Forgot your password?
typodupeerror
China

How China Built Its 'Manhattan Project' To Rival the West in AI Chips (reuters.com) 171

Chinese scientists have built a working prototype of an extreme ultraviolet lithography machine in a high-security Shenzhen laboratory, a development that represents exactly what Washington has spent years and multiple rounds of export controls trying to prevent: China's path toward semiconductor independence and an end to the West's monopoly on the technology that powers AI, smartphones and advanced weapons systems.

The prototype, completed in early 2025 by former ASML engineers who reverse-engineered the Dutch company's machines, is operational and generating EUV light, though it has not yet produced working chips. The effort is part of a six-year secret government initiative that sources described to Reuters as China's version of the Manhattan Project.

Huawei is coordinating thousands of engineers across companies and state research institutes, and recruits are working under false identities inside secure facilities. The Chinese government is targeting 2028 for producing working chips, though sources say 2030 is more realistic -- still years earlier than the decade analysts had predicted it would take China to match the West.
This discussion has been archived. No new comments can be posted.

How China Built Its 'Manhattan Project' To Rival the West in AI Chips

Comments Filter:
  • That was fast (Score:5, Informative)

    by Snotnose ( 212196 ) on Thursday December 18, 2025 @10:10AM (#65866409)
    Last year I was predicting 5-10 years until China could make chips. It took them 1. The USA is fucked, especially with that racist, demented asshole in charge.
    • China has been making chips for ages. This is about making AI chips that are comparable with Nvidia, AMD, Google, Amazon etc using modern fabs (TSMC primarily, also Intel, Samsung, Micron etc that mostly use ASML machines).

      The article suggests they're likely > 5 years behind, and that probably isn't to scale production. Also consider EUV was first used in 2018 commercially, 2019 at scale. So maybe early Chinese EUV will be a decade behind when it launches.

      Lots of the supporting ecosystem to develop to

    • The issue is you have taken China's word they have actually accomplished what they said they did. Personally, I don't believe them until I have seen it. For example, China proclaimed a breakthrough when SMIC made the Kirin 9000 processor for the Huawei Mate 60 Pro in 2023. The chip was a 7nm chip which all the "naysayers" said China could never manufacture. Except it was made using DUV not EUV. No one ever said 7nm was not possible using DUV. The main reason EUV was used was the smaller size meant low yield

      • by AmiMoJo ( 196126 )

        It's a variation on the 4 stage strategy.

        https://youtu.be/3hua1pkDmJc [youtu.be]

        1. Say you don't believe them.
        2. Say it's technically true but nothing to worry about.
        3. Admit that they caught up, but it's just a copy and they can't improve on it.
        4. Tariffs to "protect" Western industry that has been left behind.

    • by HiThere ( 15173 )

      This is a lab machine, and it's not clear that it's making large chips. I think your 5-10 year prediction of last year is probably right. There will be engineering challenges in converting a lab machine into a production machine.

      Actually, my (uninformed) prediction last year, and this year, is that it will take about a decade for China to equal the production of TSMC assuming TSMC keeps improving. But that they'll have "good enough for 90% of the market" within a very few years (and perhaps already do).

    • Last year I was predicting 5-10 years until China could make chips. It took them 1. The USA is fucked, especially with that racist, demented asshole in charge.

      From the fine summary:

      The prototype, completed in early 2025 by former ASML engineers who reverse-engineered the Dutch company's machines, is operational and generating EUV light, though it has not yet produced working chips. The effort is part of a six-year secret government initiative

      This is a prototype, not a complete product. It's the result of a 6-year program, not 1. They have not produced any working chips. Your prediction was bad and your reading comprehension is bad and you should feel bad.

      The USA is fucked, especially with that racist, demented asshole in charge.

      Yes, bu

  • So 5 years from now, they'll still be 5 years behind?

  • by petes_PoV ( 912422 ) on Thursday December 18, 2025 @10:44AM (#65866479)
    The idea that one country can develop a technology that no other can, is as flawed as it is arrogant. And by refusing to sell advanced technology, the reasons to produce domestic alternatives get stepped up a gear - or several.

    Once you accept that a competitor or adversary has both the ability and the will to create technologies domestically, that they would be prohibited from purchasing, you have to accept that the originator has lost control. What is worse is the possibility that they might just make a better version than you have.

    • The idea that one country can develop a technology that no other can, is as flawed as it is arrogant. And by refusing to sell advanced technology, the reasons to produce domestic alternatives get stepped up a gear - or several. Once you accept that a competitor or adversary has both the ability and the will to create technologies domestically, that they would be prohibited from purchasing, you have to accept that the originator has lost control. What is worse is the possibility that they might just make a better version than you have.

      And no one said any of that. In the world today, the current EUV machines are made by one company in the world. It is ASML in the Netherlands. The US nor Japan produces them, and both countries have a long history with making lithography machines. The problem was the cost of R&D and the specific strategies to make EUV was successful only for one company. Making EUV machines is not an easy task that someone can do in their garage this weekend.

      Can China copy everything ASML did? Sure. The issue is that it

      • What could be a game changer is if China finds a simpler way to do EUV. https://semiwiki.com/forum/thr... [semiwiki.com] Not a done deal by any stretch, but by forbidding China the ASML fruit, they may end up with a less expensive better tasting fruit. Another thing that is being overlooked is the optics side of the equation. Not just ASML is out of reach, they are also not able to get the Zeiss optics. Also a necessary piece of the puzzle at the moment. I think they are shadow importing Nikon and Cannon optics, although
        • by AmiMoJo ( 196126 )

          Zeiss must be very upset that the Chinese have been pushed to develop competitive lenses even faster.

  • by klipclop ( 6724090 ) on Thursday December 18, 2025 @10:54AM (#65866501)
    At this point: Missed the boat, A day late and a dollar short,That ship has sailed... All of the Trump Tariffs are self serving and he's proven he will T.A.C.O whenever pressure is high enough or if he can self profit...
  • ... massive industrial spy op to get on par, you actually prove your gross incompetence.

  • "The prototype, completed in early 2025 by former ASML engineers who reverse-engineered the Dutch company's machines, is operational and generating EUV light". "recruits are working under false identities inside secure facilities"

    China must be paying them a fortune for this theft, no wonder they want to remain anonymous. The basic technique ASML uses to create the EUV light is public knowledge and ASML even describes it, but the machinery to do it is not.

    They shine a powerful industrial CO laser at microsc

    • Once you get to UV-C wavelengths, you stop being able to use traditional optics. I doubt they're using mirrors at all. This is for the rather funny reason that the wavelength starts approaching the size of individual atoms such that you can't make a mirror that appears smooth to the light beam you're using. You start having to use things like grazing incidence mirrors [wikipedia.org] or diffraction gratings to collimate the light.

      You could also skip all that by using an electron beam to do your etching. You can easily get

      • According to their website, they "developed a brand-new optical system that uses ultrasmooth, multilayer mirrors inside a vacuum chamber. Each mirror has over 100 layers of materials that were carefully chosen and precisely engineered to maximize the reflection of EUV light.

        Flatness is also crucial for EUV reflection. The mirrors are polished to a smoothness of less than one atom’s thickness. "

        • That's a description of a fairly traditional type of mirror called a dielectric mirror [wikipedia.org]. Basically every optical telescope made in the last 75 years uses them. Even most of the higher-end hobbyist telescopes use them. I'm guessing they use something besides the normal aluminum deposited on borosilicate glass for their mirrors. It would need to have lots of electrons to reflect such high frequencies, be relatively durable, and also be shiny/reflective. If I had to guess, I'd say gallium, indium, tin, or some

      • Regarding the mirrors, an information clip of ASML I recently saw mentioned that if enlarged to the size of Germany, no point of it sticks up more than 1mm. Like the other post mentioning their mirrors are flat to the atomic level, it's just more extreme than other similarly flat mirrors.
        • That's not that special. For example, the main mirror on Hubble, which is also the same primary on the KH-11 spy satellites, is an order of magnitude smoother than this and is designed to work in optical and near-IR wavelengths. While Hubble's primary is doped with aluminum and James Webb is doped with gold, ASML's mirrors would need a different dopant.

          • That some very scientific and very expensive equipment happens to have similarly special parts makes you say that it's not special that ASML machines have that too..? Just because it's been done before, doesn't make it not special, if you ask me. Now, if I can buy such a mirror for only a little bit more than a regular one, and almost any mirror manufacturer can produce it, then I'd call it not special.
  • Chinese engineers and scientists are smart
    Attempting to prevent China from acquiring tech is futile and counterproductive
    Cooperation would be better

  • If they can't keep up with custom hardware then somebody else is going to build it. What's more because AI is a technology that by design is going to consolidate into a few players who have access to training data (google, Facebook and maybe Apple) you are going to have companies so huge they are just going to want to make their own chips.

    You've got these companies that a few changes in who makes the chips could drop their value by hundreds of billions of dollars. And you have investors buying in during
  • by business_kid ( 973043 ) <business@kid.linuxmail@org> on Thursday December 18, 2025 @12:40PM (#65866847)
    Besides EUV, there's also the question of actually making working silicon that's tested and working. The process is: make the silicon wafer; test the silicon wafer; package the few that pass testing.

    TSMC are the masters, inventing small machines and leading the world. Apple are 2nd with a working 3nm wafer fab line. Samsung have a working 7nm line, I heard they were upgrading to 5nm, but I never heard of product coming out, so they seem stuck there. Intel has a 10nm wafer fab line, but Intel can't get it working. They've rejigged things so their 10nm line is actually putting out 12nm chips but they still lose money trying to make silicon. They make their CPUs on a 14nm line.

    Here's China's progress. In 2021, Rockchip released the RK3588 on 8nm Silicon. They have that technology for many years, and I don't doubt they will improve on it. Chip making is extremely seismically sensitive, and barely noticable movement on a seismograph just wrecks stuff. Ditto power surges, lightning, etc.

    Smaller fab size = lower power usage at the same frequencies or higher frequency for the same power. But every increment downwards needs complete redesign now. It's not just about making things smaller, because thinner insulation doesn't work.
  • The (linked) article begins with:

    "It was built by a team of former engineers from Dutch semiconductor giant ASML (ASML.AS), opens new tab who reverse-engineered the company's extreme ultraviolet lithography machines or EUVs, according to two people with knowledge of the project."

    Then, much later in the long article, it explains:

    "ASML won an $845 million judgment in 2019 against a former Chinese engineer accused of stealing trade secrets, but the defendant filed for bankruptcy and continues to operate in Bei

    • Never mind. My question is answered later in the article:

      "The ASML veterans made the breakthrough in Shenzhen possible, the people said. Without their intimate knowledge of the technology, reverse-engineering the machines would have been nearly impossible."

  • Smooth seas make poor sailors. The US is creating some outstanding Chinese sailors.
  • by sudonim2 ( 2073156 ) on Thursday December 18, 2025 @01:08PM (#65866923)

    The part most people think of when they think of the Manhattan Project is the designing of nuclear devices. In reality, the bulk of the Manhattan Project was manufacturing the weapons-grade fissile material. That part of the Manhattan Project occupied roughly a third of US industrial capacity during WWII. Was this EUV R&D project consuming a third of China's industrial output? Was it even consuming an absolute amount of manufacturing capacity equal to a third of 1940's-era America? No? Then it's not China's version of the Manhattan Project.

  • I see Huawei and the Chinese Government are up to their old tricks. They stole technology wholesale from Nortel then undercut them, which destroyed Nortel. How soon before they do the same to ASML?

  • A big part of how China is moving so fast technologically is the advent of LLMs have turned the petabytes of data, emails, etc that they have stolen from western companies into usable information. They can load up corpus of content stolen from a chip producer and ask the question: "What issues arose when transitioning to ultraviolet lithography and what were the engineering changes that resolved those issues?". They can skip over many of the issues that previous companies had to work through the hard way

Feel disillusioned? I've got some great new illusions, right here!

Working...