8 ms·
I strongly doubt anyone will ever catch up with TSMC or Samsung at this point. Intel might still have a chance, but it seems to be waning. Semiconductor manufa
by bob1029 6y ago
I strongly doubt anyone will ever catch up with TSMC or Samsung at this point. Intel might still have a chance, but it seems to be waning.
Semiconductor manufacturing is very unique in that it is easily the hardest thing we do right now. Even if China snipes all the talent, tools, and process designs, they still need mountains of precise parameters that are unique to the specific installation of each tool at the specific site. Tuning these parameters relative to the process technology can take years and requires the kind of long term, open collaboration & innovation sessions that the CCP seems fundamentally incompatible with.
There are simply so many reasons this type of thing is virtually impossible to steal. If you were to ever physically tour a fully-automated semiconductor factory & experience the depth of the system engineering required to tie everything together, you would realize there is nothing as complex as that system anywhere else in the discipline of engineering.
- samfisher83 6y agoAt one point Intel was way ahead of everyone else. Saying "anyone will ever catch up with TSMC or Samsung" is a very broad think to say.
- oblio 6y ago1. That's a bold assertion you make about the CCP. I don't like them, but they have proven quite resourceful. 2. You say this as someone knowledgeable with other high tech industries? Here in IT we tend to pat ourselves on the back quite a bit, how do you know you're not falling into the same trap?
- gnusty_gnurc 6y ago> but they have proven quite resourceful when it involves oppressing people...truly generative processes? nah...it gets done when they start utilizing markets but otherwise it's fundamentally doomed
- parsadotsh 6y ago> when they start utilizing markets I don't think you're familiar with the modern Chinese economy...
- ummonk 6y agoNot to mention markets don’t really affect how TMSC or Samsung are organized internally.
- gnusty_gnurc 6y agothat's what I'm saying: chinese prosperity is largely happening when they use markets, but the popular idea that there's some unstoppable, ultra-powerful chimera of totalitarianism and capitalism - I think there's a huge burden of proof that it isn't fundamentally still organizational folly like much of totalitarian states/communism
- sudosysgen 6y agoEven the USSR under Lenin utilized markets in some cases. Communist ideology isn't fundamentally opposed to markets.
- coreblocks 6y agoThe Chinese economy is hardly communist at this stage, it's a mix of socialism and state sponsored capitalism.
- sudosysgen 6y agoAgreed.
- kasperni 6y ago> you would realize there is nothing as complex as that system anywhere else in the discipline of engineering. I've heard people say the same about the ITER reactor being assembled.
- bob1029 6y agoAs far as I am aware, ITER has no aspects of construction or operation which require manufacturing tolerances measured in nanometers.
- coreblocks 6y agoExcept for their computers :)
- Panoramix 6y agoIt depends on how you want to compare them. You could look at it from a funding perspective. The total cost of ITER through 2013-2025 is expected to be 20-65 billion. Samsung just revealed they are investing 50 billion to build an EUV foundry. This is on top of all the knowledge and assets they already have. It does not include the R&D to get to the point where you can just buy EUV machines, etchers, atomic layer deposition, etc.
- f00zz 6y agoPeople said similar things about Japan before they took the lead in DRAM development back in the 1980s.
- Nokinside 6y agoAnd then South Korea took the lead in the 1990s (SK Hynix and Samsung).
- Nokinside 6y agoWe are not talking about stealing here. They are hiring senior people with the know-how, both engineers and executives, by offering them huge compensation packages. They are hiring from both Taiwan and South Korea btw. Huawei does the same in mobile. Their know-how comes from hiring good people. They have R&D centers outside China where they employ people they hired from Qualcomm, Nokia, Ericsson and others to design their new 5G SoC and mobile technologies.
- mschuster91 6y ago> They have R&D centers outside China where they employ people they hired from Qualcomm, Nokia, Ericsson and others to design their new 5G SoC and mobile technologies. Then it is time to kick Huawei out. Dissolve or nationalize their local entities, ban banks from dealing with their money. China does not allow foreign companies to operate in China, so why should Western countries do the same? It's either a level field for all - or no field at all. The Western world has appeased China for too long in the hope China would democratize itself. This has evidently failed, it's time to pull the plug.
- nicoburns 6y agoNobody plays by the "level playing field" rules. The US likes to pretend it does, but as soon as there's an industry where it's not competitive (like agriculture for example), that goes out of the window. And IMO for good reason. Unfettered markets work in some cases. But in many cases they have lead to "winner takes all scenarios" where a single player (or 2/3 players) become dominant and the market is no longer competitive. We need a system that works against this over the long term, and some level of protectionism seems like a decent way of doing that.
- valuearb 6y agoName one. You are just repeating protectionist propaganda to justify overcharging customers to benefit protected corporations.
- 6y ago
- hn_check 6y ago"I strongly doubt anyone will ever catch up with TSMC or Samsung at this point." What timeline are we talking about here? 1 year? 5 years? Forever? Yes, of course people will "Catch up" to Samsung and TSMC, both of which are relatively recent leaders in this field. There are are loads of foundries around the world, all of which have virtually equivalent level of complexity, but it is small (literally small) advances that give one foundry or another a competitive lead. And those small advances are certainly stealable, quite aside from independent orthogonal advances.
- adriancr 6y agoImagine it's 1946 and someone would have made a similar statement regarding nuclear bombs and Russia... Where there are resources problems will get solved.
- peteradio 6y agoScientists were treated as slaves in soviet union projects... hope the tsmc engineers read the fine print. EDIT with source: https://www.theatlantic.com/science/archive/2017/05/soviet-science-stalin/525576/ https://www.theatlantic.com/science/archive/2017/05/soviet-s...
- FpUser 6y agoThere was a period when many scientists and engineers were convicted on some bogus anti-soviet charges and jailed. Then Stalin realized that many of those were world class people with unique talents and knowledge and they were allowed to work in their area of expertise but in jail like but somewhat softer conditions. In 1953 after Stalin's death the practice was eliminated (took few years).
- paganel 6y ago> open collaboration & innovation sessions that the CCP seems fundamentally incompatible with. The Soviet Union had managed to surpass the US in constructing nuclear bombs in a relatively short matter of time (I'd say that by the 1960s they were already in front), even though they were not as "open" and they didn't have as many Nobel laureates working for them as the US had.
- jeswin 6y agoThat's not accurate [1]. In the 60s and for most of the 70s, the Soviets were far behind in bombs and delivery systems. Many sources say that the US vastly overestimated the size of the Soviet arsenal at that time - and hence was very cautious. Some sources say that the US did have a fairly decent estimate, but they overstated the threat to get more funding. [1]: https://en.m.wikipedia.org/wiki/Historical_nuclear_weapons_stockpiles_and_nuclear_tests_by_country https://en.m.wikipedia.org/wiki/Historical_nuclear_weapons_s...
- paganel 6y agoI was judging things based on the Tsar Bomb, mostly.
- jjoonathan 6y agoOnce you know how to chain thermonuclear stages, you can obtain arbitrarily high yield without further advances in technology. The Americans pioneered this technique, they just didn't bother building as big of a chain. At a certain point bigger yields are less desirable than more, smaller yields. Picture an enormous spherical fireball, then picture chopping it up and using the "useless" slices that were at high/negative altitude to instead cover more ground. Once it became clear that they were past the regime of yield scaling, the Americans moved on to developing smaller warheads that could do just that: pepper targets and deliver more devastation per pound than a corresponding single warhead of high yield. Vanity metrics are always more important to the second place player, and that's what Tsar Bomba was.
- _Understated_ 6y agoThey probably don't need to "catch up" in the traditional sense. By that, I mean that if someone was starting from nothing and trying to make the same stuff and enter the same market as them, then yes, they will have a massive uphill battle and would likely fail without legislative help. If they did something different but in the same space then that's another matter. Remember, a short while ago it was all about x86 where Intel had the distinct advantage. AMD cut some of that down and is now gaining rapidly on them. Nowadays it's all about ARM but in 10 years time it may be something else and if you get in early enough and your predictions are correct then who knows... One thing I am convinced of is that if Microsoft embrace ARM wholeheartedly (I mean port everything across and let x86 go to hell) then Intel (and AMD for that matter) are in trouble.
- jjoonathan 6y agoWhat? We're talking about fabs. AMD didn't catch up, they dropped out. TSMC caught up.
- growlist 6y ago> requires the kind of long term, open collaboration & innovation sessions that the CCP seems fundamentally incompatible with I always see this but it just seems like such a reach - Nazi Germany didn't seem to do too badly on the innovation front, did it?
- eunos 6y agoThe goal is not exclusively to conquer the consumer chip industry. Ability to manufacture strategic chip industry is also very important. Remember the sanctioned big Chinese facial recognition companies? They do not really need 5nm chip for their product, I think 14nm can also work (albeit performance penalty). Also, PLA, if they ever need to manufacture AI chips for whatever reason, they need to manufacture them in mainland for supply chain security.
- mensetmanusman 6y agoTSMC’s advantage is an order of magnitude more engineers of high quality handling yield issues, the machines used to make cutting edge transistors is the same. With China doing this, TSMC’s advantage will decrease.
- jandrese 6y agoIt wasn't that long ago (10 or 15 years?) that Intel seemed to have an insurmountable lead on process upgrades. Anybody can fall behind if they get complacent or have some bad luck.
- jacquesm 6y agoOf course they will catch up. It's impossible to maintain a lead forever in a field that is slowing down as a whole. It will take some time, but they will definitely get there at some point. There is no magic sauce that makes TSMC or Samsung better at this than China can ever be. So better plan for that.
- contingencies 6y agoAlso, China gets things done incredibly quickly when state motivation aligns with technology. See: three gorges dam, national fiber optic network, mobile coverage, HK-Zhuhai-Macau bridge, 100% contactless mobile payment and instant messaging penetration, high speed trains, electric vehicles, etc. All of these examples are from the last 20 years.
- jacquesm 6y agoThat's just about the only advantage a planned economy has: reaction speed. All it takes is one decision to mobilize massive resources.
- contingencies 6y agoI would restate that as a more general advantage, possibly only when operating at China's immense scale: relative cognitive freedom in creating solutions due to a broader pool of options. Because the sum total of available regulatory, fiscal, environmental and other resources necessarily exceed that of private actors, courses of action that would be untenable in other systems or require difficult and expensive public-private-partnerships are certainly feasible. This is why, for example, the 'super app' did not emerge in a western economy and China is one of the only countries improving its forest cover.
- ecf 6y agoI was having this same thought the other day when looking at Herman Miller chairs and scratching my head. How can a CPU, with physics that are more complex than most can even barely grasp, cost 1/3 of what an Embody retails for?
- magicsmoke 6y agoManufacturing volume
- jbay808 6y agoBecause photolithography is a wonderfully scalable manufacturing process!
- coreblocks 6y agoI doubt anyone will ever catch up to Intel, look they are already producing 14nm!