https://www.technologyreview.com/2026/06/25/1139696/ibm-unveils-sub1nm-chip/
IBM has built a new prototype chip with around 100 billion transistors on an area the size of a fingernail, which is twice the density of the company’s previous state-of-the-art technology announced in 2021. The design could pave the way for faster and more energy efficient computers for years to come.
From Intel message board:
Sticking to the time-honored tradition of touting physical dimensions that bear no relation to anything actually inside the chip, the name once again scales better than the technology does.
What they're really shipping is a "nanostack architecture" built on roughly 5nm features, which they claim rivals what a genuine sub-1nm chip would deliver.
Still a genuinely impressive piece of engineering, but it seems the industry is carrying a few too many marketers on the payroll.
@qp this seems to be the truth and makes sense.
@qp+1kvz9qbzw
I actually think this is for saving his job.
The Alderon split is basically a layoff to take headcount off IBM books after over-hiring for Rapidus. Most of if not all other business deals also never happened. Even the Rapidus partnership does not look like it is going very well, and it seems they want out.
Albany is in very bad shape. But if you can keep making good headlines, maybe you can buy some more time. Classic management jugaad.
@qp I think you are right...
@n8
This feels like Huiming’s heavy-handed attempt to cement his legacy. The difference is that past semiconductor legends earned their reputations through years of hard work, genuine innovation, and commercial success. Instead, he seems determined to force nanosheets and “nanostacks” into every conversation, whether they deserve the spotlight or not. It’s like trying to force and punch a puzzle piece into a spot where it simply doesn’t belong.
The irony is that he’s chasing a model of legacy that itself has become antiquated. The era of singular, larger-than-life semiconductor icons—the Moores, Dennards, and Hus—is largely over. Today’s breakthroughs are driven by massive, collaborative ecosystems spanning companies, universities, suppliers, and research consortia, not by one individual trying to become the face of a technology.
@kn it sounds just like pouring good money down the drain - why not put all that money into a big trash barrel instead and set it on fire in the Armonk campus for all to see ? Because it is obvious that nothing is coming out of this effort and it's a real paradox that Arvind can say that IBM should no longer be chasing research patents and publications if they don't generate revenue and yet, this is exactly the thing he is doing here. This is total bullsh-t from Arvind as CEO. If there is no headway for this project in the industry, stop the project immediately and dump it. Let's face it - IBM is a flop compared to other companies and it has become like this thanks to Arvind and his Pipmunks and their rubbish culture. How many billions of dollars have been sunk into this cr-p ?
@fd
Agreed. This feels like a last-ditch effort to breathe new life into Albany’s floundering nanosheet work that Huiming has championed for years. It is his passion project.
The problem is that the industry either wasn’t interested in IBM’s implementation or developed its own versions. As a result, IBM generated essentially no meaningful licensing revenue from the technology. While IBM made genuine scientific contributions, the basic nanosheet concept wasn’t created or invented there. It was already being explored and advanced by multiple companies and research groups. IBM largely built upon an existing industry direction rather than creating it. They copied it.
From a research perspective, it’s an impressive accomplishment. From a business perspective, it’s hard to argue it was a success. After years of investment in time, resources, and manpower, where is the return on investment?
Even Huiming has acknowledged in his posts for years now has mentioned one of the most common questions he receives is when the industry will adopt IBM’s technology. The reality is that no one appears poised to adopt IBM’s specific implementation. Or ever will even entertain it.
This latest announcement feels like doubling down on the same strategy. IBM itself is saying commercial high-volume manufacturing could be five years away. In the semiconductor industry, five years is an eternity. By then, newer, better, and—most importantly—more cost-effective solutions will likely exist.
Nanosheet and stacked nanosheet technologies absolutely have a future. But if they become commercially successful, the revenue is far more likely to come from companies like Intel, Samsung, or TSMC with their own manufacturing-ready implementations just not IBM’s. So it’s almost a waste of time and money.
That’s what makes IBM’s strategy so puzzling. The company continues to invest heavily in technologies that generate impressive headlines and publications but have struggled to translate into meaningful licensing revenue or commercial return.
@fm There was very, very small improvement, if any, from P9 to P10 to entice serious customers to update. Customers saw the writing on the wall. IBM stalled, is becoming more and more a synonym of mediocrity. Customers effort focuses more on evading IBM trap of a proprietary, very expensive products. These products can easily be replaced by better ones.
@Fg. Did you read the post??? It clearly says IBM management owns this problem child.
“The net of all of this is Power Management blew it by trusting GF when it was obvious they were never going to perform.”
@em Excuses .... Excuses ... Blame someone else for IBM mediocrity and incompetence.
This is bullsh-t. They’re essentially working on the same type of research that many others are pursuing and have produced some impressive cross-sectional images. But by the conventional definition of process nodes, this isn’t truly a sub-1 nm technology. They’re playing fast and loose with the terminology and marketing, which ends up misleading people about what has actually been achieved. It’s more of a branding exercise than a genuine process-node milestone.
@F2. The lawsuit that I believe you are referencing was settled in 2025. The terms were confidential. Net net Power has gone from an 8 billion per year business to a 1.5- 2.0 billion business today. (Where did that power business go? I speculate it went to Linux on Intel due to Intels continued improvement vs Power). Taking an 8 billion business and managing it down to 1.5-2.0 billion over 10-15 years is pretty much the definition of failure. So who takes the rap for shrinking Power business by 75%?? I believe IBM Power management has to take that title for their poor business decisions.
@em well that’s not 100% true. There is a very public lawsuit of GF suing IBM. Google and Read it. May cause you to modify your story a bit.
@Ca. The backstory of Samsung making P10 is that GF cost IBM 2 years of development effort as P9 had to remain in the marketplace longer than planned due to GF withdrawing from advanced (sub 14 NM technology). That delay allowed Intel to catch up to Power technology advantages and thus rang the death knell for Power. (System I and small system P installs were replaced with advanced Intel technology) P11 as a result was late to the game and the marketplace acceptance was received via a “meh” response. P12 as a result is for the most part stalled, and there is a good chance will never see volume production. The net of all of this is Power Management blew it by trusting GF when it was obvious they were never going to perform. By the time management recovered from the GF mistake, the marketplace had passed them by. The end result of all of this is to say IBM has to make money on licensing their IP, as their ability to manufacture advanced technology solutions has passed due to the ISV’s adopting advanced Intel Linux solutions.
@ap Power10 is made by Samsung from IBM design.
Wow! This is like the third or fourth huge story IBM in less then a month! Winning! But does anyone else get a fishy vibe. Seems too compact. Why are we going all hifi in like two weeks time frame?
@ax it would be interesting to see the big song and dance routines from IBM and it's executives on the LinkedIn pages for this non-announcement.
Listen, we know IBM is good at microelectronics... the question these days is that are they going to capitalize on that? The answer is no because the Execs are just plain d-mb and only play the hype game.
The only thing IBM is good at is perfecting ways to reduce headcount without having to pay for it or get sued for age discrimination and other laws they're breaking.
This was put out today to give them one good highlight to discuss in big town hall tomorrow
In my long career at IBM, I witnessed multiple debacles related to Power. Power10 was the last chip manufactured by Global Foundries for IBM. IBM struggled to find a manufacturer for Power11 and Power12. Customers are abandoning Power platform, since, they can no longer trust the viability of IBM.
If IBM cannot manufacture products that are the lifeline of its core business, how can it claim any breakthrough advances.
@ak also will they license this out? And actually generate revenue from it? Or is it just another cool science experiment? Their Nanosheet has not been licensed by anyone. Samsung, Intel and TSMC developed their own on their own. It is basically a financial flop for the site. The site in general has a huge identity crisis. Are they R&D? Are they prototyping? Based of news just a week ago it sounded like they were HVM for quantum (impossible give size of line). So what gives?
There is industry skepticism because IBM has announced advanced nodes before (7nm, 5nm, 2nm, etc.), and demonstrating a few wafers in a research environment is very different from producing millions of chips at acceptable yield and cost.
@aa we aren’t
@ae Feels likd the Albany hype machine seems to be running at full throttle double overtime lately. Usually that much promotion means someone’s trying to distract from something.
@aa Duh ! sorry I missed the fine print...it's no obvious that this is yet another marketing announcement from IBM loser Jay Hambetta since these chips could have a higher failure rate and IBM is keeping mum about this. LOL.
Seems that a lot of folks at IBM are waiting for the government's $1 billion check from Trump.
@a8 :
Did you miss the fine print - A multi billion dollar investment is needed to build the refrigeration centers to host these energy efficient computers.
Next news you can expect is that IBM is already working on state of the art refrigeration technology to capture that market.
@OP will they burn up faster too ?
pure garbage