Closed Source Intellectual Property & the Chip Shortage
A lesson about IC Camouflaging, retirements, and legacy closed source IP
This is information which to my knowledge has not broken mainstream press nor has it reached Substack until now. One of the most significant factors that caused the legacy-node chip shortage was that Taiwan Semiconductor Manufacturing Co. (TSMC) decommissioned and discarded some or all of their lithography machine(s) capable of producing such chips around early 2020. And this was instrumental in causing the auto-chip and similar chip shortages which still plague industries that relied on those ~250nm transistor size chips.
To recap my last write-up, I summarized how black-box logic locked chip designs drive uncertainty in the commodity chip industry and also conjectured about their likely role in the ongoing chip shortage. Restating another point from that post, these black-box chips have also already killed somewhere around 90 people – that we know of – from the 2008 Toyota car crashes. Looking at the multibillion dollar losses from those deadly 2008 unintended microchip induced accelerations and the media circuses that surrounded those lawsuits, it is most likely that they have actually killed many more but the companies involved probably learned to settle alleged chip malfunctions way more early-and-often. Since then, the publicly available data on electronic car component recalls looks like this:
I just got done interviewing Luke Kenneth Casson Leighton namely because he is the founder of the Libre SoC open source computer chip project and his project is the first in a series of write-ups I am doing on potential future paradigms in computer security. His project alone could go a long way in ensuring that events like the deadly 2008 Toyota unintended accelerations never happen again but he is not only a talented leader of the fully open source computer chip project known as the Libre SoC; he is also someone who has spent a significant part of the last decade cultivating relationships with-in the fabrication sites of the island country of Taiwan and within the greater Shenzhen, China region1. So, it is fair to say Luke’s knowledge of the chip supply-chain not only runs deep but is also full of personal relationships with folks in-the-know of the real world computer chip supply-chains. That is to say Luke knows many of the people working in these supply-chains.
In the interview I recently did with Luke, he conveyed to me, rather conclusively, what is at the heart of the legacy node chip shortage which induced shortages of other products like automobiles2. This is information which to my knowledge has not broken mainstream press nor has it reach Substack until now.
Luke informed me that the reason at the heart of the legacy chip shortage was that TSMC decommissioned and discarded their legacy lithography machines; today’s most legacy nodes have a ~250nm transistor size. TSMC was one of the last — if not the last — foundry maintaining these legacy ASML lithography machines and in early 2020 TSMC decided to decommission and throw out these machines central to many of the chips used in everyday components for products like cars, medical devices and school buses because it was unprofitable for them to maintain it. Stated another way, this part of the multi-year global microchip shortage — which according to yesterdays WSJ is still ongoing — the whole shortage started because TSMC decided to throw away their older ASML lithography machines at an inopportune time for the rest of the chip and component industry.
In case you need a reminder, this caused used car prices to inflate close to 50% and new car dealer lots were sold out for months. It also had a very dangerous effect on chip supply-chains inducing an underground economy for used and black market chips. And many of these black market chips are now suspected to be in peoples brand-new automobiles.
We can only wonder if this brand-new and malfunctioning Mercedes-Benz G-Wagon got some of those underground market chips before it retailed for $158,000 in early January:
Why would TSMC trash their own lithography machine?
The reason TSMC decommissioned and tossed out their legacy node lithography machines was apparently 3-fold. (1) For months leading up to early 2020 TSMC had no orders for these legacy node chips, (2) the tolerances on these high-tech lithography machines run so, so tight that they have to continuously run 24/7 to keep them warm, wasting entire high-purity silicon wafers as they run them through the lithography machine; otherwise, the machine will begin to malfunction and (3) TSMC also wanted to make space for more advanced node lithography machines in their rather expensive and limited spaced cleanrooms3.
This of course could not be easily corrected by the market for another three reasons: (1) chip designs are not open source and thus only one company (e.g. Infineon) knows how to produce particular chip designs, (2) the only engineers who might have known how to update the legacy chip designs for more modern lithographic techniques long ago retired from their respective chip companies and (3) because commodity chip design companies have de facto monopoly power over producers who had to use these black-box chip designs as a foundation for their product’s circuitry, and so now we have the ongoing chip induced shortages across multiple industries.
What a mess!
This makes it increasingly clear much of the chip shortage could have been largely prevented if producers like car companies were able to use open source chip designs. To know exactly to what extent open source chip designs could have ameliorated the chip shortage would take a bit more investigating but after three years the picture is coming into focus and it’s pretty damning of our current closed source chip supply-chain systems. More clarity around these events would certainly be nice, however the current supply-chain systems induced by black-box chip designs — where only one company is allowed to know a given chip design — appear to be revealing that our current supply-chains are far too brittle because of this black-box nature of the modern chip industry. More robust open source supply-chain models appear sorely needed.
Given the billions of dollars being handed out by governments to solve this problem, we deserve more clarity around what exactly happened here but to review my last post on the chip shortage and my best attempt to summarize everything else that is known right now, review my previous post.
Over the next few weeks I will finish the summary write up I am doing for the Libre SoC. Then when I’m done editing the nearly six hour two part interview I did with Luke Kenneth Casson Leighton, I’ll begin to put up the material here on Substack. Once they are ready I intend to release the write up, then the two re-release interview parts over three weeks. There will hopefully be a nice place set up for you to donate to Luke and his project if you feel particularly inspired by his work, too.
The Libre SoC is the first of the four potential future paradigms in computer security for which I intend to do write-ups.
The four paradigms in total are:
Recent notable and related headlines:
April 8th, 2023 Wall Street Journel: Car Breakdowns Are Making More People Fall Behind on Their Loans
WSJ reporting subprime autoloans crises worsened by cars that don’t work
April 2nd, 2023 AutoNews: The latest numbers on the microchip shortage: North America hit hardest
Latest projections in automobile production continue to get slashed week-by-week
March 11th, 2023 Fortune: We’re going to see another chip shortage—despite the CHIPS and Science Act
Fortune Magazine reports: “any company that deals with semiconductors has to be ready for chip shortages to be the new norm. They can create a buffer through stockpiles or having multiple suppliers”
March 2023 EE Times: GM–GlobalFoundries Partnership Seen as Setting Trend
February 9th, 2023 Associated Press: GM reaches computer chip supply deal with GlobalFoundries
GM has Announced partnership with US chip supplier GlobalFoundries in upstate New York. With few other details appearing public right now, GM is “throwing out its old business model of letting parts supply companies acquire the chips with GM knowing little about them”
January 12th, 2023 Bloomberg: TSMC Chief Sees End to Chip Shortage That Hammered Auto Industry
TSMC has not yet proven correct about its predictions from January of a “further easing” in the industry wide auto-chip shortage
Here is a link to Luke’s old blogs when he was traveling that region. And in case you don’t know, the small island country of Taiwan is currently the greatest exporter in the world of the most advanced (sub-7nm transistor size) computer chips and Shenzhen is often compared as the Silicon Valley of China.
The legacy chip shortage is one of the three main broad classifications journalists in mainstream press have been using to describe the global chip shortage that has been raging since early 2020. Legacy chips are largely a colloquial term for chips that are not at the cutting edge which today is basically anything over ~100nm transistors.They are in contrast to the other chips that have been hard to keep in stock like electric vehicle chips that drive the powertrains of EVs and the advanced sub-7nm chips that go into phones, laptops, smart-watches, etc. These three broad classifications are actually very sloppy and not entirely mutually exclusive. The latter of those two — advanced sub-7nm chips — are now experiencing a chip glut and have been mostly overstocked since late 2022.
Lithography machines require better than medical grade clean rooms to ensure the best chip yields from their machines.
A great discussion continued when I linked this article to ycombinator.
https://news.ycombinator.com/item?id=35501602