网友热论:台积电制造出全球最顶尖的芯片且无人可复制,这是怎么做到的“台湾将半导体制造业作为主导产业来扶持”“得益于苹果公司的资助”“半导体行业存在着许多相互依存关系,美国控制着其中几个关键瓶颈点”
ELI5: My understanding is that 1 company in Taiwan makes the greatest chips in the world and no one else can replicate them. How is that possible?译文简介
台积电(TSMC)是全球最大的半导体代工厂,全球大约60%的半导体和90%的先进半导体产自台湾。紧随其后的是三星晶圆厂(Samsung Foundry),占据了13%的市场份额,而联华电子(UMC)、格芯(GlobalFoundries)和中芯国际(SMIC)各占6%。此处提到的仅为为其他公司制造半导体芯片的代工厂,像英特尔这样大量生产自家芯片的企业并未被计入其中。
台积电,这家成立于1987年的公司是怎么做到的呢?
正文翻译
@DarkAlman
That's not entirely true
TSMC (Taiwan Semiconductor Manufacturing Company Limited) is one of the premiere chip manufacturing companies in the world but they are hardly the only one.
They are of note because they make everything from common chips used in everything from cars to cellphones to AMDs processors. Apple, Nvidia, AMD, and Qualcomm for example all use TSMC to manufacture their chips.
But Samsung (Korea) and Intel (US) also have chip foundries of that caliber.
TSMCs secret is just that they have large numbers of highly experienced people working for them and have developed very good processes. Samsung and Intel are similar in that regard, but TSMC is special in that the Taiwanese government has in a sense made semiconductor manufacturing the countries primary industry.
Other companies like Texas instruments can also make microchips and have facilities all over the US but can't make chips as complex as Computer microprocessors. Importantly though there's nothing stopping them from investing capital to build such a facility, they just don't want to. (Developing an in house microprocessor to compete with AMD and Intel at this point would require an outrageous investment)
Israel is also another big up and comer in chip manufacturing.
Why TSMC is of note is the Pandemic showed just how vulnerable the US and the West are to losing access to TSMC production. If China for example were to invade Taiwan it could be a really big problem for the economy and availability of these chips.
This is why TSMC is building a Chip Foundry in Texas Arizona right now so chip production can happen domestically in the US.
这并不完全正确
台积电(台湾半导体制造股份有限公司)是全球顶尖的芯片制造企业之一,但并非唯一的一家。
它们之所以引人注目,是因为它们生产各种芯片,这些芯片广泛应用在从汽车到手机乃至AMD处理器等众多领域。例如,苹果、英伟达、AMD和高通等公司均采用台积电来制造他们的芯片。
但三星(韩国)和英特尔(美国)也同样拥有同等水平的芯片代工厂。
台积电的成功秘诀在于,他们拥有大量高度专业的人员,并且已经开发出非常优秀的制造流程。在这方面,三星和英特尔与之类似,但台积电特别之处在于,台湾政府在某种意义上将半导体制造业作为主导产业来扶持。
此外,像德州仪器这样的公司也能制造微芯片,并在美国各地拥有设施,但他们目前无法制造出与计算机微处理器同样复杂的芯片。然而重要的是,并没有什么因素阻止他们投入资金来建设这种高级设施,只是他们目前没有这样的意愿(要自主研发能与AMD和英特尔竞争的微处理器,在当前阶段需要进行巨额投资)。
以色列也是芯片制造业中迅速崛起的重要一员。
台积电之所以值得注意,是因为疫情表明美国和西方国家在失去对台积电生产的接触时有多么脆弱。例如,如果台海战争爆发,这可能对经济和芯片供应产生重大影响。
这就是为什么台积电目前正计划在美国德克萨斯州亚利桑那州建设芯片代工厂,以便在美国本土实现芯片生产。
评论翻译
很赞 ( 4 )
收藏
To OP: We're also talking about bleeding edge technology - it's not generally available techniques like what you would have in older industries. And the investment necessary to get to the same level, when these large, existing actors in the field are still gradually reducing their level (from "5 nm" to "3 nm" and to "2 nm" these days) - there's simply not any easy way of "just starting up something on the same level".
You have to come out of the gate as one of the best in the world, and that's generally a very hard thing to do. (TSMC: 73k employees, Intel: 124k, Samsung Electronics: 270k, but Intel and Samsung has a far wider product range than TSMC).
They're on their own level for what they do, and there's only a few actors in the world that has the resources and know how to get to that level (Samsung and Intel generally as you already mentioned).
To OP:我们正在讨论的实际上是前沿技术,这并非像传统行业那样广泛应用的技术。当该领域的大型现有企业仍在逐步推进技术节点(从“5纳米”进步到“3纳米”,再到如今的“2纳米”),要达到与之相同的水平所需的投资是非常巨大的,并不存在任何简单的途径可以轻易地“从头开始就达到同一水准”。
新进入者必须在一开始就成为全球最顶尖的企业之一,而这通常是一项极其艰难的任务。(例如,台积电拥有7.3万名员工,英特尔拥有12.4万名员工,三星电子拥有27万名员工,尽管英特尔和三星的产品线比台积电更为广泛。)
这些企业在其所从事的领域内处于领先地位,世界上只有少数几家拥有足够资源和专业技术能够达到这一层次的企业(如您提到的三星和英特尔)。
And the investment necessary
yeh, TSMC spend something like $5.5billion on R&D per year, a 3-4nm fab cost something like $20 billion
研发投资是巨大的,台积电每年在研发上的投入约为55亿美元,而一座3-4纳米芯片工厂的成本大约为200亿美元。
Am I just brain broken by big company numbers or do these seem like reasonable enough numbers to not be the only factor.
20 - 30 billion dollars can't be the thing stopping China from getting better chips
是不是大公司的数字让我感到困惑了?这些数额看似合理,并不应该是阻碍中国获得更先进芯片的唯一因素。200亿至300亿美元不至于阻止中国芯片技术的进步吧?
You spend that every year for 40 years and get in line for the latest production tech. There's only one brand of photolithography machines in the world that can enable production of the most advanced chips.
You gotta build a relationship with ASML and denmark if you want those machines.
你需要连续40年每年投入这么多资金,才能跟上最新生产技术的步伐。世界上只有一种光刻机品牌能支持最先进芯片的生产。如果你想要这种机器,就必须与ASML公司及丹麦建立紧密的合作关系。
upvoting this for truthiness. if you don't have asml basically setting up a small office in your fab, you don't get the coolest toys. sauce: worked for coorstek ceramics who makes the "slider" (outrageous huge chunk of SiC) inside the EUV machine and the previous gen twinscan's ceramic air bearing surfaces (which were alumina ceramic, but still bleeding edge tols at temperture for the years we made them)
赞同这个观点,如果没有ASML公司在一定程度上将小型办公室设在你的晶圆厂里,你就无法获得最先进的设备。来源:曾在Coorstek Ceramics工作过,该公司生产的“滑块”(一块巨大的碳化硅部件)被应用于EUV机器内部,以及上一代双工件扫描机的陶瓷气浮表面(采用氧化铝陶瓷材料制造,在我们生产那几年也是温度条件下的尖端工具)。
Sounds like trying to get a Rolex. Like do I have to buy boatloads of their shittier tools before I’ll “get an opportunity” to buy their nm stuff?
感觉就像是试图购买劳力士手表。是不是要先买一大堆他们低端的产品,才能“有机会”购买到他们的纳米级设备呢?
Yes and no.
Say you were a CEO of a company that wants to start producing cutting-edge chips. No matter how much money you have, your country better have good relations to countries like the US, Netherlands. So if you're in a country that has bad relations (china), better hope you can steal the info on how to make the machine.
Even if your country is friendly with the US and the Netherlands, the machine costs around 200 million dollars , and there are only 140 in the world as of 2022.
Even if you somehow managed to buy the machine, and get it set up correctly, you still need to hire people who can work with this very specialised machine (again 140 as of 2022 in the entire world).
After you manage to find people who can use the machine and get the machine installed, you still need to iinvest alarge amount of money in R&D to research a chip to produce.
So before you even start producing, you are more than likely looking close to 1 billion $.
是的,也不是完全这样。
假如您是一家寻求生产尖端芯片公司的一位CEO,那么无论您拥有多少资金,您的国家必须与美国、荷兰等国保持良好关系。这意味着,如果您所在国家与其关系紧张(比如中国),则可能需要设法获取制造这种机器的相关信息。
即使贵国与美国和荷兰保持着友好关系,一台此类尖端光刻机的价格大约在2亿美元左右,且截至2022年全球仅有约140台。
就算您设法购买到了这台机器,并成功安装调试,接下来还需要招聘能够操作这种高度专业设备的技术人员(同样截至2022年,全球范围内能熟练使用这类设备的人才极其有限)。
当您成功找到合适的操作员并完成机器安装后,仍需投入大量资金进行研发,以研究要生产的芯片设计。
综上所述,在真正开始生产之前,整个过程的投入很可能接近甚至超过10亿美元。因此,进入顶尖芯片制造领域不仅限于资金问题,还涉及国际关系、稀缺设备采购、专业技术人才引进以及高额的研发投资等多个复杂层面。
The fabs cost billions of dollars, and the price of the chip you sell decreases dramatically over time, if I recall correctly you get back about 50% of your investment in the first two years -- but only if you're first to the new bleeding edge transistor size. If you're six months late to the tech, you will never pay off the fab.
And that's the challenge for someone like AMD, who know what the hell they're doing. These are the most absurdly complex manufacturing processes ever created, and each generation of chips they have to reinvent it. You can't just follow the process from the last generation and make it smaller. It is absurdly difficult to do.
芯片工厂的建设成本高达数十亿美元,而且您所出售的芯片价格会随着时间推移急剧下降。据我所知,只有在率先采用最新前沿的晶体管尺寸时,大约能在前两年收回约一定比例的投资。如果技术上落后六个月,那么这座晶圆厂可能永远都无法回本。
对于像AMD这样深知其中门道的企业来说,这是一项巨大的挑战。半导体制造工艺是迄今为止人类创造的最为复杂的过程,每一代芯片都需要从头重新研发。你无法简单地沿用上一代工艺并将其缩小规模,实现这一目标的难度极其之高。因此,对于AMD这样的企业而言,在每个新世代节点的竞争中保持领先地位至关重要,否则将面临难以回收巨额投资的风险。
Another factor: There is currently only one manufacturer (ASML) for these machines and they are based in the Netherlands. There is no second player. ASML also has an export embargo for China, so even if they wanted to they couldn't sell China any of these machines.
Additionally these machines are very complicated. Somebody please correct me if I'm wrong, but a few years back China actually managed to somehow get ahold of one.
They tried to reverse engineer it and broke it when they reassembled it. They couldn't get it to work anymore and nothing of substance was learned, partly because of the insane precision required in the processes.
另一个重要因素是,目前这类机器仅有一家供应商——ASML公司,且总部位于荷兰,没有其他竞争对手。此外,ASML对中国实施了出口禁运,因此即使他们有意出售,也无法向中国提供任何此类机器。
另外,这些机器极其复杂。如果我记忆有误,请纠正:几年前,中国实际上设法得到了一台这样的机器。他们试图逆向工程仿制这台机器,在重新组装时却损坏了它,导致其无法再次正常工作。部分原因在于该过程中所需达到的精度极高,因此在尝试拆解和重组的过程中,并未学到实质性的技术内容。
ASML sends engineers on site (something like or so for its most advanced machinery) to Taiwan and those engineers remain on-site at the fabs for the entire life of the ASML machine.
In other words, it’s not just having the precision to reassemble it, but maintaining it and operating it requires an extreme
对于最尖端的设备,ASML会派遣工程师到现场(比如台湾晶圆厂),大约会有几十名工程师常驻在工厂中,负责整个ASML机器在其寿命周期内的运作。换句话说,这不仅仅是拥有高精度重新组装的能力,保持机器运行并进行维护也需要极高的理解和专业技术支持,而这并不是简单通过投入资金和人力就能迅速掌握的。
Its not just money. That helps, but there much more and it takes time.
Starting early, having a specialised workforce, recruitment and training systems, deals with companies that provide essential tools and material, investing that kind of money yearly and being backed by the right companies and goverments.
These companies also rely on some of the most complex supply chaines in the world. China is said to be generations behind and have very little outside help. Better than the russians, but a long way from parity.
这不仅仅是资金的问题。虽然资金确实有帮助,但还需要更多,并且需要时间。要从早期开始布局,培养一支专业的员工队伍,建立有效的招聘和培训体系,与提供关键工具和材料的公司达成协议,每年进行大规模投资,并得到合适的企业和政府支持。这些芯片制造企业依赖于世界上最复杂的供应链。据说中国在这一领域落后了几代,而且能得到的外部援助非常有限,尽管其情况优于俄罗斯,但仍需经历漫长的道路才能实现并驾齐驱。
There are numerous interdependencies in the semi industry, and there are a couple of choke points that the US controls. One of them is chip design software. TSMC builds the chips, but the design of those chips is done primarily on US-based software providers. Second, ASML (mentioned above) assembles the most cutting edge lithography tools in the world. But they source portions of their equipment from an ASML subsidiary in San Diego called “Cymer”.
The US controls the flow of chip design software and certain tools (like those made by Cymer) to China. Further, ASML (a Netherlands company) has agreed to back US embargoes to china. But even if ASML wanted to provide tools to china, those tools that come from the US would be off limits and therefore cripple the process.
You can throw hundreds of billions of dollars at an industry, but if you are locked out from fundamental components of the industry then no amount of money can solve it.
China would have to develop ALL of its own tools for the entire process. It is not only cost prohibitive, but it’s technically prohibitive - at least for the near term.
半导体行业中存在着许多相互依存关系,美国控制着其中几个关键瓶颈点。首先,芯片设计软件方面,台积电虽然负责制造芯片,但芯片的设计主要依赖于美国的软件供应商。其次,提到的ASML公司组装了世界上最先进的光刻工具,但他们部分设备来自位于圣地亚哥的ASML子公司——Cymer。
美国控制着流向中国的芯片设计软件以及某些特定工具(比如Cymer生产的工具)的出口。此外,荷兰公司ASML也同意支持对中国的禁运措施。即使ASML有意向中国提供工具,由于来自美国的那些工具受到限制,这将严重阻碍整个生产流程。
你可以向一个行业投入数千亿的资金,但如果被排除在该行业的核心组件之外,再多的钱也无法解决这个问题。中国必须为整个芯片制造过程自主研发所有的工具。这不仅成本高昂,而且在技术上也具有挑战性——至少在短期内是如此。
IBM was really good but they went fabless some years ago.
IBM曾经非常出色,但几年前他们决定转型为无晶圆厂模式(fabless)。
IBM went fabless because they started falling behind in the leading edge
IBM之所以选择成为无晶圆厂模式,是因为在尖端技术领域他们开始落后于竞争对手。
Same with AMD's spin-off Global Foundries. Global Foundries basically came to the conclusion that they couldn't compete at the bleeding edge anymore and AMD had to go to TSMC while Global Foundries makes things on larger process size nodes.
AMD剥离出的子公司Global Foundries也面临类似情况。Global Foundries最终得出结论,它们无法在最前沿技术上与对手竞争,因此AMD转向了台积电,而Global Foundries则专注于更大工艺尺寸节点的产品制造。
TSMC's emergence as the pre-eminent advanced chip manufacturer was largely funded by Apple. They wanted to get away from Samsung, who was also their biggest phone competitor but supplied all the CPUs for all the early iPhones. So they basically bankrolled TSMC's most advanced fabs in return for certain exclusivity. The amount of money Apple spends on advanced chips is mind boggling, and if they switched wholesale to someone else, those folks would probably surpass TSMC in 5 years.
台积电成为尖端芯片制造领头羊的地位在很大程度上得益于苹果公司的资助。苹果希望摆脱三星,因为三星既是他们最大的手机竞争对手,同时也是早期所有iPhone CPU的供应商。因此,苹果基本上为台积电最尖端晶圆厂提供了资金支持,以换取一定的独家供应权。苹果在先进芯片上的花费数额惊人,如果他们将订单全部转给其他厂商,那么这些厂商很可能在5年内超越台积电。
At this point, I’d imagine TSMC is basically self sustaining. If Apple stopped bankrolling them, someone else like Nvidia or AMD would. Even Intel is buying fab capacity from TSMC now. But yes, it takes a lot of investment over a long period of time to get to the bleeding edge and stay there.
现在来看,台积电已经基本实现了自我维持。即便苹果停止对其资助,像英伟达或AMD这样的公司也会接手。甚至英特尔现在也开始从台积电购买晶圆厂产能。但是,要达到并保持技术前沿确实需要长期大量的投资。
from "5 nm" to "3 nm" and to "2 nm"
In a couple of years it will be measured in angstroms.
从“5纳米”发展到“3纳米”,再到“2纳米”,
在未来几年内,这一测量单位可能会缩小至埃(angstroms)。
Nah, we'll have to go back to thicker chips and the tech world will lose 3-5 years of hardware progress.
Inconvenient, sure, but we're not going back to steam machine.
不,我们可能不得不退回到使用更厚的芯片技术,科技界会因此损失3-5年的硬件进步成果。虽然确实很不方便,但我们不至于倒退回蒸汽机时代。
America is going to spite the entire world because they dont want China controlling some chip factories? Sounds pretty selfish to me
美国为了不让中国控制某些芯片工厂,而打算牺牲全世界的利益?在我看来这确实显得相当自私。
No, China wouldn't risk damaging such a valuable industry
It's far more likely that they would nationalize TSMC and disrupt production to the West.
不,中国不会冒着破坏如此有价值的产业的风险。
更有可能的是他们会将台积电收归国有并扰乱对西方的生产。
They absolutely would be blown to smithereens either by
A.) China, because of carpet bombing
B.) The West, to prevent China from getting the technology
C.) Taiwan, knowing they lost to make sure China doesn't get it
它们绝对会被彻底摧毁,方法可能有以下几种:
A)中国可能采取地毯式轰炸来摧毁这些设施;
B)西方国家可能采取行动,以防止中国获得相关技术;
C)台湾可能已经有所准备,确保不让中国获得这些设施。
No, China wouldn't risk damaging such a valuable industry
They're not saying they would be blown up by China, they would be blown up by the West to stopping China getting access to the technology
Hence why a foundry is being constructed in Texas
不,中国不会冒着破坏如此有价值的产业的风险。
他们并不是在说设备会被中国摧毁,而是会被西方国家炸毁,以阻止中国获取技术。
因此,这也是为什么在德克萨斯正在建设一个铸造厂的原因。
I would think the more important resource would be the k highly skilled workers with knowledge and experience being the best in the world at what they do. The consensus seems to be that physically building the facilities isn’t the hardest part. China would want to secure those skilled and experienced workers and “motivate” them to industry secrets. Then they could replicate the facilities in mainland china, and curb any motivation the west has for reprisals
我认为更重要的资源是那些在芯片制造领域拥有世界顶级知识和经验的高技能工人。普遍认为,建设这些设施并非最困难的部分。中国希望确保这些技术娴熟且经验丰富的工人,并对他们进行“激励”以分享行业知识。然后他们可以在中国大陆复制这些设施,并遏制西方对报复的动力。
Would they be willing to do it correctly, I imagine they would be as uncooperative as they could reasonably be.
You can kiss further innovation goodbye. I don't think you can beat or threaten that out of a worker.
But the destroying to reduce their appeal to the West is an interesting idea that would unfortunately probably work.
这些工人是否会愿意配合,我推测他们会尽可能地不合作。你可以预见,在这种情况下,进一步的创新将很难实现。我认为无法通过威胁或强迫的方式来从工人那里获取技术。
但摧毁工厂以降低其对西方吸引力的想法虽然不幸,却可能奏效。
It's been reported that Taiwan has planned the destruction of the plants as a step in their defense of the island.
据报道,台湾已经计划在保卫该岛的过程中,将破坏这些工厂作为防御措施之一。
原创翻译:龙腾网 https://www.ltaaa.cn 转载请注明出处
*TSMC is currently building its North American fab plant in Phoenix, AZ
台积电目前正在美国亚利桑那州凤凰城建设其北美晶圆厂。
原创翻译:龙腾网 https://www.ltaaa.cn 转载请注明出处
Also in Dresden, Germany in the coming years. Deal is already signed.
在未来几年内,台积电还将在德国德累斯顿建厂,协议已经签署。
TSMC is still at the bleeding edge of EUV. TSMC facilities in the US are always made to be at least one node behind those in Taiwan. Taiwan made a huge gamble by putting pretty much everything into chip fab, and it worked. It's also a major national defense strategy.
台积电在极紫外光刻(EUV)技术上仍处于最前沿。在美国的台积电工厂通常会比台湾本土的工厂落后至少一个工艺节点。台湾通过几乎全力投入芯片制造,进行了一场巨大的赌博,并且成功了。这也成为了其重要的国家防御策略之一。
You forgot the part where they're making the smallest transistors with high yield that nobody else is capable of.
您遗漏了关键的一点,即台积电能够以高良品率制造出其他厂商无法企及的最小尺寸晶体管。
Intel isn't really in the same league as TSMC. They used to not be direct competitors as Intel manufactures chips for itself primarily, only recently trying to become a foundry as well and manufacture for external customers. But Intel is also multiple generations behind TSMC and in general losing to AMD and to Android. Apple even switched from Intel in its macs to an in-house chip manufactured by TSMC.
Even Samsung is still a few years behind TSMC in the leading edge. They claim they can catch up, but so far the evidence isn't there.
英特尔实际上并不属于与台积电同一级别的竞争者。过去,由于英特尔主要为自己制造芯片,并未将自己定位为晶圆代工厂,直到最近才尝试成为代工厂并开始为外部客户提供制造服务。但即便如此,英特尔在技术上仍落后台积电多代,在市场上正逐渐输给AMD和搭载高通/骁龙处理器的安卓设备。苹果甚至已在其Mac电脑产品线上从英特尔转向自家设计、由台积电生产的芯片。
即使像三星这样的巨头,在最前沿的技术节点上也仍然落后台积电几年时间。虽然他们声称能够迎头赶上,但目前还没有足够的证据表明这一点。
I mis-typed it. Intel never really gained a foothold in mobile, so they're not really losing to Android because they never had any to begin with.
AMD is the primary gainer at Intel's expense both in PCs and also in the datacenter. But given how much Mac has grown, Apple switching to their in-house chip was a fairly big loss for Intel, although that's also been a few years now at this point.
英特尔实际上从未在移动市场取得稳固地位,所以他们并没有真正意义上因为安卓而失去市场份额,因为他们从一开始就没有在这个领域拥有多少份额。AMD是主要从英特尔手中夺走市场份额的一方,无论是在个人电脑还是数据中心领域都取得了显著增长。尽管苹果转向自家芯片已经过去几年了,但鉴于Mac业务的大幅增长,苹果转用自研芯片对英特尔而言确实是一个重大损失。
AMD is quite big and has money, why don’t they invest in their own foundry?
Seems a bit strange to be so dependent on one supplier
AMD规模很大且财力雄厚,为什么它们不投资建立自己的晶圆厂?如此依赖单一供应商似乎有些奇怪。
AMD sold its stake in Global Foundries which previously made their chips.
AMD decided to focus on design, RND, and testing instead of manufacture.
That way all the costs and risks associated with fabricating chips is carried by someone else instead.
AMD之前曾持有Global Foundries股份,后者曾为AMD生产芯片。但后来AMD决定将重心放在设计、研发和测试上,而不是制造环节。这样,与芯片制造相关的所有成本和风险都由其他公司承担。
This is why TSMC is building a Chip Foundry in Texas Arizona right now so chip production can happen domestically in the US.
I think there are a few chip fabs being built in the US now. The government realized this was a critical weakness where an enemy could hobble the whole US. Also, companies like Apple and Intel have realized that outsourcing everything to overseas suppliers, even if less expensive, comes with its own set of problems so are starting to bring some production back to the US.
But, it takes years to get these factories online. It is a super complex task.
在建中。政府已经意识到这是一个关键弱点,敌人可能会借此阻碍整个美国的发展。同时,像苹果和英特尔这样的公司也意识到即使外包给海外供应商成本较低,但也带来了一系列问题,所以开始将部分生产活动带回美国本土。但是,建立这些工厂并投入运营需要数年时间,是一项极其复杂的任务。
原创翻译:龙腾网 https://www.ltaaa.cn 转载请注明出处
It's not that other companies can't replicate them, it's that the amount of startup money needed to get a top of the line chip manufacturing facility running is immense. In the s, the Taiwanese government put up a large amount of that startup money believing that it would pay off later, and it did. Lots of companies decided they'd rather pay Taiwan's company to manufacture their chips for them instead of build their own facilities.
问题并非其他公司无法复制这种模式,而是要启动一家顶级芯片制造工厂所需的初始资金极其庞大。上世纪80年代,台湾政府提供了大量启动资金,相信此举将在未来得到回报,并且确实取得了成功。许多公司宁愿支付费用让台湾的台积电为其代工生产芯片,而不是自建设施。
原创翻译:龙腾网 https://www.ltaaa.cn 转载请注明出处
You also can't replicate them either. At this point there are trade secrets, billions of R&D, and general employee expertise that make it almost impossible for a newcomer to compete with TSMC and Samsung at the leading edge.
Intel had all the resources in the world and was competitive until maybe -ish years ago when it started to fall behind TSMC. It's not entirely clear why they weren't able to achieve the same success as TSMC, whether it's a leadership issue or maybe something cultural.
除了高昂的启动资金外,实际上你也不能轻易复制台积电或三星在尖端领域的地位。目前存在大量的商业秘密、数十亿美元的研发投入以及员工专业知识,使得新进入者几乎不可能与台积电和三星相抗衡。英特尔拥有全世界的资源,在几年前还能与台积电保持竞争态势,但后来逐渐落后于台积电。尚不清楚为何英特尔未能取得与台积电同样的成功,这可能是由于领导层问题,或者是文化层面的某些因素所导致。
Semi conductors manufacturing is the most difficult mass production on the planet.
It requires tremendous amounts of investment to build and operate a fab. Each successive semi-conductor generation (often called a "node") is getting more difficult and expensive to design and produce, with the possibility of it being delayed / faulty increasing (this happened with Intel's nm process). As a result, a lot of companies dropped out, and at the bleeding edge we're left with main players: Intel, Samsung, and TSMC.
Currently, TSMC has the best process - right now it's being used for the new iPhone Pro and Apple M series and should expand to more products from AMD and and Nvidia by the end of the year.
There's also one main company left producing the machinery to make these highly advanced nodes economical: ASML. These machines are very complex, expensive, and limited in quantity per year.
TSMCs leadership role isn't a guarantee. Intel and Samsung aren't THAT far behind, and Intel in the last few years has dumped tremendous amounts of money into R&D (more than AMD, Nvidia, and TSMC combined) in an effort to close that gap. Intel claims they will take leadership from TSMC within the next months of so (but this is certainly not a guarantee and remains to be seen).
Another aspect is that Intel is only now just beginning to open their fabs to external design companies. For a long time, companies like AMD, Qualcomm, Nvidia, etc. Went with TSMC because they were the best available for them. A few years ago, Global Foundries was close enough to be considered, but they've since dropped out of the leading node market, and last gen Nvidia sourced from Samsung because their nm node was "good enough" and cheaper than TSMC's nm node and Nvidia was able to compete due to their extremely good architecture design that offset some of this discrepancy.
半导体制造是地球上最具挑战性的大规模生产行业。建立和运营晶圆厂需要巨额投资,而且每一代新的半导体技术(通常称为“节点”)的设计与生产难度和成本都在增加,延迟或出现故障的可能性也随之提高(英特尔的10纳米工艺就曾遇到这种情况)。因此,许多公司退出了竞争,在尖端领域仅剩下几家主要厂商:英特尔、三星以及台积电。
当前,台积电拥有最先进的制程技术,目前正应用于新款iPhone Pro系列及苹果M系列芯片,并预计在年内扩展至AMD和Nvidia等更多公司的产品中。
此外,全球只剩下一家主要企业——ASML,能够生产用于制造这些高度先进节点的经济高效的设备。这类机器极其复杂、昂贵且每年产量有限。
尽管台积电目前处于领先地位,但这一地位并非稳固不变。英特尔和三星并未落后太多,过去几年来,英特尔投入了巨额资金进行研发(超过了AMD、Nvidia和台积电研发投入的总和),力图缩小差距。英特尔声称将在未来几个月内夺回领导地位(但这当然不是保证,还需拭目以待)。
另一个方面是,英特尔直到最近才开始将其晶圆厂开放给外部设计公司使用。长期以来,像AMD、高通、Nvidia等公司选择与台积电合作,因为台积电是当时他们可获得的最佳选择。几年前,Global Foundries也曾接近达到尖端水平,但后来退出了领先节点市场竞争。上一代Nvidia的部分产品采用了三星的8纳米节点,因为它对于Nvidia来说已经足够好,同时比台积电的7纳米节点更便宜。由于Nvidia在架构设计方面的出色表现抵消了部分性能差距,使其能够在市场上保持竞争力。
Samsung production technology is quite close to TSMC. According to them “Samsung's 4 nm technology is two years behind TSMC's, and our 3 nm is about a year behind. But things will change when TSMC enters the 2 nm process," So quite similar and the difference will become smaller over time.
The reason one or a few companies are dominant is because of the cost. The cost to develop a process at a smaller scale is billions of dollars. The cost to build a factory for it is billions. Whe a process changes to a smaller scale the upgrade cost is almost the same as a new factory, the buildings are not the most expensive part, it is the equipment in them.
TMSC built a new factory in Phoenix, Arizona The initial investment was $12 billion and when the factory grows the total investment will be around $40 billion. The construction of the first parts started in 2021 and production is expected at the end of 2024.
Because of the high development cost a single company that can build many factories has the advantage, It can spread out the development cost over all factories and chips made.
If you try to create a competitor you will first need to get people, the ones that know how to do it and already work at the other funders. You would also need to do a lot of research to get to the level they were when the new process development started, the you will need to repeat what they did. The complied that is first and developed new technology will get lots of patents and can stop you from doing some parts exactly like they do. This makes it harder if you are late.
Fo there is one part of the semiconductor manufacturing industry only one company can do that is ASML in the Netherlands. They mate the photolithography machins that is used to project the pattern onto the wafers other processes is to add and remove material. No other company can make machines that project them for the smaller processes.
三星的生产工艺与台积电相当接近。据称,“三星的4纳米技术落后于台积电约两年时间,而我们的3纳米技术大约滞后一年。但当台积电进入2纳米工艺时,情况将会有所改变。”两者差距虽存在,但随着时间推移会逐渐缩小。
一两家公司在市场中占据主导地位的原因在于高昂的成本。开发更小尺寸工艺的成本高达数十亿美元,而建造相应的工厂也需要数十亿美元的投资。当工艺升级至更小尺度时,升级成本几乎等同于新建一座工厂,建筑本身并非最昂贵的部分,真正贵重的是其中的生产设备。
台积电正在美国亚利桑那州凤凰城建设一座新工厂,初期投资为120亿美元,随着工厂扩建,总投入预计将达400亿美元左右。该工厂部分设施建设始于2021年,预计到2024年底开始投产。
由于高昂的研发成本,能够建立多个工厂的单一企业具有优势,它们可以将研发成本分摊到所有工厂及生产的芯片上。
若试图创建一个竞争对手,首先需要找到人才,即那些已经在这个领域工作并熟知业务的人才,并且还需要进行大量的研究以达到当前先进工艺的发展水平,之后再重复领先者所做的工作。先行研发并取得新技术突破的企业会获得大量专利,可以通过专利阻止后来者完全复制其做法,这无疑增加了后入局者的难度。
在半导体制造业中,有一部分只能由荷兰的一家公司完成,那就是ASML。ASML制造用于光刻工艺的光刻机,这种设备将图案投影到晶圆上,其他工艺则包括添加和去除材料。对于更小尺寸的制造工艺,目前没有其他公司能够制造出能实现此类精密投影的机器。
Why is ASML the only one?
为什么ASML是唯一的供应商?
Your understanding ist wrong. ASML ist the only company that makes the machines to produce chips. It's a durch company.
So all the world ist not dependend on a taiwanese company. It's a dutch company.
您的理解有误。ASML 是唯一一家生产芯片制造机器的公司,它是一家荷兰公司。因此,全世界并非仅依赖一家台湾公司,而是依赖于这家荷兰企业。
For the same reason TSMC is the leading manufacturer, or iPads are better than other tablets. You pour your profits into R&D, so your product gets better, so more people work with you, so you gain invaluable experience and even more profits, which you keep pouring into R&D, which keep increasing the lead between you and your nearest competitor.
ASML成为领先供应商的原因与台积电成为顶级制造商、iPad优于其他平板电脑的原因相似。你将利润不断投入到研发中,使得产品变得更好,从而吸引更多合作伙伴,积累宝贵的经验并获得更多利润,这些利润又被继续投入研发,进一步拉大了与最接近的竞争对手之间的差距。
There were companies trying to make EUV lithography (now the most cutting edge) possible. Canon, Nikon and ASML. Canon and Nikon stopped development as they thought the technology would never be feasible at mass production levels. ASML continued. Samsung, intel, tsmc all invested heavily in ASML so they could get the technology to work. They managed to do it. Now they are the sole supplier of the most cutting edge technology.
曾有3家公司尝试开发极紫外光刻技术(如今是最尖端的技术),它们分别是佳能、尼康和ASML。佳能和尼康认为该技术无法实现大规模生产而停止了研发工作。但ASML坚持了下来。三星、英特尔和台积电都对ASML进行了大量投资,以帮助其攻克这项技术难题。最终ASML成功做到了,并且目前成为了唯一能够提供最尖端极紫外光刻技术的供应商。
You'll often find this with very specialized knowledge. Once a company has it setup, the barrier to entry is so high that not too many other people are doing bother even trying to get into that business. This is especially true if the business is pretty open.
What I mean by that is TSMC (The Taiwan) company will make the computer chips for anyone. If you're a US company and want to make a chip. You send TSMC the design and the money and they will make it for you. They're very open.
You couldn't do this with Intel (at least historically). Intel only chips for Intel.
So when AMD (Intel's competitor) wanted to make chips, they just sent their design to TSMC instead of building their own chip factories. TSMC grew really fast and powerful like this because they kept improving their process, having specialized knowledge, and making chips for anyone.
As others have said, Samsung and Intel both have similar knowledge.
It may not be the case forever. There has been some tension between China and USA/Taiwan and there were some restrictions on the kind of chip TSMC could make for China. Now there is an incentive for China to be able to make it's own chips. So China has been investing heavily trying to duplicate the process. The USA is also trying to get more chip production in the USA itself as it doesn't want to be so reliant on Taiwan in case stuff happens.
在非常专业化的知识领域中,这种情况很常见。一旦一家公司建立起这样的技术壁垒,其他竞争对手进入该领域的难度就非常高,因此没有太多人愿意尝试。如果这个行业相对开放的话,这种现象尤为明显。
例如,台积电(台湾积体电路制造股份有限公司)会为任何需要的人制造计算机芯片。如果你是一家美国公司,想要生产一款芯片,只需将设计和资金发给台积电,他们就会为你制造出来。台积电的业务模式非常开放。
相比之下,英特尔(至少在历史上)仅为自己生产芯片,不对外提供代工服务。
所以当AMD(英特尔的主要竞争对手)想生产芯片时,他们选择将设计方案交给台积电而非自建晶圆厂。台积电凭借持续改进生产工艺、积累的专业知识以及为任何人制造芯片的方式,迅速成长为一个强大企业。
如其他人所说,三星和英特尔同样拥有类似的知识和技术。
但这种情况可能不会永远持续下去。中美之间以及中国大陆与台湾之间存在一定的紧张关系,曾有一些针对台积电为中国大陆生产某种芯片的限制措施。因此,中国现在有了自主研发芯片的动力,并已大力投入资源试图复制这一过程。同时,美国也希望增加本土芯片生产,以减少对台湾地区的依赖,以防万一出现状况。
As other responders have said, it’s not that other vendors don’t have leading edge transistor technology. So what TSMC does can, to a degree, be replicated. But what is very difficult to replicate is the combined volume and reliability at which TSMC does it. It’s a weird side-effect of semiconductor manufacturing: the more volume (number of wafer starts) you have, the more reliable (higher yield) the semiconductors produced will be. TSMC has been a leader at this for years, putting a steady percentage of their profits into technology and reliability improvements that bring them more and larger customers, which bring them higher profits and … around it goes. The only other companies which achieve a similar degree of volume and quality are the remaining DRAM and Flash makers: Samsung, SK Hynix and Micron. There are no non-DRAM, non-Flash foundries that are better than TSMC in terms of volume and reliability.
正如其他回答者所说,其他供应商并不缺乏先进的晶体管技术。因此,某种程度上可以复制台积电的技术。但是,非常难以复制的是台积电在组合产能和可靠性方面的优势。这是半导体制造中的一个奇特副作用:产能(晶圆开启数量)越大,所生产的半导体可靠性(良率)就越高。多年来,台积电一直在这方面处于领先地位,稳定地将利润的一部分投入到技术和可靠性改进中,从而吸引更多更大的客户,带来更高的利润……如此循环。实现类似产能和质量水平的其他公司仅有一些DRAM和闪存制造商:三星、SK海力士和美光。就产量和可靠性而言,没有非DRAM、非闪存厂商能比得上台积电。
TSMC, the company in question doesn't design chips, and is dependent on a Dutch company for the machines.
That's not to say they don't have their own process, one that's given them a competitive edge, but it's not like other "foundries" (the name for this type of company) couldn't achieve the same result next year. TSMC has been ahead of their competition, but they do have competition.
Said Dutch company (ASML) however does not. No one can make current gen high end chips without a machine from Eindhoven, and despite the Chinese stealing company secrets, that's unlikely to change any time soon.
台积电,也就是我们讨论的这家企业,并不设计芯片,其制造工艺依赖于一家荷兰公司所提供的机器。
这并不是说台积电没有自己的独特生产工艺,正是这一工艺赋予了他们竞争优势,但也不能认为其他“晶圆代工厂”(这种类型公司的名称)明年就无法取得同样的成果。台积电在竞争中确实领先,但他们确实面临着竞争压力。
然而,上述那家荷兰公司(ASML)的地位则独一无二。如果没有来自埃因霍温的设备,目前任何公司都无法生产出高端的最新一代芯片。尽管中国存在窃取该公司商业机密的情况,但在可预见的未来,这种情况不太可能发生改变。
Most nations know the broad strokes of how to make a nuclear weapon but the how is so nuanced that it’s almost impossible to reverse engineer without incredible financial resources.
Same with chips. They have insane in-house knowledge curated over decades that is so sensitive, should China attack Taiwan, they would more likely destroy their processes rather than let them fall into c** possession. We can see the final result, but trying to reverse engineer it without access to the process is challenging
The US is now committing huge financial resources to try to catch up with state-side competition.
大多数国家都知道制造核武器的大致方法,但细节非常微妙,几乎不可能在没有极大财力的支持下进行逆向工程。
芯片也是一样。台积电拥有经过数十年培育的疯狂内部知识,这些知识非常敏感,如果中国(大陆)攻击台湾,他们更有可能摧毁他们的制程,而不是让它们落入***的手中。我们可以看到最终的成果,但在没有获取到完整的制程的情况下进行逆向工程是具有挑战性的。
美国现在正在投入巨大的财力,试图赶上国内的竞争。
Intel and Samsung can. The 'nanometer' thing is mostly marketing fluff. It actually doesn't refer to any single measurement on a transistor. Thus, there is no part on a 3nm chip that is 3nm. TSMC just has the largest capacity to produce a chip that is in the '3nm class'.
Intel 和三星也具备相关技术。所谓的“纳米”概念在很大程度上是一种营销术语,实际上它并不指代晶体管上的任何单一尺寸测量值。因此,在一块3纳米芯片上,并没有哪个部分的实际尺寸是3纳米。这里的“3nm级别”是指台积电拥有最大的产能来生产这种最尖端的芯片。尽管如此,台积电能够在“3nm等级”的芯片制造方面保持领先,拥有最高的产量和最先进的工艺水平。
That largest capacity aspect is critical to success though. TSMC has the highest yields, which means its actually economical for them to produce 3nm chips. Samsung could get there, Intel has had plenty of issue with yields in multiple generations.
产能规模这一方面对成功至关重要。台积电的良品率最高,这意味着他们制造3纳米芯片在经济上是可行且有利可图的。三星有可能达到这一水平,而英特尔在多个产品世代中都遇到了良品率问题。
Pretty much the biggest reason is that there isn't enough volume in chip manufacturing to justify it. If we added similar chip manufacturer, it wouldn't double the amount being produced because there aren't enough people to consume all of them. That means the investment would likely not be returned. All it would do is hurt TSMC, but it would hurt the investor more.
主要原因在于当前芯片制造业的需求量不足以支撑更多类似制造商的加入。即使增加一个类似的芯片制造商,由于市场需求有限,并不会导致产量翻倍,因为没有足够多的消费者来消化这些额外的芯片。这意味着投资可能无法得到回报。这样做不仅会冲击台积电的地位,但对投资者来说损失可能会更大。因此,在当前市场环境下,新增同类芯片制造商并不划算。
The Taiwanese government made the strategic economic and military decision that relying on other people's inate love freedom and philosophical support for western-style democracy was all well and good, but being as indispensable to what they thought the modern global economy would be as possible was better.
As such the Taiwanese government invested metric fucktons of money and resources and social capital in promoting the growth of the semiconductor industry and a few other sectors.
Others could replicate their success, but they would have to be able to, if not match their drive and intensity, at least match (or more likely exceed) their resource expenditure in order to even have a chance at succeeding. TSMC was a very large bet for the government of Taiwan, and it took decades of investment before it really began to pay off. If you aren't under the same sort of existential pressures that they are ,it is probably a lot more difficult to maintain that level of investment for decades before seeing results.
The US used to be the world leader in semiconductors, and it is beginning to reinvest there, but even in such historically fertile grounds and even with the resources of the US, results are not guaranteed, especially given how fractious our political process has become.
台湾政府在战略上做出了一项经济和军事决策,认为依赖他人对自由的内在热爱以及对西方民主理念的支持固然重要,但成为现代全球经济中不可或缺的一环更为明智。因此,台湾政府投入了巨额资金、大量资源和社会资本来推动半导体行业和其他几个关键领域的发展。
其他人可能可以复制他们的成功,但他们必须能够,至少在资源支出方面能够匹敌(或更可能是超过)他们的投入,才有机会成功。对于台湾政府来说,台积电是一个非常大的赌注,花费了几十年的投资才真正开始取得回报。如果你没有面临与他们相同的生存压力,要在数十年的时间里保持如此巨额的投资才能看到成果可能会更加困难。
美国曾经是半导体领域的世界领导者,现在正在重新投资,但即使在如此历史悠久的沃土上,即使拥有美国的资源,也不能保证取得成果,尤其是考虑到我们政治进程变得多么分裂。
Its not
TSMC is a factory / machinery provider, other people buy time on its machinery and make THEIR designs.
TSMC doesn't design chips, it simply mass produces them cheaply and reliably.
It also doesn't make the machines it makes chips with, it buys them.
So REALLY, TSMC just buys a huge amount of chip making machines, employs a very skilled workforce to use said machines, and leases those machines operational hours on favourable rates to others.
Chip foundries are expensive and to be profitable you need to run them 24/7/365
Intel makes 70% of the worlds PC's it can run its fabs none stop, AMD makes 20%, better and cheaper to forward order hours for new products then pick up slack hours for the back orders.
台积电并非芯片设计者,它更像是一个工厂和设备提供商。其他公司购买台积电的机器使用时间,并利用这些设备生产自己的芯片设计。台积电并不参与芯片的设计工作,而是专注于高效且低成本的大规模生产。同时,台积电所使用的芯片制造机器也是采购自其它供应商,而非自行制造。
实际上,台积电所做的就是大量购入芯片制造机器,雇佣一批技术熟练的员工操作这些机器,并以优惠的价格将机器的运营时间租赁给其他客户。半导体晶圆厂的运营成本高昂,为了实现盈利,通常需要全天候(24小时/7天/全年无休)运作。
英特尔生产的芯片占据了全球PC市场70%的份额,因此可以不间断地运行其晶圆厂;AMD则占据约20%市场份额。对于新产品的订单,提前预定生产线工时更为经济有效,而对于补单或追加订单,则可以在非满负荷时段进行生产。
原创翻译:龙腾网 https://www.ltaaa.cn 转载请注明出处
This is grossly oversimplifying how difficult it is to do what TSMC does. They are by far the leading foundry in the world at the leading edge. The manufacturing aspect is much more difficult than the actual design of the chip. Hence why there used to be a lot more foundries and most of them have stopped competing at the leading edge because they couldn't achieve sufficient yield.
There are plenty of companies like Amazon, Google, Apple, etc who have designed their own chips, but there are only TSMC and maybe Samsung that can actually produce them.
对台积电所做的事情进行了过于简化。台积电目前是全球领先的尖端晶圆代工厂,其制造工艺远比芯片设计更为复杂艰难。过去有很多晶圆代工厂,但由于难以达到足够的良品率,大多数已经停止在最尖端领域与台积电竞争。虽然有许多像亚马逊、谷歌、苹果这样的公司设计出了自家的芯片,但真正能够生产这些芯片的只有台积电和可能的三星。
Chip fabrication is extremely expensive, building a fab alone costs billions of dollars, so there are only a handful of companies that are in the field at all. TSMC, Samsung, Intel, GloFo, Huawei, Micron, a few smaller ones.
Each fab has its own proprietary processes and technologies, and it takes years of research and huge sums of money to improve those processes.
TSMC got its lead by consistently pouring more money into R&D than its competitors during the s and s, and now that it's on top it keeps spending on R&D so that others can't catch up.
芯片制造是一项极其昂贵的业务,仅建设一座晶圆厂就需要数十亿美元的投资,因此在这个领域中活跃的公司数量极为有限。包括台积电(TSMC)、三星(Samsung)、英特尔(Intel)、格芯(GlobalFoundries)、华为海思半导体和美光科技(Micron)在内的几家大公司,以及一些规模较小的企业。
每座晶圆厂都有各自独特的工艺流程和技术,而要改进这些生产工艺需要经过数年深入研究,并投入巨额资金。在上世纪90年代至21世纪初,台积电通过持续投入比竞争对手更多的研发资金,逐渐建立了领先地位。如今台积电继续保持高额的研发投资,确保其他公司在技术上难以赶超其领先地位。