As he says here, the decision not to take risk and jump to EUV with TSMC and Samsung was the mistake that started this downfall and it snowballed. The decision happened somewhere in the pathfinding stage into 10nm.
When Gerald Marcyk was running Intel components research back in the
2001 Intel was already considering using EUV lithography for the 45nm technology node that would come out 2007. It turned out that the switch to EUV could be delayed and delayed again. Intel thought they could get away with it this last time and it was a wrong decision.
Intel, Samsung and TSMC invested billions into ASML in 2012. In exchange Intel got 15% ownership, Samsung 10% and TSML 5% if I remember correctly.
They don't have any significant ownership anymore.
A lot of people were hoping someone would come up with a better way to make "extreme ultraviolet" (soft x-rays, really) light than the horrible kludge used now.[1] Syncrotrons were tried. (Whatever happened to Lyncean?[2]) What's used now looks like a physics experiment, not a production process. Drops of tin are hit with lasers to emit X-rays, which are then focused with mirrors on the mask and wafer. This takes a setup the size of a house that produces very little power at the output end.
Much of the industry was hoping something better would come along. But so far, it hasn't.
What's stopping them from using hard x-rays? Those are really easy to emit. (Genuine question; not trying to suggest that they're all idiots who just never considered using hard x-rays.)
Interaction with the mask material, I imagine. With hard x-rays, most of the photons generated will just sail right through, not interacting at all, or (worse) ionize the material being exposed.
Personally, I'm surprised that there aren't ways to generate this energy by nonlinear mixing of light from two lasers that are separated by the desired frequency.
Lasers are an Electro-Magnetic emission of which the output is a synchronized at a specific and regular frequency.
If you have two lasers at about 700nm and 1000nm, and they 'start' at a synchronized point, then 7mm (7000nm) of emission later they'd have a repeat of the cycle begin again.
During that cycle the interference pattern would probably have an interesting shape, but I very much doubt it'd be useful for semiconductor production. Maybe a math or physics graduate could elaborate on if there's a name for this kind of pattern and if there are any uses I'm not aware of.
That’s a linear mixing, non-linear allows for the absorption of two photons followed by the emission of one with twice the energy. That allows for a laser pumped frequency doubler.
Actually most consumer oriented lasers (for example green lasers) work like this, but I don’t know if it works for frequencies as high as soft x-rays.
An option would be to replace the laser and tin with some tape. [1] I know tape is a consumable item but is considerably less professional compared to a laser.
A laser is not just light. It is a specific wavelength. X-Rays are a spectrum. The EUV/soft X-ray is a part of that spectrum. There's also other properties that you want from the laser. But this tape-x-rays won't be easy to control or uniform in wavelength. Making it not a laser, even if you could collimate it.
Intel invested lots and lots in novel EUV alternatives and they didn't work out. They took risks, they spent big on R&D, and I think you can certainly blame mismanagement but it's wrong to imply Intel wasn't taking risks.
Intel first started cheaping out as far back as 2003-2005.
157i could've been almost as good as EUV, at much lower cost, but Intel fully consciously decided to forego it because "nobody can beat us on 193i yet," and of conern that 157i consortium will give superior litho to potential competitors.
As someone with no background knowledge, I first looked up "193i" intel which lead me to the term 193nm immersion lithography. After that doing the same for 157nm was trivial.
I would have been more wowed by the $120 million if I didn’t know B2 bombers cost $2 billion each. I wonder how much more safe and secure we would be with lots of Twinscans 3400B’s and a few less B2’s?
EUV is hard, but isn't part of the reason why only ASML can make those machines because of the CRADA they have with DOE as a member of the EUV LLC? I believe this agreement was why ASML wasn't allowed to sell their solution to China recently. I bet Intel will have no trouble buying EUV to bring in-house - especially if that serves US strategic interests.
They not only have no truble in buying it they likely have one or two systems.
But this systems are not easily produced neither fastely so between ordering them and getting them over a year might pass.
More importantly just having them doesn't allow you to actually use them properly. They are in a way quite a "raw" tool and the exact details of how you use them make major differences. And even if you know the best way to use them, that still doesn't mean you know the best way to create chips with them. Because chip layout (in it's details) has to be designed to fit the production process. Not doing so will lead to bad yield and or bad max. perf clock etc. Sometimes just improving how you layout some things while still using the EUV system the same way can make a major difference.
So it's not easy at all to switch.
I'm still positive they will long term manage to do so.
But it will take some time.
Which is why some parts which are not to much of their core business but can largely profit from TSCMs manufacturing process will be produced by them. And maybe as a bonus they will learn a trick or two from TSCM.
ASML only makes a few machines every year, you can't just buy them off the shelf
IMHO banning sales of ASML's product to China is chest thumping protectionist nonsense - if you're worried about the rise of China's military then you should be tangling their economy up more with that of the West so that they have more reasons not to go to war rather than driving them to be self sufficient.
The next war will be fought with AI. It's essential the West stays ahead in chip technology, one of the few advantages we still have. It could be the difference between winning or losing a war with China.
> tangling their economy up more with that of the West so that they have more reasons not to go to war
This strategy failed, China wasn't interested in becoming like the West. We're in a new cold war, it's time people wake up to this new reality.
I can’t decide whether to respond with either the (purported) Albert Einstein quote: “I do not know with what weapons World War III will be fought, but World War IV will be fought with sticks and stones.”
Or Robert E Lee: “it is good that war is terrible, otherwise men would grow fond of it.”
Your fondness seems obvious though. I think I will go with a current colloquialism: “Just stop.”
Nobody wound a cold war, everyone ends of worse than if they'd never had one ... "Interesting game, the only winning strategy is not to play at all" .... How about you just admit that China is different from the West, you don't have to make them be like you to be friends with them. And if you trade with them enough that they can't do without you then they have to listen to you when you talk about things like Hong Kong and the Uighers ...
What does 'becoming like the west' have to do with economic cooperation and peace? Are you saying the western model is the only possible valid and morally acceptable model for the entire world?
We are in a cold war because, like back in the day with the USSR and the old cold war, China doesn't want to turn into an Asian copy of the US. I'm no fan of PRC but I wish more countries would do the same. Traveling a lot today and you'll see American cultural imperialism everywhere. Besides, we do need another power to compete with the US so it can't do what ever the f** it wants to. Sadly the only real competition at the world scene today is PRC (and in a few areas, Russia). I for one would rather see more power on the hands of Russia and PRC if the only alternative were complete Americanization of the world.
Globalization is fine in some areas but when a generation can live off of pizza and burgers and yet have never tasted an Italian pizza or non-McDonalds style burger the world is a much poorer place. Everyone is turning into Americans. Asian Americans, Scandinavian Americans, etc.
> I for one would rather see more power on the hands of Russia and PRC if the only alternative were complete Americanization of the world.
America is far from perfect, so many things are broken it is insane, but to prefer to have a stronger Russia or China is bewildering to me. Everything that is bad about America is equally bad or worse in both of those places plus an undeniable loss of liberty. Cronyism, inequality, racism, imperialist tendencies, environmentalism, etc.
I'm not blind to the problems America has, but I've also lived in China and it boggles my mind that people think this way.
I don't agree at all that the things that are bad in the US are as bad or worse with those countries, first of all because it is a misunderstanding that it is "in the US" because the problem I was referring to with the US is outside the US.
The US is doing everything it can to control and dominate the world and we need a counterweight to stop this ASAP. It does much more harm than good. What happens inside US borders is another matter. The problem is when the US can do as it pleases which is basically what it is like today. I dislike Russia too but if Russia makes the US pause before invading a country or the PRC did the same it is a win in my book. The US is an extreme aggressor and someone need to force its fleets back where they belong and stop its political steamrolling.
Besides the problem of using its military to "make right" we are also seeing a death of culture because of Americanization. If half of huge US businesses were exchanged overnight to other strong nations businesses the world would become a better place. McDonald's and Disney doesn't belong outside the US as the dominant local business. Local burger and movies does. We live in a time where many children think American food and films are the norm, instead of something foreign, new and exotic as it should be. Every culture is becoming one culture. It smells a lot like something PRC does that the US is extremely critical about but is doing much more itself..
Of course I'd rather see the EU as a power that could put the US in its place but I'd also rather see Russia or PRC doing so if the alternative is no-one doing it.
ETA:
Saying imperialist tendencies are worse in Russia and PRC than the US is mindboogling. Russia and PRC does a lot of bad things but they are light-years behind the US in imperialism.
> ETA: Saying imperialist tendencies are worse in Russia and PRC than the US is mindboogling. Russia and PRC does a lot of bad things but they are light-years behind the US in imperialism.
Russia, in the last decade, annexed a region of another country. China is in the active process of annexing a region of the South China Sea for resources and using its newfound economic strength to project its power via the Belt and Road initiative.
I think you're focused on avoiding cultural hegemony to the actual detriment of the liberty of mankind. These are not countries that even pretend to have liberty as a core tenet - being an outspoken critic in both places is dangerous, Russia even flexed its power and killed people in the UK!
I'd rather eat McDonalds for the rest of my short, fat life than be at risk for being killed for being an outspoken critic of the government.
> Globalization is fine in some areas but when a generation can live off of pizza and burgers and yet have never tasted an Italian pizza or non-McDonalds style burger the world is a much poorer place. Everyone is turning into Americans. Asian Americans, Scandinavian Americans, etc.
For what it's worth, after chain restaurants become too dominant, the next level seems to be hipster burger joints, microbreweries and fusion cuisine restaurants.
Yes but the most influential ones like movies and music doesn't seem to go that route sadly. With Disney+ children everywhere will grow up even more American.
> tangling their economy up more with that of the West
That might have worked in the past but doesn't anymore as China actively counters this.
Selling them such machines will just bring short term production benefits to china and potentially gives them a chance to have it a bit easier to copy the technology.
But long term China will have it's own production methods and for it's military they will only use this sources. Currently they are at 28nm as far as I know. Which might be still quite a bit away from stat or the art but is already fairly useful for many use cases.
> if you're worried about the rise of China's military then you should be tangling their economy up more with that of the West so that they have more reasons not to go to war rather than driving them to be self sufficient.
Military power, like all other things, is based on economic might. If you want to slow the rise of China's military while it remains a totalitarian system then allowing further economic integration between the West and China is a silly idea compared to using sanctions to decouple two economies over time.
Though the rise of China's military is definitely not the reason why economic decoupling and sanctions should be deeply considered.
It would take years (even decades) to duplicate ASML machines. At least in the absence of unprecedented tech transfer (most likely through hacking and HumInt work). It's no where near a foregone conclusion that banning ASML exports won't slow down SMIC's attempts to build leading edge nodes that get used to supply China's military.
I think the key counterpoint to your argument is “like all other things.” That’s the thing you’re calling a silly idea: wealth, trade, immigration, etc. I think it’s unclear which side wins, but getting Goebbels out of the White House certainly seems like a step back towards the “silly idea.”
Prior to WWI, European economies were pretty deeply entangled and people even did not have to have passports to cross borders. Only Russia and Turkey required passports from travellers.
Austria-Hungary even produced steel and guns for their future enemy, the Royal Navy...
I don't think they're worried about China's military so much as China's independence from the rest of the semiconductor supply chain. When you give China a machine, it's only a matter of time before they reverse engineer it and make something slightly worse but much cheaper. This kind of theft is backed by the government. So the expectation that they will remain forever dependent upon ASML is wishful thinking.
ASML machines include a large amount of technology from the US, and so can be controlled under US export control laws.
A country is free to develop all those technologies locally and not face US export control, but that's such a difficult task that nobody has achieved it.
That doesn't matter. If they want to export TO the US they are subject to US export laws. The US has a history of their economical and political power forcing companies that have no business with the US to do their bidding to stay in business. Just look at the Iran sanctions.
> When you give China a machine, it's only a matter of time before they reverse engineer it and make something slightly worse but much cheaper
> This kind of theft
You called what they did “reverse engineering.” But then you next called it theft.
Why?
Reverse Engineering is not theft. It’s a follower route, where you build the same thing that someone else had built. You don’t even need their blueprints. You can just work off of their output, and work backwards.
One famous reverse engineering company was Compaq, which reverse engineered the IBM PC, with the nod from Microsoft to make their own PC compatible clone. Was this also theft? And was this also a crime in your worldview?
Another example is the atomic bomb. Once America proved that a nuclear fission bomb was possible, then it was only a matter of time before another country reproduced it.
Compaq did a clean room reverse engineering effort of the IBM BIOS for interoperability. Under US law this is completely legitimate.
China can and should do clean room reverse engineering for whatever technology they want to, and sell the resultant work back to the US (contingent on other IP restrictions of course)
And the US and Netherlands are well within their rights to block equipment equipment to China on grounds that it will be used by a unfriendly military.
The atomic bomb example is not just about "the principle has been proven", it was also a story of Soviet Union spies within the Manhattan Project.
> if you're worried about the rise of China's military then you should be tangling their economy up more with that of the West so that they have more reasons not to go to war
That's not really how things work. You become powerful first through economic hegemony, and then you can assert military might.
ASML EUV machines are the combined result of basically everybody in the business except maybe Canon and Nikon. Motorola, AMD, Micron, Infineon, IBM, Intel, Samsung, TSMC, lasers and plasma physics from LLNL.
I think National Security might have something to do with manufacturing of electronics moving outside the US. The war on crypto called for manufactures to either provide two versions of their product, or move manufacturing to a place they could export a single product: https://www.eff.org/pages/decrypting-puzzle-palace
So? Don't most manufacturers already have to provide multiple product versions if they sell internationally, to handle different electricity/plug differences around the world (among many other things)? Maybe there is an effect, but it isn't as if products sold worldwide come in one version that works everywhere.
Corporations love excuses to differentiate (read: charge more, especially for local captive customers) for products. Big Pharma uses all sorts of US government programs to run interference for them - patents, import/export restrictions, legal buying requirements, etc - all to charge more for the US version of products they sell dirt cheap elsewhere.
Whatever "National Security" effect was present causing manufacturing to move overseas was dominated by maximizing profits: wringing savings out of the supply chain and lower labor costs. The result was predictable.
I see Intel's complaint about this a way to drum up or steer business their way, hobble their competitors that rely on Asian chip manufacturing, seek future government intervention to help them - or accomplish all of the above.
Most of the electrical power supplies I have these days accept voltages anywhere from 100-250VAC. Its just a matter of having the right prongs on the end for the target market.
The electrical differences are isolated from the chip level manufacturers no processor that would have been affected by crypto export bans would be taking raw wall voltage.
IMHO (ok not very humble) any ware on crypto, weather it's in the past or modern approaches to it is stupid af and I want to give a metaphorical kick to anyone pushing it.
What many people which push it don't seem to realize is that such approaches undermine national and personal security and any benefits gained for it are likely outstripped by the long term drawbacks.
Drawbacks like making it easier for foreign actors to enact industry espionage or find ways to influence or blackmail people.
The idea of a government accessible backdoor which can't be accessed by foreign powers is just ridiculous. Especially if the access doesn't require physical access to the devices in question.
What is the point of "spend spend spend" without a strategy and top-down direction to it?
All we seem to care about in the US is "jobs". Generic jobs. Working class low wage jobs. Whatever jobs gets the public off a politician's back.
Not a lot of focus on whether those are useful jobs in long run, or strategically important jobs. Or we seem to end up after "spend spend spend" with some new financial derivative products, people employed to buy/sell them to each other (paid handsomely), and people who are left on the sidelines who didn't get to participate mostly.
At least other countries (I'm looking at Asia), less free and dictatorial they may be, decide what industries to make a strategic national priority and go after it.
Rather than aimlessly spending public money it might be better to encourage better corporation across the various companies.
I've been reading quite a bit of Dr Deming's thoughts recently and he had the view point that there should be corporation through out the supply chain, not just trying to squeeze every last penny. He felt it would be better for market and society as a whole.
Import substitution is usually a loser policy for developing nations. Also subsidies can frequently run afoul of trade agreements because they're basically the same as a tariff in terms of disadvantaging exporters.
National security is not really a valid concern but could maybe be used to justify divestment from China. But that just means diversification of import partners. We could be building fabs in Mexico like we do with the auto industry. Or, as he's actually suggesting, europe.
Yes, but some of the ROI of that intervention should be able to be captured by the US (as a representative of the people) in the form of equity or secured loans.
Insisting on stimulus of private industry without equity or interest is a great way to socialize losses and privatize gains, as they say.
I never understood the existence of bailout programs without equity. If the banks, car and semiconductor companies mess up and you keep them alive then you deserve to have a say in their business and owning their shares is the easiest way to achieve that control.
Occasionally I think about the tradeoff between municipal purchases that support the community (with expanding degrees of community) or which are the best for the task. Prior to EVs, for example, should parking enforcement be performed in a Prius to maximize long term TCO or should it use a GM/Ford/Dodge and keep money in the US.
At least in the US we've found that major global competitors control relatively critical roles in production of goods in the medical space. Who is responsible for quantifying the risk of being cut off from components?
I'm stoked. I've been a bit of an Intel hater lately but more competition is good for everyone. Though I'm guessing other people are excited about these directions as well because their stock is pretty high at the moment.
There have traditionally been cycles and we all win.
I remember years back when AMD became competitive with intel and then intel turned around and came out with fast cpus, and we had all kinds of choices.
I think the PC is a really good thing we don't want to lose it. When we have competition it is better for customers with more choices, and employees who actually have to compete, and there are standard interfaces new companies can develop for.
Honestly, what I worry about is Apple and their spiral into themselves and their closed ecosystem.
I think Apple was at their best when the adopted the PC architecture and interoperability made everything better.
> there are standard interfaces new companies can develop for.
Not so much in x86/x86_64 though. This ISA is a minefield of patents if you want to do anything state-of-the-art. AMD and Intel have mutual licencing agreements with each other, which prevents any upstarts from competing with them. Any competitors will have to license ARM or invest into RISC-V.
American fabs have focused on too many feel good measures to competitively compete with the Asian fabs. The new TSMC Arizona HR director came from Intel and one of the first questions out of his mouth was about TSMC's commitment to diversity. TSMC doesn't have a commitment to diversity...it has a commitment to hire the best and brightest to bring value to their end customers.
By all means, hire every diverse candidate you can if they’re the best and the brightest. Diversity and inclusion is a sure fire way to hire mediocre people. Every culture and every race has some brilliant folks. Hire them.
Exactly. In some cases getting a diverse background of opinions and viewpoints can make your company better. Like a clothing company or social media company. However when you are talking about advancing technology through modifying physics and scaling up systems then you simply need the best and brightest.
The company has to react both tactically and strategically. Short term, they may have to use overseas production while the domestic capability improves.
TSMC knows Intel is a fab competitor, and everyone involved in the relationship will be thinking first about how to protect TSMC interests and trade secrets, and secondly servicing Intel.
> "Having 80% of all supply in Asia simply isn't a palatable manner for the world to have its view of the most critical technology," Mr Gelsinger said.
Eh, the whole set of FAANG plus Intel itself is based on American West Coast, and this guy's take is that so many chips being made in East Asia is not "palatable" for the world?
There's nothing wrong with saying "That's a large pie they're eating and we want a bigger slice for ourselves." We could do without this kind of self-serving pseudo-cosmopolitanism.
I think you missed the point. He isn't going to outright say "If China decides to follow-through on their threats to invade Taiwan it will cause a worldwide disruption to technology".
But at the end of the day, that IS reality, and it's a systemic threat to both Europe and the US. The fact it's self-serving doesn't make the statement any less factually accurate or important.
That's not really a counterpoint though. It is kind of bad for the world that FAANG++ and for example most of the movie industry is centered in US (west coast)?
Also with tensions rising there it is not good if 80% of the chips are being made in a war zone. And bad for everyone but China if China takes control of Taiwan.
He's really just deflecting from Intel dropping the ball. He didn't talk about how we got here, he's just campaigning for support for a company that's had a hard time executing by playing the politics card.
I'm not an Intel hater, but this lack of ownership and deflection isn't something I like to see in CEOs.
Intel spends $13.5B a year on R&D. AMD spends $2B. TSMC spends about $3B.
Spending more cash isn't the problem. I know it's easy to blame the bean counters for everything, but in the chip-making business, today's problems are created 5 years ago, and if Bob did a good job, we'll see it years after his departure.
I think Intel are more focused upon not surfing the bleeding edge, more the bulk of other silicon production that is usually on many generations behind. So offering those the opportunity to step up in nodes for them but also using proven nodes which , whilst not bleeding edge - will do ok. That with any bleeding edge needs, being open to outsourcing on a case by case bases for their own parts.
I think for many, they view Intel based upon how well they do running computer games and think the stock resolves around that when we have learned that their income streams are more widely spread.
As for alternative to TSMC, that gets down to getting ahead on nodes or when nodes get to a stage that they won't get any smaller. Also the work with IBM is interesting and whilst IBM left the FAB industry, they still do research and that combination may help Intel catchup or surpase the current cutting edge TSMC offer.
I do feel what Intel is going for is opening up their 14nm(insert plus signs to taste) as they shift their core onto the 10nm. They had a lot of investment in 14nm and milked it well, but now they have the opportunity to open that up and gain value from that instead of the usual retooling one node into a new node and some node changes will be more impacting and might well be the case. Let alone the downtime and then to get the plant particle free after all the work - very costly. So to cash in on that as is more and just build new plants from scratch can for the accountants become very much the right direction to go and that is how it seems to be. More so when politically getting subsidies to build new plants over retooling an existing one can make that option even cheaper.
However it is being driven, it's good for Intel and the industry as a whole as more choice. Equally I'm very interested in how Intel and IBM will collaborate and have a good feeling about all this.
TSMC is far too important to far too many parties. We're getting dangerously close to a direct US/China war for Taiwan. That wouldn't be good for anybody. Both sides will claim it's about something else, but really it's all about the chips.
Both sides will arm for war, mostly as a deturant and bargaining point.
Espionage at a corporate and nation state level will be used to erode the value of the problem; probably by every advanced nation state.
The other 21st century and beyond area of competition will be energy tech (of all sorts), because of that old science fiction meme of the only two things an advanced species need being space and energy.
Arm has a better business model because there are so many great computer engineers in the world and with ARM they all create new SoCs.
Because of Arm's model there are many diverse ARM SoCs, and ARM is the leader in everything from wireless keyboards up to the fastest smartphones. ARM is also in the fastest supercomputer. In the areas of desktops and servers, ARM will keep gaining popularity because there will be many new ARM-based designs for every 1 produced by Intel or AMD. Intel will never catch up when they are competing against many excellent companies that are all sharing some of their code.
This could be great. Personal anecdote time, when I was using ESP32 there were some things it couldn't handle due to a very unconfigurable WiFi interface -- captive portal advertisement via DHCP specifically. (Before people get their knickers in a twist, it was connecting to a box that had no internet access, so there was no hostage situation there.) It turns out there are basically no open source or even American options that replace that chip. A lack of competition leads to a stymied market.
There are tons of options that aren't ESP32s. Many just aren't in the package that you are used to with a ESP32. Nordic, NXP, SiLabs and Microchip all have pretty comparable offerings at a similar price point for bare ICs (and if you are doing anything outside prototyping/small volume, you shouldn't be buying modules IMHO anyways).
If you are talking about ESP32 modules, there are still options but they are typically a bit more expensive, like $1-2 more last time I checked. However, there are ARM based wifi/bt modules that you can get as cheap as $1.50-$2.00 for an FCC certified one. Check out Fanstel BlueNor [1] as an example. The huge advantage being you get to use ARM tools instead of the Xtensa stuff, which depending on your workflows can be a definite plus (though I believe they recently added xtensa support in LLVM, so that might be a bit of a moot point).
They are definitely going to be difficult to compete with on the new ESP32-C3 line price-wise once they ramp it up though. At the end of the day though the ESP offerings have a bit of tradeoffs when you look at battery life and performance compared to other offerings. I believe they are working or may have fixed the power usage/sleep problems though on the C3s.
When it is all said and done though, I think the ESP32-C3 is going to probably take over the sub $1-2 IOT space, at the end of the day batteries are cheap and if you are designing your product to be plugged in... who cares.
What personally burns me with Espressif is they charge so much for their ICs. Like sometimes the modules are cheaper than the bare chips, even at 1k quantity, which is kinda irritating when you are space constrained.
> I've just learned that the ESP32 is manufactured on a TSMC 40mm process. That's essentially caveman tech.
But it's cheap, and MOQs are still achievable for mortals.
This is why.
There are way more reasons older litho processes persist beyond just them still making tons of money.
40nm for a microcontroller is rather impressive, given that the most are made with anything up to 180nm.
It takes more effort to design a chip with scaling down of geometries, lead times, and test samples are getting longer to get. Mask sets get more expensive. Less fabs are available to produce the chip. Some devices become unavailable at lower geometries.
Probably the only reason that line is still running would be validated long term stability products, or niche hardware designed with special application compatibility (E.G. radhard, higher end mil / aerospace concerns, maybe automotive).
Semiconductor tooling is rather expensive and I imagine that updating this to a more modern line would require overhauling and re-balancing many systems. It might help someone not familiar with the industry to imagine each technology node as an entirely new model of car.
There are some projects where people try to etch silicon in their garage, but the node they are targetting is from last century. Anything newer and you need toxic and corrosive chemicals, and increasingly exotic equipment. The yield isn't great either.
I'm all for redistributing fab capacity geographically. But I thought it became concentrated in Asia because of the lower cost for high quality engineers due to LCOL. How will Intel change this equation to make it profitable in other geographic regions?
USA has some of the lowest cost of energy / electricity in the world and one of the best rail-freight systems in the world.
Our passenger rail sucks and our people are expensive. But significant manufacturing remains in the USA (yes, we manufacture more today than ever before, despite the hoopla about outsourcing). There's a reason for that: USA has some pretty good advantages if you know what to look for.
I mean... only when we (Americans) feel like getting things done.
A lot of it is just arguing with ourselves over what to do. But if a consensus is reached, Americans can get it done pretty quickly. There's just so much discussion / debate / noise now.
America has a ton of capacity and we've got no true shortage of good (qualified) people.
Profit / rent-seeking and protectionism are the downfalls of my country (America). I identify that as the root fault of all of our social ills and inability to get work done outside of crisis and people with the right kind of crazy middle-finger to typical business interests drive to be the real life versions of our best superheros. I really do hope Elon Musk gets our species to building-across-the-street backup replication level.
I'd say that's one of the reasons why Europeans are so unhappy with the US. I know I am. They have so much potential to get shit done, but they squander all of that potential.
Yeah, it's not cost of living, it's that everything is there... it's a chicken-egg because even if you move the physical fab, everything you need still needs to be transported or you need to convince people to move with you.
This isn't a one year project, this could be a 25 year project to get the entire pipeline from raw resource to a piece of silicon.
There are some problems with doing business in Latin America. Corruption (which manifests itself as bureaucracy that can commonly be accelerated by greasing palms) and taxation are both fairly high in many places. It's not all about how much you pay your workers.
In my opinion the potential for a greater Pan-American economy is the number one counterargument to Chinese global dominance. We have almost everything we need, right here in the Americas.
Pan-American anything often boils down to just good ol’ NAFTA +
Colombia.
Brazil and Argentina feed China, benefit from China’s presence re: the US, and Buenos Aires is as far away from Miami as London is. Throw in the guay’s cause geography, and Bolivia because Bolivia, most of South America’s productive land isn’t counter arguing.
Because any reasonably committed extent of that economy ends at the Amazon, it’s just NAFTA + Colombia + Peru + Chile. Colombia’s a great get and arguably already in progress, but Chile is tiny and Peru is far.
The rest is some of the most picturesque, violent, non trivially developed (expensive!) tropical geography on earth that’s been in the pits for centuries even by extractive colonial standards for a reason, with the Caribbean part anchored by Cuba and Venezuela.
Total population addition less than Mexico, spread across a region the size of Australia, with the “easy” wins a 1000 miles away tucked into the Andes.
Corruption actually makes doing business EASIER for a company with stockpiles of cash. It can even help you avoid taxes. So it is not a problem, it's an opportunity.
There's a threshold where it becomes near impossible to do business. I've seen laptops intended for employees... "disappearing" in customs. When a complaint was made, the reply was to not make a fuss, otherwise more laptops would sprout legs and run away.
Corruption also tends to come with lots of bureaucracy. At every level you have palms to grease. Things tend to take a while unless said palms (which sometimes change often) are properly and generously greased. Then you have different government agencies competing to get the... "funding" - and your company is caught in the crossfire. Good luck getting the components you need imported in a timely manner. That is, until you have finished setting up your... uh, "supply chain"? Which can take a while, because it's obviously all behind the scenes. Hope you have a good "guide".
In the end, you have a government that's working _against_ your company. They will (ironically enough) want to extract maximum value from you and your company. If the place is bad enough, your competitors won't be playing in a fair field either. It's not only how much you can offer, but who you know and what sort of dirt you have on them.
When corruption is really high, neither the police or the justice system will be fair to you. You better hope you have identified everyone that needs palm lubrication - otherwise, you know that particular judge you forgot about? Your competitor has them in their pocket.
Since the police in really corrupt places tend to be ineffective, another way a competitor can cripple your operations is to make you vanish. If it is a place where violent crime is common(not a bad assumption), then it was just another unfortunate statistic.
You don't really want that kind of "opportunity". It's not worth it to decrease taxation. If anything, you will be taxed just as much, except that's now off the books.
Not only you need stockpiles of cash, but you also need to understand the dynamics of the place you are trying to get into. Otherwise, you are just a wealthy(and clueless) mark.
Of course, said company would have to be based entirely in the target country. It couldn't be US-based with a foreign branch, otherwise anti-corruption laws would apply.
Source: born in a pretty corrupt country. Seen a fair amount of shady crap. Trust me, you don't want to deal with that stuff.
Labor costs in a modern fab are a pretty small part of total costs. The fab construction itself as well as filling it with semiconductor manufacturing equipment (like those ASML EUV lithography machines) swamp labor costs.
Taiwan and Japan being the giants in the field is due to one thing only: Government policy. Both governments invested heavily in research labs, universities and industry (TSMC was a joint government-Philips venture), setting them up for where they are now. Cost of living isn't relevant here.
This have been bothering me for a while. What would happen if for any reasons the biggest fabs are destroyed? It would take years to get more than half of a century of accumulated knowledge just from patents, and that's not even counting industrial secrets and all the economical machine needed to support such an industry.
There have been projects focused on securing Humanity's technological knowledge in case of a worldwide calamity like GitHub storing all its source code in microfilms, but I can't think of any project entirely focused on building from scratch and kick-starting the industrial motor of modern civilization.
> What would happen if for any reasons the biggest fabs are destroyed?
it "just" takes some years and heaping piles of money to replace them. though probably more if you're trying to replace them all at once, as there aren't a lot of folks who make the equipment to rebuild the fabs.
> It would take years to get more than half of a century of accumulated knowledge just from patents, and that's not even counting industrial secrets and all the economical machine needed to support such an industry.
sounds like you're more concerned about loss of the personnel and records, rather than the buildings.
> I can't think of any project entirely focused on building from scratch and kick-starting the industrial motor of modern civilization.
i mean, "world peace" is generally concerned with preventing it from being destroyed in the first place.
conventional wisdom is that taiwan bases a chunk of their national security on nobody wanting to see TSMC ruined.
> it "just" takes some years and heaping piles of money to replace them. though probably more if you're trying to replace them all at once, as there aren't a lot of folks who make the equipment to rebuild the fabs.
You can't buy an economy. You'll need a petrochemical industry for example for all the insulators and other substances that go into the making of chips and you'll need mines for precious metals or recycle them from scratch and of course an energy industry. Just those few examples are their own are giant economical pyramids on their own rights, and would be very hard to replace even if most relevant university graduates made it alive to this point.
> sounds like you're more concerned about loss of the personnel and records, rather than the buildings.
I don't know you say this, I was more focused on the process as a whole on a macro scale.
> i mean, "world peace" is generally concerned with preventing it from being destroyed in the first place.
I never mentioned war in particular. Could be a solar flare, an asteroid, a plague (and look how we were prepared for this one).
> ...I can't think of any project entirely focused on building from scratch and kick-starting the industrial motor of modern civilization.
That's a national defense-scale ambition, and it is akin to asking, "what are the minimal constraints to reproduce X tech tree with Y years steady state sustainability?" You're basically asking what it takes to establish an independent human colony without support from home.
The Long Now project is addressing a small sliver of the question by looking into what it takes to long-term sustainably perform some acts of information storage and dissemination.
Folks into preparedness kick the can on the problem, assuming some certain level of infrastructure remains to pick over the remains of to rebuild/restart. The ones who can hold out for 10-20 years independently before rebooting over the remains of civilization are considered extremely well-funded and hardcore, but their generations-spanning outlook is pretty dim. Even the primitive skills-oriented teams have no good solution to how to pass those skills onto future generations, other than bigger communities to spread the information loss risk, and these teams are typically too small (a couple dozen at most) to look into that direction for a solution.
But no one to my knowledge is working on the overall tech tree problem. Our entire civilization runs as a proprietary binary blob, with no reproducibility in the event of a DR incident.
Even if you "just" "settle" for "only" a 1940-era tech base, we're probably looking at a minimum 45.3M people to participate. The 1939 "Census of Manufactures" identified 453 separate industry classifications. Guesstimate 100,000 absolute minimum number of people per classification, and that's going to be a lowball number of active participants. Toss in dependents and families, plus ancillary support and we're probably around 60M minimum? US 1939 population was 130.9M, so I'm guesstimating we can't get away with less than 50% to create a sustainable "reaction", 66M in round numbers. I'd probably peg it around 75-80% for robustness, and take all the information encoding technology we have today to increase the density of knowledge and know-how per worker to increase the robustness and recovery speed simultaneously instead of decreasing the number of people in the enterprise. Even with the benefit of perfect 20/20 hindsight of what worked and what did not in the past 80 years, it will likely still be a long slog.
To give you a concrete glimpse into what "from scratch" entails, just look at the Gingery book series of what it takes to go from charcoal and scrap metal to a basic machine shop. Getting charcoal from bio-matter is a pretty straightforward task, if you've practiced it before. If you want to extend "scratch" to pulling ore and refining it, now the effort level gets to pretty large scales. Then coming up with refractory-grade clay. Don't forget how to feed yourself in the meantime. The complexity scale just to get beyond subsistence level from scratch is quite harrowing.
Civilization is the most complex artifact the human species has created to date. Re-creating it from scratch would be a pretty tall order that no one has yet to offer even a proof of concept upon. I do welcome discussion on how to solve this with alternatives.
I wasn't aware of the long now project, I'll look into it.
The point is that it's not totally from scratch and there's a very good proportion that can be passed on easily and a lot of trial & errors and research doesn't have to be done again.
Mathematics can be learned with only books, important concept such as calculus doesn't have to be rediscovered.
Major scientific discoveries don't have to be reinvented.
Hard learned mistakes can be learned from History without making them.
Prospected resources ___location can be saved and are static by nature.
Also assuming that nothing whatsoever will remain is unlikely. Any such project should provision a set of plans to accommodate multiple level of destruction from the Bronze Age to pre-WW2 technology.
Having some kind of roadmap with essential milestones like making concrete and the Haber process will save a lot of time. Disseminating the information in different ways such as stone tablets (like the Voyager Golden Record) to micro films could be done in a large scale for redundancy and could also be done for relatively cheap.
> The point is that it's not totally from scratch and there's a very good proportion that can be passed on easily and a lot of trial & errors and research doesn't have to be done again.
The catch is there is a world of difference between what is in written form and operational practice, and yet another world of difference in operational practice at scale. Most of what is unwritten in civilization are operational practice, and doing so at scale; most of that is encoded behind commercial enterprises' walled gardens of information.
As a concrete example, figure out what it would take to reconstruct the melt blown material manufacturing equipment. Or what it would take to construct an MDX lab. Or "just" a machine shop like Fireball Tools' capabilities. That kind of knowledge is extremely slow to percolate into an information transmissible form, if it ever makes it there. We don't have mind state capture and replay, unfortunately.
There is lots of material on an even "lower" tech base-oriented activity, small-scale farming and/or ranching. And out of the tens of thousands of people in the developed world every year who want to take the plunge to grow their own food, there is insufficient "cookbook" recipes they can follow to enable them to easily do so, and of the many who do take the plunge, most do not stick it out past five years.
People get a taste of this dynamic anytime they venture into hobbies that require some form of self-sufficiency, like hiking, camping, blue water sailing, and so on. The logistical tail to support those activities to where they can "launch" is quite extensive.
I agree infrastructure-damaging incidents are extremely unlikely to uniformly take out all supply chains across the board. But planning-wise, there is no way to tell which parts of infrastructure would be adversely affected. The best we have been able to do as a species is to route around the damage on a best effort basis, relying upon the diversity/redundancy of sourcing and logistical alternatives, like with the COVID-19 pandemic and the Suez Canal backlog.
As a "DR plan for civilization" posture however, that is insufficient. Like Microsoft Word, such a planner has to cater to lots of different scenarios (users), each with its own unique set of induced infrastructure gaps (requirements), ending up with a Franken-plan (Microsoft Word). It's not an easy or compact problem to solve as far as I can tell, though I'd be happy to be shown wrong. If we ever want to make off-world settlement a reality in the far future, then we have to solve this problem to some degree.
The thought has occurred to me that making such a plan too compact holds within it a dark pattern: if it is too accessible to too-small an entity, then it encourages elites with sufficient resources and megalomania to pursue their own ideations of a world created in their own image. I suspect without humaniform robots and GAI however, that won't become a possible reality. I believe the scale and complexity of mind states to transfer around (densities of which are limited by human lifespans and in the base case, available energy) to accomplish the goal dictates pretty large population requirements out of reach of all but combinations of developed-world nation-states.
It isn't all dire. We've solved this before, up to around the 17-18th century, arguably tapering off the competence around the 19th century. We haven't dusted off the outdated DR plan and updated it for our new "platform".
The other big problem is that countries that don't produce complete consumer electronics products have to export their ICs to China. If you're not a full-stack country, you have to get along with China.
Why do we need to keep secrets? If chips are distributed evenly, there will be no advantage given to anyone but lots of people can reap the benefits.
I don't like nationalist arguments, but I do think it is a bad idea for critical tech to be concentrated in a single place. It would be good for there to be advanced manufacturing sites on all six populated continents owned by different countries.
Because every geopolitical competitor is keeping their own secrets. Makes no sense to disadvantage yourself when your peer and near-peer competitors aren't.
I don't understand this thinking. Why would we want to impoverish others? After all, it's not like these power brokers give a shit about the American public, it's about great power competition.
Is it better to be beholden to another party, or be the one the other party is beholden to? Which also means you are more self-reliant. If trade relations soured to the point no more chips exported to your country, how would your country handle that? How long would it be before other fabs could even come close to competing?
I agree. I think realistically, the US will always have its own fab capability for the foreseeable future. I think if the capability were spread around the world though as I described above, it would be pretty difficult to cut off any one country.
If you wanted to get really out there, I wonder if it would be possible to place certain capabilities under international control. I think this is already the case with the ITER fusion project, we don't usually do this with commodities.
first of all, I don't think "impoverish" is quite the right word. if you have a 9700k and I buy a 10900k, my cpu is faster, but your cpu is no slower than it was yesterday.
past that, I don't really understand what there is to explain. if it would be good for everyone to have $1000, it would be even better (as an individual) to have $10000. assuming you care about money/chips/whatever. if you don't it obviously doesn't matter.
edit: ah, I see what you are really asking is "why should we care who wins a power struggle among the world's elite?". I guess maybe we shouldn't, but personally I'd rather be on the winning team, even if I think the "teams" are silly.
I don't see how depriving others of technological advancement comports with any kind of justifiable theory of social governance, it's just a method of dominating others.
I am not convinced we even benefit much from these kinds of antics. We have a few more consumer goods, but fundamental goods (education, housing, medicine, food) are sky-high in price, there are few worker protections, and huge portions of the budget are dedicated to illegal military adventures. I think cooperating and sharing technology widely leads to a richer world.
When I drive around the city I see a Sprouts store and signs all over the place that say "Locally Grown" or "Buy Local".
There's a reason for this. Every time you pay money that goes into the community there's a higher chance for local prosperity.
If we start eating beets from China some beet farmer may be rich and his family will definitely be happy. But why not buy local beets that enriches our communities.
So at some point Sprouts made the decision to instead of buying beets from a Chinese farmer, they will buy them from a local one. However Sprouts, as a national chain of grocery stores, could have decided one day to just make an extra $1 Billion by selling MOSTLY out of country food. They don't do this for 2 reasons - fear of backlash that all the food is from China and 2 these people do actually understand the local connection.
For some reason this logic does not translate to silicon valley companies. Those companies have made the choice that because of the price difference, it's best to buy it from far away. When you look at everything that's going on in the USA - the deep divides between rich and poor, between left and right - most of that is BECAUSE of economic inequity. If we fix the economic inequity, a lot of political shit will just go away. We have a reason to fight because things AREN'T going so well. Everyone has a different fix.
I personally think the fix is to come up with policy that Sprouts has: BUY LOCALLY GROWN.
The main reason we hear about the rich getting richer and the poor getting poorer is because those rich execs at Walmart and tech companies are paying people in China, India and wherever else they can pay the lowest possible price and resell it to Americans. It makes perfect sense why the 1% is becoming richer and richer. And raising taxes on them isn't the answer - it's to stop the exporting of American dollars to sell shit to Americans. Duh?!
How is that? As the article explains, they have just announced their plans to directly compete against TSMC/Samsung as a manufacturer of chips for other companies (rather than just manufacturing their own designs).
To me, this suggests Gelsinger is confident in the company’s ability to get their 7nm-and-beyond processes back on track and become competitive once again.
> this suggests Gelsinger is confident in the company’s ability to get their 7nm-and-beyond processes back
Or it may suggest that the company wants to say anything to reassure investors that it will change course and stop losing market share. I wouldn't read any particular guarantee into these statements.
That’s a fair point. A comment [0] in a parallel thread points out a third possibility: that Intel believes x86 is on its way out, and they’re attempting to pivot before their only significant revenue source dries up.
I hope for the semiconductor industry’s sake they really are getting 7nm back on track this time, and that this new manufacturing business will be competitive. Perhaps that’s unrealistic to expect given their track record lately, but at least Gelsinger seems to deserve more confidence than Swan.
Rocket Lake is still on 14nm because of issues with their 10nm process and the new guy a week in is talking smack about them shipping 7nm product and compete against the likes of TMSC/Samsung ???.
They don't have to do 10 and 7 on the way to 5, right? I assume they can just skip some of the work, especially if the newer process is totally different anyway.
Yes, they're working on multiple processes. They're working on both 7nm and 5nm, and it's expected that 10nm's yield issues will effectively be solved by releasing 7nm.
Wow he really said that. Kinda stupid if you ask me, people may take it personally. This is why when you mean "We need a more distributed supply chain" you say "we need a more distributed supply chain" and don't single out a geoethnic region.
> “other worlds that don’t speak Indo-European languages”
There are hundreds of Indo-European languages spoken in Asia. One of them (Hindi) is the 4th most common first language in the world, after Mandarin, Spanish, and English[1]. (I'm actually surprised Arabic isn't higher on that list but maybe the various regional variants count as different languages?)
I'm not sure where you're getting your geography and linguistics information from.
Asia is a hyper broad term that could include areas from east to Russia to just west of Turkey except Middle East. Some thinks Central Asia is not or whatnot.
“Continent” refers to single contiguous land encircled by seawater, that aren’t islands, at least as I think of the word.
> “Continent” refers to single contiguous land encircled by seawater, that aren’t islands, at least as I think of the word.
To the best of my knowledge, you are literally the only person who uses that definition. By that definition Asia, Europe, and Africa would be a single continent. North and South America would be a single continent. What school did you go to that they taught you there are 4 continents in the entire world?
Isn't that six? Eurasia, Africa, Australia, North America, South America, and Antarctica, with India noted as "sub-continent" and Asia sometimes distinguished within Eurasia by mountain ridges.
By your own definition that can't be true. North and South America are a single continuous land mass. You can walk from the top of North America to the bottom of South America. The same with Eurasian and Africa - they are 100% a connected land mass - you can walk from Europe to Asia to Africa, no problem. So by your own definition there are only 4 continents. To the best of my knowledge, almost everyone else says there are 7-8 continents (depending on whether you count Zealandia or not).
These are joined together by land, and are not individually encircled by sea. The Panama Canal doesn't count, as it is manmade and would silt up if not maintained.
"Indo" in "Indo-European" comes from India, one of the two largest countries on the Asian continent. You might classify some languages spoken in India as.. Indo-European. Shocking. Those Indians who speak those languages.. are Asian! Surprise!
Since you missed third grade geography, the other continents in addition to Asia are Europe, North America, South America, Africa, Antarctica, Australia, and recently discovered to be its own continent, Zealandia
Good try though, projecting your own racism onto the topic at hand while simultaneously revealing incredible ignorance
In Britain they call people from India Asians all the time. Immigrants from the Indian subcontinent are the biggest nonwhite group, and that’s the word they are known by.
In the US, I think people often use “Asian” to euphemistically refer to people they used to call Orientals. The latter term is considered problematic. The new term hasn’t improved their treatment, though.
Found this definition by UN sub-dividing Asia into five(six) sub-regions[1], containing nations as Kazakhstan, China, Philippines, India, Saudi Arabia(, Russia) to name one from each.
I think the "India" prefix in the terminology you mentioned and the fact that euphemistic American "Asian" including only East/Sountheast Asians indicate that there is a widely recognized and practical distinction between the groups, which also aligns with the UN definition, despite both partly sharing the literal expression.
> In the US, I think people often use “Asian” to euphemistically refer to people they used to call Orientals.
I wonder why that is; while at the same time, Kamala Harris is widely considered the first Asian American Vice President, because her mother is a Tamil American (as am I).
How is his statement any different from saying too much of our oil comes from the Middle East or too much of our wine from Europe? Or are all those also forbidden statements?
Oil dependence on ME has always been an issue. Guess what the US did... They're now one of the biggest oil producers in the world. What was your argument again?
As for wine, wine isn't as fundamental/crucial to modern civilization as computer chips and oil.
The headline is a bit editorialized. Here's what he actually said:
> It also needs more geographically balanced supply. Today there's a heavy bias to Asia. And as we've seen, coming out of some of the disruptions and challenges, the world needs US and European supply in a more balanced way. It's the right thing for the global supply chain requirements, for both commercial as well as for government and defence use as well.
Isn’t he being politically aware by not naming the Asian country ? Because as a technical person what’s the concern if a lot of chips are made in the continent of Asia
Asia, in addition to being a continent, is an interlocked geopolitical sphere, with concerns like local wars, which can be amortized over by having factories in several countries too far apart to fight each other practically.
When you are a CEO of one of the biggest companies in the world (or even just a commenter on HN), even the smallest thing that you say can be misinterpreted by people. Pat said great things, but the newspapers picked the most controversial part of course.
Pointing out that there is a concentration of chip manufacturing in one region (or even country) is not xenophobic. It is globalist, and it is capitalist, but it's not xenophobic any more than it is to say that the concentration of social networks coming out of silicon valley is a problem (it is!).
When Gerald Marcyk was running Intel components research back in the 2001 Intel was already considering using EUV lithography for the 45nm technology node that would come out 2007. It turned out that the switch to EUV could be delayed and delayed again. Intel thought they could get away with it this last time and it was a wrong decision.
EUV is super hard. There is only one company that can make EUV lithography machines, Dutch company called ASML. TWINSCAN NXE:3400B costs ~$120 million https://www.asml.com/en/products/euv-lithography-systems/twi...
Intel, Samsung and TSMC invested billions into ASML in 2012. In exchange Intel got 15% ownership, Samsung 10% and TSML 5% if I remember correctly. They don't have any significant ownership anymore.