@HPC Podcast Archives - OrionX.net - HPC News Bytes – 20260126

Episode Date: January 26, 2026

- Intel earnings, outlook - Industry-wide supply shortage - Shrinking windows of vendor price quotes - Micron fuels upstate New York tech hub - Chip investments in US regions [audio mp3="https://orionx.net/wp-content/uploads/2026/01/HPCNB_20260126.mp3"][/audio] The post HPC News Bytes – 20260126 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 as Sheen Khan of OrionX.net. Intel share prices at one point last week jumped 11 percent ahead of its Q4-2020 earnings report with results that generally beat analyst expectations for the quarter, but its guidance for the next quarter was below and list expectations. In its fourth quarter, Intel reported revenues of $13.7 billion, which was above the expectations of $13.4 billion, and it reported earnings per share of $0.15, nearly double the $0.8 that was expected.

But for its next quarter, it projects revenues in the range of $11.7 to $12.7 billion. The midpoint of that range is $12.2 billion, which is below the average. average the so-called consensus expectation, which was 12.6. And that was enough for the stock to drop by 17% the next day. Intel stock finished the week, just over $45 a big drop from the previous day when anticipation was high, but it is nevertheless a significant increase from where it has been in recent months when Intel share prices were mired in the $20 range for the better part of last year. Intel stock price reflects many factors. Most importantly, the insatiable demand for processing for AI.

AMD's stock has more than doubled over the past year, and Nvidia, who shares dip below $100 in April, is now just under $190.10. So Intel's stock price rise is part of a general trend. I guess we have to add here that our coverage is not investment advice. As you said, the primary reason for the decline was the forecast for the first quarter of 2026, which was a surprise given how well Q4 of 2025 had gone. So why would they set lower expectations?

Reasons given and discussed in the media include manufacturing yield on the advanced 18A node, which we have discussed here, which is in line with industry norms, but lower yield means fewer chips to sell. Next reason was lack of customers for Intel's foundry business. This is despite reports in recent months of Apple. and Microsoft signing up. So it looks like investors really wanted a big announcement here. Such an announcement might be expected later in 2026, they said. It is also true that the stock had gained quite a bit as a percentage,

so expectations were high about the manufacturing business, desktop servers, and AI. But probably the most important reason, and a good segue to our next story, was supply shortage. Intel said they see demand that they cannot meet, and especially for AI data center server chips. Presumably, this is not just related to the 18A Fab, but also to older production lines, where capacity was previously cut and product inventory has depleted. Yes, the computing industry is experiencing a severe shortage of high-performance components,

specifically DRAM and NAND flash memory driven by the AI frenzy. This is a significantly disruptive development. It is considered a structural scarcity, which means it's expected to be here for many months, probably into 2027, and it is creating rapid double-digit price increases and severe supply chain volatility. Memory prices were up about 50% last quarter, and analysts expect another 50% this quarter. We are seeing marketing campaigns by vendors to help customers optimize the memory and SSDs that they, already have, highlighting hybrid configurations without loss of performance, or otherwise sway customers away from large memory or large SSD configurations. Many vendors are finding that they cannot provide long-term valid price quotes and have resorted to, quote, allocation only ordering,

unquote, where suppliers stick to contracted volumes and they only guarantee pricing and supply for short windows as little as 48 hours, because their suppliers. don't know what the price will be. This is a seriously disruptive situation and recalibration for many customers and vendors alike. That is very true. On one hand, customers' business processes often depend on spending at least a few weeks, if not a few months, evaluating price quotes before placing a purchase order. On the other hand, major vendors have increasingly moved towards selling complete systems. So customers cannot build their own inventory of members. or SSD and buy motherboards and racks later.

The dominoes start with the exploding demand for AI data centers, which consume some 70% of high-end memory chips in the form of high bandwidth memory, HBM, which is used for GPUs, and high-grade Nant flash memory, which is used for enterprise class solid-state disks, SSDs. The next domino is shortage of HBMs because they need advanced packaging and assembly, Really, TSM's chip on Wayfair on substrate, so-called COWAS, and remember the C, WNS in there. And that capacity is booked until 2027 and beyond. And new plants are not coming online until that time frame, the 2027, 2028 timeframe.

The C in COWAS includes the GPU and the HBM. Now, the HBM itself is a 3D stacked structure of memory dyes. These are placed next to each other on the W in Kowas, which is called the interposer, an intermediary substrate between logic memory dyes and the main package substrate, which would be the S in Kovas. Major manufacturers of both of those chips, HBMs and Nan Flash, are SK Hynix, Micron, and Samsung. For Nan Flash, we can also add the longstanding manufacturing joint venture between Kioxia, formerly Toshiba and Western Digital in Japan and YMTC in China.

So vendors naturally want to prioritize the AI data center business for all the expected reasons. That is where the new demand is. The demand is growing. It's expected to continue to grow. And it's a high revenue, high profit segment. But they hit a zero-sum situation because there is significant overlap in the supply chains of HBMs and NAND flash chips. Since they both use the same memory wayfair capacity, and only a few vendors make them. So prioritizing HBM production for GPUs reduces the available capacity for traditional memory products like NANN flash for SSDs and conventional DRAM. But the price of all of them goes up, and so does the margin with it. Related to these problems, but not delivering anything like short-term relief

is news from upstate New York that Micron has broken ground on a $100 billion project to build a, quote, megafab in Onondaga County, specifically the city of Clay on the outskirts of Syracuse. Micron said the project is the largest private investment ever in state history and will be the most advanced memory fab in the world. The megafab will be comprised of up to four fabs and is expected to be the largest semiconductor facility in the U.S. Shaheen, for those of us who live in the region, it's all very exciting. The micro-news follows by a few years Global Foundry's expansion plans and relocation of its headquarters to Saratoga County just north of Albany in New York. There is also Corning Glass out here that is a world leader, and IBM has long been active in the state, having had its beginnings in Endicott and Poughkeepsie

and its famous research center in Yorktown Heights. Between those companies and research, University such as RPI, Sunni, Albany, Syracuse University, and your alma mater, Cornell, supplying young talent. Upstate New York is becoming quite the tech hub, not to mention the many wineries around Finger Lakes and ski resorts and such. And there's also horse racing. So yes, you all have to come visit. Indeed. Micron started site preparation in 2023, got various clearances last year and officially broke around a couple of weeks ago. They expect the first manufacturing facility to begin operations in 2030, and the second one by

233, and all four running by 2041 to 2045. Micron is planning for 100% water reuse and 100% renewable energy for their campus. These projects will generate lots of ongoing construction and technology jobs and investments by other supply chain vendors. But the bigger news, of course, is the national effort in the U.S. to build out tech manufacturing capacity. And the Micron Megafab is another beneficiary of the Chips Act passed in 2022. Looking at other regions in the U.S., besides New York, where Micron's economic and manufacturing presence will more than rival IBMs at its peak, there's obviously Arizona, which has the largest number of announced projects with over $60 billion of investment,

There's another $60 billion in Texas with major facilities from Samsung and Texas instruments. Oregon houses several Intel plants, so there's about $36 billion in Oregon by Intel, and smaller amounts on the order of $150 million by microchip and HP. There's also the $2 billion Intel plant in Ohio, part of what was called Silicon Heartland to build two leading-edge factories with very aggressive timelines to go into production right around now, but it has since slowed down and is now projecting to go online in the 2030-2031 time frame. Idaho is where Micron is headquartered and it's investing $15 billion for a new R&D and manufacturing facility there. New Mexico is where Intel is investing

$4 billion for advanced 3D packaging. Utah is emerging as a hot place with Texas instruments expanding its capacity in a multi-state $60 billion plan. And there are several smaller but still significant projects in Colorado, Minnesota, New Hampshire, Virginia, and North Carolina. In addition, you may recall that the U.S. government established eight regional innovation hubs or microelectronics commons across the country, which includes Massachusetts, Indiana, North Carolina, and California to focus on prototyping and workforce training for defense applications.

Of course, California remains a global powerhouse in all manner of technology, including chips and R&D. 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.