@HPC Podcast Archives - OrionX.net - HPC News Bytes – 20250811

Episode Date: August 11, 2025

- TSMC employees caught in 2nm espionage charges - Trump calls for Intel CEO to resign - Curtain fall for Tesla's Dojo supercomputer - Paper review: Generative AI as a Geopolitical Factor in Industry 5.0 - David Patterson on federal cuts for research - Jack Dongarra on supercomputing, AI, quantum computing, and geopolitics [audio mp3="https://orionx.net/wp-content/uploads/2025/08/HPCNB_20250811.mp3"][/audio] The post HPC News Bytes – 20250811 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 HPC Newsbytes. I'm Doug Black of InsideHPC. And with me is Shaheen Khan of OrionX.net. Listeners of this podcast know that Shaheen and I are not scandalmongerers, but it's kind of unavoidable this week, as two stories from the world of HPC AI have scandalous overtones. The first involves TSMC in a national security investigation on a trade secret espionage case. According to local media reports in Taiwan, nine TSMC engineers were involved in a plan to photograph confidential process diagrams. Hundreds of screenshots were captured at a

TSM Foundry, where the advanced two nanometer production lines are located, according to the coverage. The other case involves new Intel CEO Lipbu Tan, whom President Trump has called on to resign after saying he is, quote, highly conflicted, unquote, due to business ties between the previous companies where Tan worked and with China. Shaheen, let's go through these one by one. Yes. On the TSM story, the suspects were reportedly arrested at a coffee shop near a TSM Fab in Taiwan, as one of them took photos with his smartphone of confidential information from the screen of a company laptop. They allegedly planned to share TSM trade secrets with Japanese chip company Rapidis, which we have covered

here, and which was founded by a who's who list of major Japanese vendors, and said it was on track to ship in 2026 and for volume production in 27. It would seem far-fetched that they would do such a thing. One of the accused is a former TSM employee who reportedly organized and planned a lot. Another appear to have suddenly become rich. Between carrying out their plot in public view and showing off sudden wealth, we'd have to say that the accused are not very good at covering their tracks. The real threat in theft of intellectual property remains what nation states do, however. Various studies and government reports indicate that while competitors and insiders cause their share of damage. State-sponsored actors are the most significant source of IP theft, and that is measured

in trillions of dollars. As for Lip Bhutan at Intel, he was CEO of Chip Design Software Giant Cadence from 2009 to 2021. Starting in 2015, the U.S. banned the sale of American components to China's National University of Defense Technology. But Cadence workers sold sensitive semiconductor design tools, software, and other technology to the university from 2015 to 2021, according to the Justice Department. Last month, Cadence agreed to pay $140 million to resolve the charges. Tan also is the founder of investment firm Walden International and began investing in Chinese electronics and manufacturing companies in the early 1990s. Walden also was an early investor in semiconductor manufacturing International Corp or SMIC, China's partially state-owned chip manufacturer.

In fact, TAM served on its board until 2018. One would expect Intel's board would have gone above and beyond to vet candidates for such a nationally and globally significant appointment. Intel, for its part, issued a statement and reiterated the deep commitment of the company, its directors, and CEO to advancing U.S. national and economic security interests, and reminded everyone how much they are investing in the U.S. I'll just read a part of it, quote, we are continuing to invest billions of dollars in domestic semiconductor R&D and manufacturing,

including our new fab in Arizona that will run the most advanced manufacturing process technology in the country and are the only company investing in leading logic process node developments in the U.S., end quote. It's been a wild year for Elon Musk, who's been in the roiling headwaters of politics and business since the start of the new U.S. presidential administration in January. As he has left the political arena to focus on his companies, now we have a Bloomberg story reporting that one of his biggest development projects,

the Dojo's supercomputer, looks, quote, dead, unquote, as executives involved in the effort have left Tesla for a competing startup. Dojo was an in-house AI chip development project at Tesla intending to create supercomputers to train its autonomous driving AI models. Tesla hired top chip architects and tried to build AI chips and systems and clusters, not just for the car, but for servers in the back end, saying it could optimize it better for what they needed

and in the process would reduce reliance on companies like Nvidia. Among those who joined but have since left was Jim Keller, now at Tense Torrent. He left seven years ago, and his successor left Tesla in 2023. Bloomberg said the Dojo team has lost about 20 workers recently and remaining Dojo workers are being reassigned to other projects within Tesla. It's true that building your own chip is a lot more feasible and affordable than it was in decades past,

and we should expect big cloud providers will carry on with internal chip development. So maybe Tesla's sister company, XAI, could pick this up. But regardless, the story shows that it is not so easy to go too far on vertical integration. Shaheen found a story on the archive website entitled, quote, generative AI as a geopolitical factor in industry 5.0, sovereignty, access, and control. According to the paper's abstract, industry 5.0 marks a new phase in industrial evolution, emphasizing human centricity, sustainability, and resilience through the integration of advanced technologies. Within this evolving landscape-generative AI and autonomous systems are not only transforming industrial processes, but also emerging as pivotal geopolitical instruments.

The impact of AI on geopolitics is very complex, and generally we should welcome academic studies that analyze it. The paper considers industrial competitiveness, supply chain, military and economic warfare, data protection regimes, and sovereign AI. think that like many others, the paper is too preoccupied with, quote, industry and the manufacturing economy versus an emerging information economy. For example, as you said, it uses the designation industry 5.0, which skips over how information revolution is impacting everything well beyond just industry. But Gen AI's transformative impact on geopolitics is valid, and there the paper leads a very good discussion. We all know AI can bring many benefits, but can also cause problems. The paper highlights issues at the global level. It says, for example, quote,

a fragmented global AI landscape driven by divergent values, technological blocks, and regulatory asymmetries, threatens both interoperability and shared progress, end quote. To avoid the problems, especially at a global level, the paper makes policy recommendations for governments, industry, and international bodies. There were news articles with perspectives from two of the leading lights of our industry. One is an opinion piece in the publication The Hill by David Patterson, whose labs at UC Berkeley have made huge contributions to computer science and the IT industry. Concepts like Risk CPUs, raid storage, and the Risk Five chips, and much else have come out of that

lab. David's piece is on the dangers of cutting federal government research funding. He talks about the ROI on taxpayer dollars that supported his labs and how that investment, roughly $100 million, has had big economic impact on the order of a trillion dollars or more, and that the work of those companies led to new jobs and massive tax receipts for the government. Beyond that, such work across many disciplines beyond computer science has been crucial in keeping the U.S. at the forefront of economic strength and resilience. The other news article is an interview with Jack Dungara at a conference in Europe that first appeared in the Italian edition of Wired Magazine and then in the U.S. edition. It discusses how high-performance computing

is no longer confined to labs and is now vital for training AI models. It covers AI's role in accelerating scientific discovery and simplifying programming and also touches on quantum computing, future systems, and geopolitics. Jack sounds cautious about quantum computing, seeing it as nascent and even overhyped, while seeing the future as including various accelerators and specialized capabilities, including CPUs, GPUs, quantum, neuromorphic, optical computing, and others.

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.