His channel has awesome videos in general, very informative and no way boring.
Already saw this one, nice to see it getting more exposure. And oh boy, China was already working on cpu independence, it looks like they are rapidly accelerating their efforts and thus results.
Like he said they have the talent all they have to do is throw bodies at it. I can see them reach parity sooner rather than later but that's what he said.
they have the talent all they have to do is throw bodies at it
IMO this is underselling it. PRC at ~530k/700k of domestic IC talent they think they need for complete indigenous industry according to IC2018 white paper (IIRC same year they started prioritizing talent production to first-level discipline), and now academic overcapacity is spamming ~30k IC talent per year who'll aggregate expertise and integrate into PRC IC. This makes PRC basically the only actor projected to have no IC talent shortage in medium/long term, and indeed only one with enough aggregate talent for fully indigenous semi supply chain. For reference every western semi power is projected to have 100-400k semi talent shortfall by 2030s.
At end of the day, EUV is not black magic, a handful of countries with ~40% the population of PRC, took their time to coordinate/develop LPP for 30 years, with limited resources and industrial policy. Broadly integrated circuits hasn't even been a particularly well compensated sector so questionable if western semi was even stacked with best talent. Before first commercial EUV system was released ASML had ~10k employees, Zeiss SMT had ~3k, Cymer had ~1k. It wasn't a massive commercial/strategic undertaking. For reference, Boeing/Lockheed/Airbus had like 100-150k in 2010s. Hard to say how fast PRC can close gap / reach parity with Manhattan level effort.
That is wrong. The development of EUV technology was primary a government investment involving leading European union, American and east Asian government institutions.
These consisted of major programs such as SEMATECH, INVENT, EUCLIDES, ASET-EUVL involving leading organisations such as Lawrence Livermore National Laboratory (LLNL), Lawrence Berkeley National Laboratory (LBNL), Sandia National Laboratories, Himeji Institute of technology, IMEC, CEA-LETI.
High level research is primary dominated by master/doctorate graduates. China has a 25-64 year-old population rate of only 1.2% of the population compared to 15% of OCED nations.
This is missing the point. I highlight headcount of companies commercializing EUV vs another high complexity industry like aerospace for simple scale comparison of bodies involved to rollout and sustain commercial EUV. Obviously institutional research was involved, the 530/700k is for all sectors of industry including academia. If anything adding up all the headcount of research orgs would be smaller than broader commercial sector. At the end of the day if you want to commercialize sector you still need all the other heads. There's plenty for research, but 100,000s of fab monkeys and other technical staff for every other piece of the semi supply chain. That's what every other semi actor (TW, JP, SKR, US) are projected to be short on, as a bloc.
If your point is access to high level talent, then PRC is, without exaggeration, in process of developing and exploiting the greatest "high skill" demographic dividend in recorded history, and will be able to milk stupid disproportional amount of high level talent (relative to competitors, likely even western bloc in aggregate) for decades while they remain in workforce i.e. well past 2060s. Their high skilled talent advantage is borderline insurmountable short of AI take off.
Focusing on 25-64 year old misses/skews picture, focus on 25-40. PRC tertiary enrollment is split by generations, i.e. pre 00s workforce was like low single digit % tertiary because PRC poor and tertiary not prioritized, but post 00s academic reforms (building out tertiary + R&D institutes), and they've been brrrting tertiary, at PRC population denominator scale. Overcapacity if you will. Currently PRC tertiary enrollment is ~60% vs 10-20% in 00s-10s, but larger cohorts then, i.e. 10-20% of 15m+ births then vs 60% of ~10m births now (births as in accessible 18 year olds). PhD levels exploded as well, currently ~80K STEM PhD vs ~50K US. The 00s-10s cohorts are 40s by now, large reason why PRC rapidly catching up in every high end sector last few years is because they finally have a lot of master/doctorate level talent, not just graduated, but with years of lab and/or industry experience to get commercialization rolling. We're basically in time frame where PRC has reached parity or exceeds US in aggregate STEM (and PhD) workforce gap (especially doctoral level) set to grow larger for decades. At current trendline, PRC is going to spit out 120m tertiary @60% of 200m births from past ~20 years in next ~20 years, with ~40% STEM, it means they'll be adding ~50M STEM, aka roughly as much as US is projected to add population total (i.e. births + immigration). As in if every US newborn and immigrant is STEM, it will merely keep pace with PRC at current trends. PRC talent pool large enough to outcompete with OECD block (since they have the numbers and ability to coordinate in one jurisdiction). Again, that's just STEM, there's 80M other tertiary workers.
I highlight the 2018 IC white paper is to show PRC is actively producing high level talent specific to IC, at scale. Quick correction that IC was only elevated to first-level discipline, i.e. massive expansion in IC doctoral program in ~2020 (not 2018). They went from handful to ~30 new PhD programs for IC in ~2020, with talent accelerator plan to get to ~50-60 by this year. The amount of "high level" talent specific to IC, i.e. hungry (and patriotic) 30-40 y/o PhDs in prime of career is exploding - full time doctorates candidates started graduating last year, many of whom intern/work for PRC IC companies throughout. Current goal is increase to 30-50k IC talent with ~1000-2000 PhDs per year for high end R&D. This is why people meme about US export controls & sanctions helping/forcing PRC semi to coordinate and indigenize... previously we had 10+ years of PRC developing shit all human capital with respect to IC, instead throwing 100s of billions in Big Fund to buy (and squander) expensive western semi equipment. Now PRC forced to also spend 100s billions to develop stupendous amount of high end domestic talent so they can build everything themselves.
It is missing the point. I highlight headcount of companies commercializing EUV vs another high complexity industry like aerospace for simple scale comparison of bodies involved to rollout commercial EUV. Obviously institutional research was involved, he..............
These industries are not comparable. The early and current research within the EUV industry exclusively required the highest levels of expertise across multiple disciplines, including optics, laser physics, cryogenics, precision engineering, metrology, and materials science. While aviation is primary dictated towards large scale manufacturing.
If anything adding up all the headcount of research orgs would be smaller than broader commercial sector. At the end of the day if you want to commercialize sector you still need all the other heads. There's plenty for research, but 100,000s of fab monkeys and other technical staff for every.......................
In accordance with data from the UNESCO Institute for Statistics the total compilation of researchers in the western bloc involved with EUV is 5.3 million verses 2.2 million from China. While for the year 2000 it was 3.1 million. Still larger than China today.
Focusing on 25–64-year-old misses/skews picture, focus on 25-40 and new tertiary cohorts. PRC tertiary enrollment is split by generations, i.e. pre 00s workforce was like low single dig.....................
The link I provided above showed the education ratio for 20-34 years old. The S&E degrees awarded per 100,000 was 14 for China; 60 United States; and about 80 for the rest of the developed world. The developed world also has persistent unaccounted migration from the undeveloped world which is also going through an education transformation.
At current trendline, PRC is going to spit out 120m tertiary 60% of 200m births from past \~20 years in next \~20 years, with \~40% STEM, it m
Half of China's tertiary are from short-cycle tertiary education. Not full tertiary education. While the rest are predominately ISCED-6. In terms of pure numbers, India actually leads in graduates and is set to grow over the next 10-20 years over China in STEM. It's clearly not an accurate assessment given disparity in education facilities.
aviation is primary dictated towards large scale manufacturing
There are large scale and complex R&D efforts in aviation, especially leading edge like military aviation for engine cores taps into many of those fields. Before current western semi indy policy turned on money printer, aviation sector R&D better resourced and funded.
compilation of researchers in the western bloc involved with EUV is 5.3 million verses 2.2 million from China.
What are relevant #s involved in EUV research + related semi supply chain specifically? Because it's not 5 MILLION by order of magnitude. Don't know if you're just crawling chatbots to find irrelevant stats where OECD/west number is bigger like total tertiary academic educators (I assume inclusive of humanities). Sector relevant info / estimates UNESCO wouldn't even have (you'd have to dig from industry specific writing for figures), i.e. simply appealing to UNESCO stats and alleging 5M+ involved in EUV doesn't pass basic smell test, it's projecting faulty logic analysis to the point of not even being wrong.
education ratio for 20-34
Stats from 2020s outdated relative to PRC speed. Regardless, absolute values > ratios for strategic competition. It matters who has most talent. PRC is producing ~4x more STEM vs US now, around OECD level combined. Geo/domestic politics limit west ability to add talent, i.e. not likely for US (or semi bloc powers) to increase annual immigration by millions of STEM to keep pace with PRC aggregate talent production, regardless if developing world increasing total talent supply. They could, but current short/medium term reality based on what's in pipelines is PRC heading towards IC talent oversupply and western bloc undersupply. And western bloc being increasingly immigration unfriendly.
tertiary
High end IC specific talent generation is, i.e. graduate/doctoral level is by definition >ISCED5 aka not short-cycle. PRC has relevant high-level IC talent production pipeline, when I say IC PhDs graduating in 4 years, I mean they're hamming full time PhDs, i.e. not leisuring part time, i.e. they're not doing lesser programs. More relevant, is PRC also has the industrial supply chains and ability to coordinate brains for industrial policy successfully, something LBH India isn't in same league in.
not an accurate assessment
Look at China. Look at India. Look at relative PRC vs US catchup recent times, PRC rapidly moving up various global R&D / S&T / innovation indexes, including western pubs controlled for quality, moving up supply chains etc, reaching parity or now leading in various sectors. All in last 10 years. All because prior 10 hammered out relevant tertiary talent. Look at actual increasing PRC competitiveness. That pool of talent will be 2x-3x in next 20 years. Why dance around irrelevant UNESCO stats or ISCED definitions, when reality / accurate assessment has obvious PRC bias. We know the talent PRC generated and continues to generate can deliver results, fast. We know PRC adding them in large numbers. We know last 10 years of PRC going from fraction to parity aggregate talent numbers with US already enabled them to rapidly catch up across the board in spectrum of sectors. And we know we're in phase where PRC aggregate numbers exploding past parity with US, and in next 20 years their aggregate numbers will be multiple times more. It's pretty clear whose assessment is likely to be correct.
Don't know if you're just crawling chatbots to find irrelevant stats where OECD/west number is bigger like total tertiary academic educators (I assume inclusive of humanities).
I advise you don't attack and call the data bias from United Nations Educational, Scientific and Cultural Organization (UNESCO) just because it does not align with your narrative. In particular when you provided no counter source to disprove it.
What are relevant #s involved in EUV research + related semi supply chain specifically? Because it's not 5 MILLION by order of magnitude.
How is providing direct data showing the explicit number of researchers involved in research not evidence of academic and research capital. How else do you want to measure it, by number of lab coats?
High end IC specific talent generation is graduate/doctoral level which by definition >ISCED5 aka not short-cycle. PRC hasrelevanthigh-level IC talent production pipeline, when I say IC PhDs graduating in 4 years.... Stats from 2020s outdated relative to PRC speed.......
You're not listening. The core point of my argument is that extreme ultraviolent lithography (EUV) research is conducted at the highest academic level, requiring a doctoral or master's human capital expertise and decades of experience rather than just a bachelor's degree.
China will take decades to reach this level of human capital that western nations involved with EUV development have based on the current trend in ISCED-7 and ISCED-8.
I'm at awe at his rate of production, 1-2 long-form videos a week and they mostly sound very competently researched to a layman like me. A part of me does wonder if this is actually more of an indictment on his quality of research, cause YouTube does not have a shortage of slop producers churning out content.
The proof is in the pudding. I'd be more convinced if they were shipping these things. There's a Zhaoxin KX-6640MA powered PC on Aliexpress right now, and you can get PRC-made RISC-Vs, and that's not nothing. But it's also not showing any signs of this acceleration that some are touting.
Their goal isn't for consumer devices, its to produce cutting edge supercomputing chips and AI chips, the likes being used by Huawei & Tencent to run their models, not reliance on NVIDIA.
I always find it funny, that many U.S Americans (and especially the Left) seems to have a real hatred and some furious Asiaphobia (Sinophobia, to be precise) and are somehow afraid of everything Chinese and (alleged) espionage from Far East in general.
The joke is, that it was mainly no other country but the U.S. itself, which deported expertise and manufacturing of goods towards Far East since the Sixties and turned it up to 11 in the Seventies, only to outsource fabrication and manufacturing in general in exchange for corporate America to reap unheard of profits – Outsourcing coding-skills onto Indian low-wagers (especially Californian companies did that), the electronic expertise to Japanese low-wage work-horses or semiconductor-expertise in general and basically everything else in the academic STEM-fields to everything Far East.
Who could've possibly foreseen that, after decades of out-sourcing expertise to Far East, said very knowledge ends up there?!
Reddit's infamously notorious Anti-Chinese sentiment can really get exhausting at times, not allowing to freely discuss technology unbiased.
Absurd. China bans all western media and social platforms and undertakes tens of billions in external propaganda campaigns annually and you dare attempt to argue that the west is bias.
We have no physical prototype of any workable EUV machine, or any detailed documentation of it's existence with the only evidence consists of photos of a small machine and claims on TikTok. Any other instance this would be consider nonsense, but due to mass propaganda even a small box is able to generate mass propaganda sentiment.
Where did I claim that China is the good guy here? Did I refuted any Chinese propaganda?
You may have missed it in your baseless rant, but I was solely speaking on the fact, that China is basically the world's workbench, and how that came to pass by the West's very own hands and work of doing, by outsourcing expertise since the Seventies.
More like hardware enthusiasts who understand tech & geopolitics more than the average gamer.
We can see beyond the anti-China rhetoric coming from the US. Heck, we even know about the fact US gov funds billions of $ on anti-China propaganda in the West.
So we analyze these tech development with a lens that can see through the bullcrap.
I know it's an enormous catchup that China needed to nail. Economic,strategic capital and scientific challenge never seen since the economic comeback of US economy and never tried since after US dominated post cold war economics
But all US was betting on is that ASML expertise and worker shortage on photolithography worldwide would be enough for China to be starved out on RnD. But in capitalism you can solve problems by throwing money and attracting expertise. And China throwing both money and political will to train themselves for EUV. The last castle left in technological parity. I'm not commenting how much talent China attracted to learn how to build a foundation cause thats for the future us to find out. But the 10 year catchup estimating people give is in for a surprise imo.
It’s absolutely embarrassing how much we’ve fallen behind. The last time America was “great” we put the best minds together and created the most unconscionable weapon ever created. If you want to make America “great again” you have to invest in science and tech just like we did back then, and then establish utter dominance of Putin and the Middle East. Democracy either stands alone, or we take fascists and oligarchs down with us
The physical weapons themselves might not. But the newest chip for AI might decide future wars with data analysis, autonomous weapons and R&D.
Big shame if this amazing tech Ressource would go to that. But there are really cool alternatives, like Deepmind moving from Protein simulation and development to whole cells.
I've heard of a big TCP/IP network that seems pretty great. It can be accessed via submarine optical cables running at a few dozen terabits per second, via mobile satellite terminal, or maybe with the battery-powered gadget in your pocket using one of two different wireless protocols and infrastructures. Perhaps someday you'll use these Elbonian innovations to communicate with like-minded persons.
It's always faster to buy or steal or deduce existing knowledge than it is to be the entity developing new knowledge & processes. It's one of the weaknesses of being the first-mover or the dominant player in any market, they will have to overcome obstacles carving out a path that everyone else can easily follow behind, or spend the most to maintain a lead while others can spend less for the same results.
The US's strategy seemed far more viable when nodes were on two year cycles. But now that nodes are taking three years and progressing toward four it is going to give far more time for other interested parties to catch up, or at least close the gap.
developing new knowledge & processes. It's one of the weaknesses of being the first-mover
There's also consideration first-mover made the right call, i.e. PRC is also developing alternative EUV methods like SSMB/synchrotron which may leapfrog existing tech. IIRC, a few years ago, there was interview with ASML engineer who drove their EUV efforts. The TLDR is on paper / initial analysis - building synchrotron expensive, so they opted for plasma. But then it was many paragraphs of unplanned technical challenges. Whether engineering around said challenges piled on cost is for more technical people to speculate, but what we know is ASML EUV cost 200M USD each, next gen pegged 400m. Which is about the cost of multiple small or single medium size synchrotron facility in west. For reference, PRC's latest large scale synchrotron facility HEPS is 1.3km circumference, with 60+ beam lines for 90 tool/experiment stations, for ~700M USD. The conceptual design for SSMB right now is a much smaller ring designed to power multiple tools at 1 kW. It's not just node timelines - it's questionable if ASML route is going to remain dominant in terms of economics, only that it's the only working method despite sunk cost.
Be carefull and not fall in easy trap of cost comparison. ASML EUV machine is not only “making EUV”. The surce of radiation is the less expensive stuff, even if you use a syncrotron Line as surce you will need all the optics and control to use it for photolitography, that is where the cost explode. Then you need a foundry to be built around the syncrotron etc. I am not saying is not feasible or stupid, do not get me wrong. I am just saying is not 1x ASML = 2x syncrotron, so use syncrotron.
Isn't one of the arguments for synchrotron being that the beam quality higher so there's less challenge to optically optimize the final beam? There's also no tin to blast around so you dont have to constantly replace these very expensive mirrors. But obviously there are many thing in the pipeline that need to be mastered as well.
Yes, that is true. As I mention I am not saying it is wrong going in that direction. I am just saying to not simplify too much the math related with cost. I am pretty sure that as soon as the use of syncrotron is much convenient than the use of ASML equipments (or say similar alternatives) foundry will move accordingly.
My understanding is the selling point of SSMB according to Tsinghua is price per tool is suppose to be lower because beam manipulation on synchrotrons mature/explored technology. Another comment below mentioned, less complex mirrors when strong collimated beam from source. IIRC original napkin math during EUV exploration assumed synchrotron more expensive because assumption was EUV as infra cost more than EUV as tools that can be shipped on planes, but PRC very good at driving down infra costs. That said, PRC also seems very good at driving down tool costs, see recent account their domestic DUV supposedly 1/30th price of ASML. Whether to be believed or not, semi is a capex heavy game, hence imo economics will be large determiner of node timelines/adoption going forward.
The US's strategy seemed far more viable when nodes were on two year cycles
Increasingly the "strategy", to the extent it can be called one, is to assume you can maintain technological dominance solely by reactively crippling competitors instead of investing in cultivating talent domestically. Pretty obvious where what ends.
I'm not commenting how much talent China attracted to learn how to build a foundation cause thats for the future us to find out.
And with what is going on in the US right now, this will only intensify. The brain drain that is already happening will embolden China further. It will help Europe to an extent as well, but I think China will be the primary winner.
In the short term, China might not even need to develop their own EUV machines, DUV machines could suffice. There was a DUV-based version of TSMC's 5nm process. The efficiency gain between TSMC's 5nm and 3nm, the most advanced mass-production node on the planet, is only roughly 25%. This difference can be compensated for by a slight increase in chip size and a minor reduction in clock speed.
Agreed that a 5nm node would be competitive for a lot of applications, for a long time to come. It's under-appreciated here how useful legacy nodes can be, because there's a bias here to full-power, latest gaming-suitable CPUs.
An almost-extreme example is the Skywater 130nm process node with open-source PDK. 130nm, in general, was the node of the Pentium III, first used in 2000-2001. It's still probably better than what's used for a lot of lower-end microcontrollers today.
mysticzoom@reddit
His channel has awesome videos in general, very informative and no way boring.
Already saw this one, nice to see it getting more exposure. And oh boy, China was already working on cpu independence, it looks like they are rapidly accelerating their efforts and thus results.
Like he said they have the talent all they have to do is throw bodies at it. I can see them reach parity sooner rather than later but that's what he said.
dirtyid@reddit
IMO this is underselling it. PRC at ~530k/700k of domestic IC talent they think they need for complete indigenous industry according to IC2018 white paper (IIRC same year they started prioritizing talent production to first-level discipline), and now academic overcapacity is spamming ~30k IC talent per year who'll aggregate expertise and integrate into PRC IC. This makes PRC basically the only actor projected to have no IC talent shortage in medium/long term, and indeed only one with enough aggregate talent for fully indigenous semi supply chain. For reference every western semi power is projected to have 100-400k semi talent shortfall by 2030s.
At end of the day, EUV is not black magic, a handful of countries with ~40% the population of PRC, took their time to coordinate/develop LPP for 30 years, with limited resources and industrial policy. Broadly integrated circuits hasn't even been a particularly well compensated sector so questionable if western semi was even stacked with best talent. Before first commercial EUV system was released ASML had ~10k employees, Zeiss SMT had ~3k, Cymer had ~1k. It wasn't a massive commercial/strategic undertaking. For reference, Boeing/Lockheed/Airbus had like 100-150k in 2010s. Hard to say how fast PRC can close gap / reach parity with Manhattan level effort.
Ducky181@reddit
That is wrong. The development of EUV technology was primary a government investment involving leading European union, American and east Asian government institutions.
These consisted of major programs such as SEMATECH, INVENT, EUCLIDES, ASET-EUVL involving leading organisations such as Lawrence Livermore National Laboratory (LLNL), Lawrence Berkeley National Laboratory (LBNL), Sandia National Laboratories, Himeji Institute of technology, IMEC, CEA-LETI.
Schematic of the EUVL development history. | Download Scientific Diagram
High level research is primary dominated by master/doctorate graduates. China has a 25-64 year-old population rate of only 1.2% of the population compared to 15% of OCED nations.
Higher Education in Science and Engineering | NSF - National Science Foundation
dirtyid@reddit
This is missing the point. I highlight headcount of companies commercializing EUV vs another high complexity industry like aerospace for simple scale comparison of bodies involved to rollout and sustain commercial EUV. Obviously institutional research was involved, the 530/700k is for all sectors of industry including academia. If anything adding up all the headcount of research orgs would be smaller than broader commercial sector. At the end of the day if you want to commercialize sector you still need all the other heads. There's plenty for research, but 100,000s of fab monkeys and other technical staff for every other piece of the semi supply chain. That's what every other semi actor (TW, JP, SKR, US) are projected to be short on, as a bloc.
If your point is access to high level talent, then PRC is, without exaggeration, in process of developing and exploiting the greatest "high skill" demographic dividend in recorded history, and will be able to milk stupid disproportional amount of high level talent (relative to competitors, likely even western bloc in aggregate) for decades while they remain in workforce i.e. well past 2060s. Their high skilled talent advantage is borderline insurmountable short of AI take off.
Focusing on 25-64 year old misses/skews picture, focus on 25-40. PRC tertiary enrollment is split by generations, i.e. pre 00s workforce was like low single digit % tertiary because PRC poor and tertiary not prioritized, but post 00s academic reforms (building out tertiary + R&D institutes), and they've been brrrting tertiary, at PRC population denominator scale. Overcapacity if you will. Currently PRC tertiary enrollment is ~60% vs 10-20% in 00s-10s, but larger cohorts then, i.e. 10-20% of 15m+ births then vs 60% of ~10m births now (births as in accessible 18 year olds). PhD levels exploded as well, currently ~80K STEM PhD vs ~50K US. The 00s-10s cohorts are 40s by now, large reason why PRC rapidly catching up in every high end sector last few years is because they finally have a lot of master/doctorate level talent, not just graduated, but with years of lab and/or industry experience to get commercialization rolling. We're basically in time frame where PRC has reached parity or exceeds US in aggregate STEM (and PhD) workforce gap (especially doctoral level) set to grow larger for decades. At current trendline, PRC is going to spit out 120m tertiary @60% of 200m births from past ~20 years in next ~20 years, with ~40% STEM, it means they'll be adding ~50M STEM, aka roughly as much as US is projected to add population total (i.e. births + immigration). As in if every US newborn and immigrant is STEM, it will merely keep pace with PRC at current trends. PRC talent pool large enough to outcompete with OECD block (since they have the numbers and ability to coordinate in one jurisdiction). Again, that's just STEM, there's 80M other tertiary workers.
I highlight the 2018 IC white paper is to show PRC is actively producing high level talent specific to IC, at scale. Quick correction that IC was only elevated to first-level discipline, i.e. massive expansion in IC doctoral program in ~2020 (not 2018). They went from handful to ~30 new PhD programs for IC in ~2020, with talent accelerator plan to get to ~50-60 by this year. The amount of "high level" talent specific to IC, i.e. hungry (and patriotic) 30-40 y/o PhDs in prime of career is exploding - full time doctorates candidates started graduating last year, many of whom intern/work for PRC IC companies throughout. Current goal is increase to 30-50k IC talent with ~1000-2000 PhDs per year for high end R&D. This is why people meme about US export controls & sanctions helping/forcing PRC semi to coordinate and indigenize... previously we had 10+ years of PRC developing shit all human capital with respect to IC, instead throwing 100s of billions in Big Fund to buy (and squander) expensive western semi equipment. Now PRC forced to also spend 100s billions to develop stupendous amount of high end domestic talent so they can build everything themselves.
Ducky181@reddit
It is missing the point. I highlight headcount of companies commercializing EUV vs another high complexity industry like aerospace for simple scale comparison of bodies involved to rollout commercial EUV. Obviously institutional research was involved, he..............
These industries are not comparable. The early and current research within the EUV industry exclusively required the highest levels of expertise across multiple disciplines, including optics, laser physics, cryogenics, precision engineering, metrology, and materials science. While aviation is primary dictated towards large scale manufacturing.
If anything adding up all the headcount of research orgs would be smaller than broader commercial sector. At the end of the day if you want to commercialize sector you still need all the other heads. There's plenty for research, but 100,000s of fab monkeys and other technical staff for every.......................
In accordance with data from the UNESCO Institute for Statistics the total compilation of researchers in the western bloc involved with EUV is 5.3 million verses 2.2 million from China. While for the year 2000 it was 3.1 million. Still larger than China today.
Focusing on 25–64-year-old misses/skews picture, focus on 25-40 and new tertiary cohorts. PRC tertiary enrollment is split by generations, i.e. pre 00s workforce was like low single dig.....................
The link I provided above showed the education ratio for 20-34 years old. The S&E degrees awarded per 100,000 was 14 for China; 60 United States; and about 80 for the rest of the developed world. The developed world also has persistent unaccounted migration from the undeveloped world which is also going through an education transformation.
At current trendline, PRC is going to spit out 120m tertiary 60% of 200m births from past \~20 years in next \~20 years, with \~40% STEM, it m
Half of China's tertiary are from short-cycle tertiary education. Not full tertiary education. While the rest are predominately ISCED-6. In terms of pure numbers, India actually leads in graduates and is set to grow over the next 10-20 years over China in STEM. It's clearly not an accurate assessment given disparity in education facilities.
dirtyid@reddit
There are large scale and complex R&D efforts in aviation, especially leading edge like military aviation for engine cores taps into many of those fields. Before current western semi indy policy turned on money printer, aviation sector R&D better resourced and funded.
What are relevant #s involved in EUV research + related semi supply chain specifically? Because it's not 5 MILLION by order of magnitude. Don't know if you're just crawling chatbots to find irrelevant stats where OECD/west number is bigger like total tertiary academic educators (I assume inclusive of humanities). Sector relevant info / estimates UNESCO wouldn't even have (you'd have to dig from industry specific writing for figures), i.e. simply appealing to UNESCO stats and alleging 5M+ involved in EUV doesn't pass basic smell test, it's projecting faulty logic analysis to the point of not even being wrong.
Stats from 2020s outdated relative to PRC speed. Regardless, absolute values > ratios for strategic competition. It matters who has most talent. PRC is producing ~4x more STEM vs US now, around OECD level combined. Geo/domestic politics limit west ability to add talent, i.e. not likely for US (or semi bloc powers) to increase annual immigration by millions of STEM to keep pace with PRC aggregate talent production, regardless if developing world increasing total talent supply. They could, but current short/medium term reality based on what's in pipelines is PRC heading towards IC talent oversupply and western bloc undersupply. And western bloc being increasingly immigration unfriendly.
High end IC specific talent generation is, i.e. graduate/doctoral level is by definition >ISCED5 aka not short-cycle. PRC has relevant high-level IC talent production pipeline, when I say IC PhDs graduating in 4 years, I mean they're hamming full time PhDs, i.e. not leisuring part time, i.e. they're not doing lesser programs. More relevant, is PRC also has the industrial supply chains and ability to coordinate brains for industrial policy successfully, something LBH India isn't in same league in.
Look at China. Look at India. Look at relative PRC vs US catchup recent times, PRC rapidly moving up various global R&D / S&T / innovation indexes, including western pubs controlled for quality, moving up supply chains etc, reaching parity or now leading in various sectors. All in last 10 years. All because prior 10 hammered out relevant tertiary talent. Look at actual increasing PRC competitiveness. That pool of talent will be 2x-3x in next 20 years. Why dance around irrelevant UNESCO stats or ISCED definitions, when reality / accurate assessment has obvious PRC bias. We know the talent PRC generated and continues to generate can deliver results, fast. We know PRC adding them in large numbers. We know last 10 years of PRC going from fraction to parity aggregate talent numbers with US already enabled them to rapidly catch up across the board in spectrum of sectors. And we know we're in phase where PRC aggregate numbers exploding past parity with US, and in next 20 years their aggregate numbers will be multiple times more. It's pretty clear whose assessment is likely to be correct.
Ducky181@reddit
Don't know if you're just crawling chatbots to find irrelevant stats where OECD/west number is bigger like total tertiary academic educators (I assume inclusive of humanities).
I advise you don't attack and call the data bias from United Nations Educational, Scientific and Cultural Organization (UNESCO) just because it does not align with your narrative. In particular when you provided no counter source to disprove it.
What are relevant #s involved in EUV research + related semi supply chain specifically? Because it's not 5 MILLION by order of magnitude.
How is providing direct data showing the explicit number of researchers involved in research not evidence of academic and research capital. How else do you want to measure it, by number of lab coats?
Number of R&D researchers per million people, 2022
High end IC specific talent generation is graduate/doctoral level which by definition >ISCED5 aka not short-cycle. PRC has relevant high-level IC talent production pipeline, when I say IC PhDs graduating in 4 years.... Stats from 2020s outdated relative to PRC speed.......
You're not listening. The core point of my argument is that extreme ultraviolent lithography (EUV) research is conducted at the highest academic level, requiring a doctoral or master's human capital expertise and decades of experience rather than just a bachelor's degree.
China will take decades to reach this level of human capital that western nations involved with EUV development have based on the current trend in ISCED-7 and ISCED-8.
Do you have any data to disprove this?
iForgotMyOldAcc@reddit
I'm at awe at his rate of production, 1-2 long-form videos a week and they mostly sound very competently researched to a layman like me. A part of me does wonder if this is actually more of an indictment on his quality of research, cause YouTube does not have a shortage of slop producers churning out content.
pdp10@reddit (OP)
The proof is in the pudding. I'd be more convinced if they were shipping these things. There's a Zhaoxin KX-6640MA powered PC on Aliexpress right now, and you can get PRC-made RISC-Vs, and that's not nothing. But it's also not showing any signs of this acceleration that some are touting.
trololololo2137@reddit
zhaoxin x86 chips are mostly for legacy purposes. the real innovation is happening in ARM, MIPS and RISC-V
PhoBoChai@reddit
Their goal isn't for consumer devices, its to produce cutting edge supercomputing chips and AI chips, the likes being used by Huawei & Tencent to run their models, not reliance on NVIDIA.
okieboat@reddit
Cool that this sub is just a CCP propaganda sub. Hilarious.
Arachnapony@reddit
reality has a chinese bias
Helpdesk_Guy@reddit
I always find it funny, that many U.S Americans (and especially the Left) seems to have a real hatred and some furious Asiaphobia (Sinophobia, to be precise) and are somehow afraid of everything Chinese and (alleged) espionage from Far East in general.
The joke is, that it was mainly no other country but the U.S. itself, which deported expertise and manufacturing of goods towards Far East since the Sixties and turned it up to 11 in the Seventies, only to outsource fabrication and manufacturing in general in exchange for corporate America to reap unheard of profits – Outsourcing coding-skills onto Indian low-wagers (especially Californian companies did that), the electronic expertise to Japanese low-wage work-horses or semiconductor-expertise in general and basically everything else in the academic STEM-fields to everything Far East.
Who could've possibly foreseen that, after decades of out-sourcing expertise to Far East, said very knowledge ends up there?!
Reddit's infamously notorious Anti-Chinese sentiment can really get exhausting at times, not allowing to freely discuss technology unbiased.
Ducky181@reddit
Absurd. China bans all western media and social platforms and undertakes tens of billions in external propaganda campaigns annually and you dare attempt to argue that the west is bias.
We have no physical prototype of any workable EUV machine, or any detailed documentation of it's existence with the only evidence consists of photos of a small machine and claims on TikTok. Any other instance this would be consider nonsense, but due to mass propaganda even a small box is able to generate mass propaganda sentiment.
Anti-Western sentiment - Wikipedia
Helpdesk_Guy@reddit
Where did I claim that China is the good guy here? Did I refuted any Chinese propaganda?
You may have missed it in your baseless rant, but I was solely speaking on the fact, that China is basically the world's workbench, and how that came to pass by the West's very own hands and work of doing, by outsourcing expertise since the Seventies.
Strazdas1@reddit
Reality also has entropy and chaos bias.
ryanvsrobots@reddit
The alternative is to live in denial about China approaching semiconductor fabrication parity.
If anything this is the alarm bell and you're saying "there's no fire, it's propaganda" while you're being surrounded by smoke.
PhoBoChai@reddit
More like hardware enthusiasts who understand tech & geopolitics more than the average gamer.
We can see beyond the anti-China rhetoric coming from the US. Heck, we even know about the fact US gov funds billions of $ on anti-China propaganda in the West.
So we analyze these tech development with a lens that can see through the bullcrap.
PM_ME_UR_TOSTADAS@reddit
Fragile Murican ego lol
doscomputer@reddit
anonymous voting systems are always sold to the highest bidder
Helpdesk_Guy@reddit
My oh my … Even looking at the other side, is no considered treason, or what?
icygreengiant@reddit
If true ASML will finally get competition
MongolianBatman@reddit
I HAVE 3000 SIBLINGS.
Sevastous-of-Caria@reddit
I know it's an enormous catchup that China needed to nail. Economic,strategic capital and scientific challenge never seen since the economic comeback of US economy and never tried since after US dominated post cold war economics
But all US was betting on is that ASML expertise and worker shortage on photolithography worldwide would be enough for China to be starved out on RnD. But in capitalism you can solve problems by throwing money and attracting expertise. And China throwing both money and political will to train themselves for EUV. The last castle left in technological parity. I'm not commenting how much talent China attracted to learn how to build a foundation cause thats for the future us to find out. But the 10 year catchup estimating people give is in for a surprise imo.
HelloItMeMort@reddit
It’s absolutely embarrassing how much we’ve fallen behind. The last time America was “great” we put the best minds together and created the most unconscionable weapon ever created. If you want to make America “great again” you have to invest in science and tech just like we did back then, and then establish utter dominance of Putin and the Middle East. Democracy either stands alone, or we take fascists and oligarchs down with us
sharkyzarous@reddit
if the greatness of America is a WMD attack, i beg to god never let America be great again.
Thorusss@reddit
The physical weapons themselves might not. But the newest chip for AI might decide future wars with data analysis, autonomous weapons and R&D.
Big shame if this amazing tech Ressource would go to that. But there are really cool alternatives, like Deepmind moving from Protein simulation and development to whole cells.
Qaxar@reddit
It's crazy how jingoistic Americans are.
PhoBoChai@reddit
The combination of both arrogance & ignorance is exemplified in these typical Americans.
pdp10@reddit (OP)
I've heard of a big TCP/IP network that seems pretty great. It can be accessed via submarine optical cables running at a few dozen terabits per second, via mobile satellite terminal, or maybe with the battery-powered gadget in your pocket using one of two different wireless protocols and infrastructures. Perhaps someday you'll use these Elbonian innovations to communicate with like-minded persons.
justbecauseyoumademe@reddit
5 people were responsible for making the Internet as we know it,
3 of them were not American but British and Polish.
Vint Cerf and Bob Kahn did create TCP/IP but not after Paul Baran and Donald Davies created packets as we know it and Tim Berners-Lee made the WWW
Tasty-Traffic-680@reddit
Nope, it's all blue jeans and plastic cheese. /s
Kougar@reddit
It's always faster to buy or steal or deduce existing knowledge than it is to be the entity developing new knowledge & processes. It's one of the weaknesses of being the first-mover or the dominant player in any market, they will have to overcome obstacles carving out a path that everyone else can easily follow behind, or spend the most to maintain a lead while others can spend less for the same results.
The US's strategy seemed far more viable when nodes were on two year cycles. But now that nodes are taking three years and progressing toward four it is going to give far more time for other interested parties to catch up, or at least close the gap.
dirtyid@reddit
There's also consideration first-mover made the right call, i.e. PRC is also developing alternative EUV methods like SSMB/synchrotron which may leapfrog existing tech. IIRC, a few years ago, there was interview with ASML engineer who drove their EUV efforts. The TLDR is on paper / initial analysis - building synchrotron expensive, so they opted for plasma. But then it was many paragraphs of unplanned technical challenges. Whether engineering around said challenges piled on cost is for more technical people to speculate, but what we know is ASML EUV cost 200M USD each, next gen pegged 400m. Which is about the cost of multiple small or single medium size synchrotron facility in west. For reference, PRC's latest large scale synchrotron facility HEPS is 1.3km circumference, with 60+ beam lines for 90 tool/experiment stations, for ~700M USD. The conceptual design for SSMB right now is a much smaller ring designed to power multiple tools at 1 kW. It's not just node timelines - it's questionable if ASML route is going to remain dominant in terms of economics, only that it's the only working method despite sunk cost.
BauCaneBau@reddit
Be carefull and not fall in easy trap of cost comparison. ASML EUV machine is not only “making EUV”. The surce of radiation is the less expensive stuff, even if you use a syncrotron Line as surce you will need all the optics and control to use it for photolitography, that is where the cost explode. Then you need a foundry to be built around the syncrotron etc. I am not saying is not feasible or stupid, do not get me wrong. I am just saying is not 1x ASML = 2x syncrotron, so use syncrotron.
Priximus@reddit
Isn't one of the arguments for synchrotron being that the beam quality higher so there's less challenge to optically optimize the final beam? There's also no tin to blast around so you dont have to constantly replace these very expensive mirrors. But obviously there are many thing in the pipeline that need to be mastered as well.
BauCaneBau@reddit
Yes, that is true. As I mention I am not saying it is wrong going in that direction. I am just saying to not simplify too much the math related with cost. I am pretty sure that as soon as the use of syncrotron is much convenient than the use of ASML equipments (or say similar alternatives) foundry will move accordingly.
dirtyid@reddit
My understanding is the selling point of SSMB according to Tsinghua is price per tool is suppose to be lower because beam manipulation on synchrotrons mature/explored technology. Another comment below mentioned, less complex mirrors when strong collimated beam from source. IIRC original napkin math during EUV exploration assumed synchrotron more expensive because assumption was EUV as infra cost more than EUV as tools that can be shipped on planes, but PRC very good at driving down infra costs. That said, PRC also seems very good at driving down tool costs, see recent account their domestic DUV supposedly 1/30th price of ASML. Whether to be believed or not, semi is a capex heavy game, hence imo economics will be large determiner of node timelines/adoption going forward.
Exist50@reddit
Increasingly the "strategy", to the extent it can be called one, is to assume you can maintain technological dominance solely by reactively crippling competitors instead of investing in cultivating talent domestically. Pretty obvious where what ends.
FourDimensionalTaco@reddit
And with what is going on in the US right now, this will only intensify. The brain drain that is already happening will embolden China further. It will help Europe to an extent as well, but I think China will be the primary winner.
Typical-Yogurt-1992@reddit
In the short term, China might not even need to develop their own EUV machines, DUV machines could suffice. There was a DUV-based version of TSMC's 5nm process. The efficiency gain between TSMC's 5nm and 3nm, the most advanced mass-production node on the planet, is only roughly 25%. This difference can be compensated for by a slight increase in chip size and a minor reduction in clock speed.
pdp10@reddit (OP)
Agreed that a 5nm node would be competitive for a lot of applications, for a long time to come. It's under-appreciated here how useful legacy nodes can be, because there's a bias here to full-power, latest gaming-suitable CPUs.
An almost-extreme example is the Skywater 130nm process node with open-source PDK. 130nm, in general, was the node of the Pentium III, first used in 2000-2001. It's still probably better than what's used for a lot of lower-end microcontrollers today.
Aleblanco1987@reddit
really interesting video, asianometry always bring quality content