The challenges Intel faces to compete with TSMC, Samsung Ever since CEO Pat Gelsinger announced Intel was opening its fabs to contract manufacturing, the question has been: for whom? Who was going to pay to have their chips made by an American giant with such a spotty reputation for execution and whose process tech has been not only falling behind TSMC and Samsung, but even Intel's …
I believe Intel's fab business is guaranteed to be successful since it's only a matter of time before China invades Taiwan and TSMC's fabs are destroyed. Companies like NVidia will have few other optons (Samsung is mostly booked ful too) to produce advanced semiconductors like the kind TSMC produces. For intel it's merely a waiting game.
If I were TSMC I'd set up a U.S. holding company as a matter of prudence to enable moving the company to the U.S. wholesale. Their entire IT operations should be hosted in the cloud with fail-over to another region if Taiwan is attacked.
The biggest problem for electronics production i the rest of the world is going to be passives (resistors, capacitors in all their different varieties, diodes, inductors, all the other components I can't remember) because 99.9% of the worldwide production of those is happening in China.
But many of the manufacturers are Western which can easily move their production to other Asian countries. So far they haven't because they're simply lazy and imprudent.
They're hoping things will blow over. But they won't.
The first large Western passive electronic component manufacturer to move their operations out of China is stand to win big if Taiwan's invaded.
"easily move their production"
Yeah, just pick up a production line and move it. What could go wrong.... Said no engineer ever.
You don't just move a production to somewhere else. It takes YEARS to move production locations for anything involving high volume production. It's not just the machines you have to move, it's also the people and/or their knowledge, your supply lines and logistics, your facilities (waste disposal/treatment becomes it's own consideration at this scale for instance). Then comes re-qualifying products from your new line against products from the old line. You're customers aren't going to like even the slightest shift in "mean" values, even if the overal value stays in the spec range. They're probably going to want to receive a few batches of the new stuff, along with continued delivery of the old stuff until they can qualify that the new stuff is up to snuff. And even then they might just go "bollocks to this, the other guy is 0.001 cent cheaper per unit and if I'm going to invest in qualifying a new part I might as well move over". And then you lose the customer entirely.
And also no, most of those manufacturers aren't Western. The factories and machines in them are very likely not the property of the companies selling the components in the west. They're all contracting production to various producers in China that actually own the means of production (and even if the Western countries did "own" them, China would require them to make the company actually owning the factory fully Chinese and would make it extremely hard to actually move the factory line).
TSMC is building a new fab in Arizona and I'm pretty sure has full contingency plans in case of China invading. I doubt China is ever going to be that bold though. It's going to hurt them (and the rest of the world) tremendously. Certainly there's no indication it's going to happen in the short term. Russia's extremely hard time in Ukraine probably has helped indicate the folly of such an endeavor too.
You're forgetting that their Great Leader has pinned himself down on reuniting the Motherland and Taiwan. Taiwan is easily defended by determined armed forces so logically speaking doing so could only result in a bloody nose. If the U.S. pitches in its military might China stands no chance of taking the island and will easily be dispatched.
But the fact is these things aren't decided by logic but by political incentive.
So China's only hope is bullying, bribing or threatening the U.S. not to intervene. Threatening economic upheaval might work, but it would have to be measured against the U.S. loss of influence in the world and especially Asia. If the U.S. were to leave Taiwan to its own devices it would de facto cede world dominance to China. I don't see that happening.
A recent interview with Shang-Yi Chiang, former Vice President of R&D at TSMC (also held positions at TI, HP, and SMIC) had insightful commentary on the speed of bringing up a new node.
"We all take two years to develop one generation, how come you guys can do it in one or one-and-a-half year?" And they asked if some of your customer transfer technology to you or what not? And I told him, "No," I told him that, "That's not true." I think he probably implied we steal technology from customer, the way he talk.
And I say, "I'll tell you why." I said that, "When we develop one node, basically you have some learning cycles. First, you do some simulation. And you have some idea, then you run wafers to prove that. So, you run a group of wafers according to simulation and you have some splits. The wafer runs through the fab, they come out and you measure them, you analyze them, and you try to improve and you run this again. This again, you run. So, this is learning cycle." At that time, "It takes about six learning cycle, roughly, to complete one generation." Of course, you had some short loops and not just one. I said that, "My R&D wafer in the fab run much faster than yours, because my R&D engineer works three shifts and you only work one shift. So, your R&D wafer move eight hours a day, my work/move 24-hours a day. So, my wafers go three times faster, even if you are twice smarter than me, I still beat you up." <laughter>
https://www.computerhistory.org/collections/catalog/102792671
There are different cultures. Some companies decide that the best course is to screw your customers for every last cent and deliver as little as you can get away with, as late as possible, pay your suppliers as little as possible, demand more every year and pay as late as possible - the only thing that matters is this quarters earnings. you need an MBA to understand why this is optimal
Then there are companies that have relationships with customers that last decades, where the customers and suppliers are partners in the project and if everyone works together then everyone does better out of the deal. This is basically communism and explains why Japanese, and European engineering companies do so badly.
It's a big gamble with public money, at some point the US government may realise that propping up Intel is just throwing good money after bad. If Intel can pull off the near-miracle of producing competitive nodes at a decent price, the strategic value of an indigenous subsidised processor source may just keep it going. If it can't then it's only a matter of time before Uncle Sam pulls the plug.
I don't see why it would be such a "near-miracle" for Intel to catch back up. To me it's always looked like Intel wrong-footed itself by betting the farm on EUV litho becoming available about 10 years before it was actually competitive with multi-patterning and thus didn't put enough resources into developing sub 7nm nodes through DUV litho systems. TSMC did go the multipatterning route and appears to have developed a process that was much easier to switch to EUV once those tools became available. Meanwhile Intel had to basically develop the 11, 9 and 7 nm nodes in a panic and seems to have gone a route that didn't allow for easy switching to EUV not that these tools are available. They've been working hard on retracing their steps and getting back to competitive processes for lower nodes through EUV. They should be able to get there, Intel certainly has the institutional knowledge to make it happen.
Intel’s problem is that it’s had a succession of CEOs who were money-men and bean-counters. Take the cash reserves and use them for stock buybacks and dividends… what could go wrong? It’s not as if in an industry where you have to build a humungously expensive fab plant every couple of years you’re ever going to have a need for large piles of cash you can draw on, right?
For the last few years Intel’s been proudly repeating Grove’s old mantra of “Only the Paranoid Survive”, but with the passing of years everyone forgot what it actually meant in real terms and started to think of it as just a catchy marketing slogan, to jazz up content-lite Powerpoint foils along with a graph of Moore’s Law.
Pat “gets” all of this, and with his background he knows full well what needs to be done. I worry however that he’s been placed in the invidious position of someone being given the helm of the Exxon Valdez (or if you prefer a more recent analogy, the Ever Given) and being told “All yours! The shoreline is a half-mile off the bow and we’re sailing at full throttle towards it, make sure not to hit it, good luck! We’re all counting on you!”
I write as an ex-Intel person who had a bloody brilliant couple of engineering decades with the company until a recent round of layoffs caught me (and my manager) by surprise…
Intel had a huge advantage by being vertically integrated.
The fabs could be late to the next generation because they had a guaranteed customer in Intel. The designers could rely on the next generation because Intel had fabs that would always be able to produce it.
As it takes hundreds of millions of dollars to tape out a complex chip, Intel is unlikely to find many customers until after its new FAB lines have proven themselves (if they ever do). This will add a delay of 6 months to a year (or more) from the 20A FABs coming online to the point of first commercial customers. Unfortunately for Intel - the tape outs are specific to a particular process - one for TSMC 2nm will not work on Intel 20A and vice versa. For a customer the choice is between investing over $500million on a TSMC tape out from a manufacturer with a good success record or over $500million on an Intel tape out from a manufacturer with a poor success record.
Intel will probably have to massively subsidise its first customers - or it will not get any.
TSMC is likely to suffer from it's own success. It might well be there's customers that would LIKE to spend 500 million USD for a TSMC tape-out but can't actually get that contract because TSMC simply doesn't have the capacity. If Intel can show it can do 20A, it might well be a choice of "wait 2 or more years for capacity to become available at TSMC and spend 600+ million USD, after inflation or spend 500+ million USD and get wafer starts within a few months"
I assume Intel will be it's own customer if there are no others willing to do risk production.
After all Intel will need to use 20A node for it's CPUs/GPUs as well eventually. If and when production is stable, there may be others willing to become intel's customers for that node.
Not sure how many other customers they are going to get.
No competitors in the CPU / GPU or other markets that Intel play in are going to want to share info, even if it's just order sizes and launch dates, with intel.
Or are you going to trust your fab slots if Intel decide they need more product capacity for themselves? (I'm betting there are lots of caveats in their contracts)
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
