card technology, pioneer Frederick Rosen Buhl, a Norwegian engineer. I seem to recall we covered it here. But anyway, Bull was always a highly respected technology company with serious system design capabilities. So the name change is welcome and comes with good brand equity. ASML, including Zyce as a key supplier to ASML, ARM, mistral, SAP, and now Bull make a great anchor portfolio and technology foundation for European sovereign tech. The long and storied saga of Intel and Data Center GPUs took a turn last week

when it was reported that CEO Lipu Time said the company plans to build graphics, processing units, again. Under previous leadership, Intel has made several attempts to take on NVIDIA in the GPU arena, including with the much-publicized Data Center Macseries GPU code name Pontavecchio, which is used in one of the three X-Scale superiors. supercomputers in the U.S., this is the Aurora system at Argonne National Lab. Recurrently, Intel's attempts in data center GPUs has been stymied by delays, cancellations, and product roadmap changes, even as NVIDIA and AMD have charged ahead with GPU offerings

at rapid cadence. More recently, there are reports that Intel is in the process of acquiring AI chip company, Sanbanova, but how that acquisition will fit into the company's overall GPU strategy remains unknown. Intel does a good job on the desktop and on-chip GPUs for gaming and increasingly for AI. And maybe another acquisition can accelerate its presence in data center AI. But anyway, it's clear that Intel cannot sit it out. Liputan managed to Wu Qualcomm executive Eric Demers, where he was SVP of engineering. He joined Intel last month as SVP of GPU architecture as part of their data center group headed by Kework Kishishian, who himself joined Intel last September coming over from ARM.

As a prelude to their upcoming conference called MTECH AI 2026 in April, on the MIT campus, the editors of MIT Technology Review have published five predictions for AI. They are, to run through them briefly, AI becomes invisible. This is a make or break year for AI. AI agents, LLMs will drive scientific breakthroughs, the rise of vibe coding, and AI reasoning levels up. Of course, the use of LLMs and science has become more and more important over the last five years, and you were mentioning incredible progress with vibe coding.

So it seems like a good list of topics, if not quite a prediction. You know, one reason AI is so irresistibly important is because it promises to impact everything. So it's hard to narrow down a list. But yes, this is a good list and on-brand for a technology publication. Everything becoming smart, connected, always on. Those have been promises of the tech for decades. AI-infused applications versus just AI applications is where it will go. Embedded AI, native AI, ambient AI, and perhaps other phrases all point in that direction, where hardware and software becomes smarter, and in that sense, AI is just there invisible. We also have NVIDIA's GTC conference coming up next month, and we'll see what big things

come out of that. But AI progress has been incremental since LLMs came on the scene, and the last big step function before that was deep neural networks. Agentic AI is an incremental next step, and reasoning and planning will get better, but what we need are two or three really major advances, like LLM. whoever comes up at that might keep it quiet, but it will inevitably show up in how well the AI works. All right, that's it for this episode.

Thank you all for being with us. HPC Newsbytes is a production of OrionX in association with InsideHPC. Shaheen Khan and Doug Black host the show. Every episode is featured on Insidehpc.com and posted on Orionx.net. Thank you for listening.