@HPC Podcast Archives - OrionX.net - HPC News Bytes – 20260525

Episode Date: May 25, 2026

- CHIPS Act awards for Quantum Tech - IBM foundry for superconducting quantum - Global Foundries factory for various quantum modalities - NSF X-Labs, other initiatives - Emergence of what might be called a National Discovery Infrastructure - Cerebras IPO - IPO Mania: SpaceX, OpenAI, Anthropic - ~$650B: Just 4 hyperscalaers's announced AI investments for 2026 - Intel + Tenstorrent acquisition rumors and implications [audio mp3="https://orionx.net/wp-content/uploads/2026/05/HPCNB_20260525.mp3"][/audio] The post HPC News Bytes – 20260525 appeared first on OrionX.net.

Welcome to HPC Newsbytes, a weekly show about important news in the world of supercomputing, AI, quantum computing, and other advanced technologies. Hi, everyone. Welcome to HBC Newsbytes. I'm Doug Black, and with me, of course, is Shaheen Khan. The U.S. stepped up its support for the quantum industry. The Department of Commerce last week signed nine letters of intent, totaling more than $2 billion in Chips Act incentives. to support quantum foundries and quantum computing companies developing utility scale fault-tolerant quantum systems. These funds will support a portfolio of quantum companies, including two domestic foundry companies, global foundries and IBM, and seven quantum computing companies across a range of modalities,

including atom computing, DRAC, D-Wave, inflection, Psy Quantum, Continuum. Minium and Rigetti. Each will receive $100 million in incentives except Iraq, which will receive 38 million. The biggest planned award, however, is $1 billion for IBM to establish a new quantum foundry for superconducting waferes. Global foundries will receive $375 million to establish a secure domestic quantum foundry for multiple modalities, including superconducting, trapped ion, photonic, topological, and silicon spin. Other countries are not sitting idle either and pouring money into advanced technologies and fast-tracking defense-focused procurements. France, for example, put an additional one billion euros in its National Quantum Initiative called Plan Quantique,

which helps its local vendors like Allison Bob, Pascal, Quantella, Cubit, Soft, and others. Along these lines, the National Science Foundation in the U.S. announced a one-and-a-half billion initiative called X Labs. It's focused on what they called milestone-driven research programs, initially including AI-enabled scientific imaging, quantum sensing, quantum interconnects, integrated photonics, and advanced scientific instrumentation. Brookhaven National Laboratory provides a very concrete example. Its future electron-ion Collider is being designed with AI and machine learning embedded across their systems and workflow, from accelerator operations and detector systems to data capture and analysis.

X-Labs joins the Trillion Parameter Consortium, the Genesis Mission, and other initiatives like the National AI Research Resource, N-A-I-R-R, DOE American Science Cloud, M-S-C, National Quantum Initiative, and the RANANF. related quantum super computing integration, and national research platform NRP, which is based at UC San Diego, San Diego Supercomputer Center. They all have a common theme and cover different parts of a large push by the U.S. government to make AI, HBC, quantum, and scientific instruments into what may be called a national discovery infrastructure. These initiatives cover different parts of this national discovery infrastructure. Research organization, various test beds, lab integration,

science-focused AI models, access, data distribution, etc. A.I. Chipmaker Cerebra Systems debuted its IPO earlier this month on the NASDAQ exchange with a market capitalization of $95 billion on its first day of trading after initially pricing the offering at a $40 billion, fully deluxe diluted valuation. The Cerebris stock chart in the 10 days since the IPO has been fascinating. After topping out at $386 on debut day, the stock has receded steadily to its current price of $256. But you can't claim Cerebus. They initially priced the stock at 185. The IPO generated tremendous excitement catching the AI chip wave. But cooler heads have since prevailed after analysts declared loudly that the stock price does not make sense when held up against anticipated revenues.

Shaheen, we've talked about Cerebrus often over the past four-plus years. They're an interesting company with an interesting dinner plate-sized wafer scale engine that always seemed to lead that group of AI chip companies that includes San Bonova and GROC, along with Tenth Storrent, if you want to include them. I first met the cerebris folks at the SCC conference in 2019 and was amazed by their chip, which is essentially an entire uncut silicon wafer functioning as a single processor. The advantage it has is eliminating the data and memory bottlenecks of traditional GPU clusters, which are cut, of course, into postage stamp sizes and must constantly fetch data from off-chip memory.

Cerebrose keeps all its memory on the one chip. Because the Cerebrose chip is so large, an entire AI model can live on a single device, reducing communications delays from data traveling between different chips. And programming is simpler because developers don't have to split AI models across hundreds or thousands of individual processors. For a long time, the company seemed something of a novelty and confined to workloads of the science experiment type.

But Shaheen, things have definitely changed for cerebrus. Well, yes, the tech buildout is massive and relentless. Technology megatrends like AI, telecommunications, with 5G going to 6G, quantum computing, quantum technology in general, biotechnology, space tech, fusion, and on and on, are exciting and full of promise. They are also highly relevant to national security and national security and national. competitiveness in a rapidly changing technopolitical and geopolitical environment. All of that creates the gold rush backdrop we see today.

It is against that backdrop that Cerebrus went public, and then, of course, SpaceX, OpenAI, and Anthropic are preparing to do the same. The latter three are expected to add up to some $4 trillion in valuation, requiring something in the vicinity of a trillion dollars in invested money. Cerebrus's valuation went over $100 billion for a bit. A lot of that is because of their massive $24.6 billion backlog coming from mostly OpenAI. Their other big customer is G42, which used to have a small ownership too.

So, heavily concentrated business and not so high margin right now. Their margins on hardware and cloud are reportedly in the 40s, plus or minus, or even 30s, while notable peers are in the 60s or more. But advocates see them growing into that valuation. If it goes as planned, they would deliver the backlog in the next few years, add more backlog, move from the current 5-nanometer fab to 3-nanometer fab at TSM, and scale their cloud offerings, both of which would improve margin. Note also that not having Kuda was obviously not an obstacle.

They worked with OpenAI and Gimlet Labs to point. and optimize OpenAI's model to the cerebrous hardware, taking advantage of the massive on-wayfar SRAM memory for static random access memory, and building cloud APIs so other developers don't need to get under the hood. Between just Amazon, Google, Microsoft, and meta, spending on infrastructure was ballpark $350 billion in 2025. Their announced spending in 2006 is ballpark $650 billion.

Add others, and you can see a number nicely over $700 billion in 2026, whether or not all that will be spent or is actually able to be spent is another matter. But it sure exposes every element of the supply chain, down to how many factories exist for some item and how long it takes to ramp it up, etc. It also calls into question where all that money is going to come from. what gets sold to generate it and what impact this might have on the market. Chip industry chess pieces keep moving. The latest piece of potential news is that Intel may be interested in acquiring AI chip company 10 Storren.

This is interesting for a number of reasons. For one, Ten Storan is a highly regarded AI chip startup whose standing was significantly boosted five plus years ago when CEO Jim Keller joined the company. It's interesting, too, that Keller is a form. Intel executive, he was senior vice president of the Silicon Engineering Group until mid-2020. This acquisition, if it comes together, reflects the marked change of circumstances for Intel under CEO Lipbut Tan. On the first day of trading in January 2025, Intel stock was at $20. It's gone up 6x since, giving Intel many more options on the acquisition front. It was only last year that

we heard rumors Intel would split its chip design and chip manufacturing businesses with the latter, possibly acquired by Broadcom or TSM. We have not heard rumors of that type for quite some time, Sheen. We should note, too, that earlier this year, Intel announced strategic partnership with AI chip company Sanbanova to build heterogeneous AI inference systems. Intel is obviously intent to get into the AI compute arena in a bigger way. It's definitely a significant. It's definitely a significant event and a good move for both companies if it pans out. As you said, such a deal would bring Jim Keller back to Intel. Keller is a chip architect with quite a stellar record.

He's credited with reinvigorating AMD twice by creating the K-8 architecture way back and then later the Zen architecture, designing Apple's first custom iPhone processors, and then Tesla's self-driving chips, and then propelling Tenth Tentorrent, with a focus on cost-effective AI chips and systems. He's quite the living legend in technology,

one of the very few engineers whose designs have transformed multiple multi-billion-dollar industries. Tense Tornent is a prominent Risk Five company, such a deal with catapult Risk Five to the front lines of the chip industry. It also anchors the combination of Risk Five and AI squarely in the US. At SC-25, Tense Torrent had chips and systems on display. but also it has a focus on automotive chips, which is a big area by itself and has been a big area for Intel over its history. Intel spent about $15 billion buying MobileI, which reestablished its leadership in the automotive industry and continues to own some 77% of it, having taken it public a few years ago and then sold off shares occasionally.

Ten Storrent is a private company, so assessments of its size and momentum are not official. All told, the company has raised $1.18 billion in funding so far. And last we heard, it was pursuing another $800 million raise with a pre-money valuation of $3.2 billion. So that gives you a flavor of the valuation now and how that might inform the size of a deal we might expect. All right, that's it for this episode. Thank you all for being with us. HPC Newsbytes is a production of Orion X, Shaheen Kahn and Doug Black, host the show. Every episode is posted on Orionx.net. If you like the show, please rate and review it.

Thank you for listening.