8 ms·

The US has lots of advanced fabs already. This isn't the part of the supply chain we're missing. This does nothing to reduce our dependence on China for manufac

The US has lots of advanced fabs already. This isn't the part of the supply chain we're missing. This does nothing to reduce our dependence on China for manufacturing. Edit: and this isn't actually a cutting edge fab. By the time it opens in 2024 the cutting edge process will be 3nm or beyond.

sremani 7y ago
I understand, but fabs are realm of competition with China and their 2025 plan. I am big fanboy of Robert Lighthizer, and with Phase 1 effectively ripped we are going to see a lot of moving things. May not happen immediately, first it will be slow then sudden.
- arcticbull 7y ago
  Fabs aren't really a realm of competition with China. Leading fabs are in Germany (GlobalFoundaries in Dresden), Taiwan (TSMC's primary fabs), South Korea (Samsung) and even the US (GlobalFoundaries) and Singapore (GlobalFoundaries).
  - wtallis 7y ago
    GlobalFoundries isn't really "leading"; they got out of that race a while ago and are now focusing all their R&D on niche fabrication technologies that aren't in competition for building CPUs, GPUs, or smartphone SoCs. And you didn't mention Intel, which is the only serious competition to TSMC when it comes to advanced fab technology for those kinds of chips.
    - arcticbull 7y ago
      Yeah, I didn't mention Intel, I know they have fabs domestically but they're also moving to TSMC after 10nm [1] so I figured they were getting out of the game. Good call-out re: GF. [1] https://wccftech.com/intel-abandoning-10nm-after-dg1-planning-to-use-tsmcs-6nm-and-3nm-for-next-generation-xe-gpus/ https://wccftech.com/intel-abandoning-10nm-after-dg1-plannin...
      - wtallis 7y ago
        That's a bullshit rumor from a bullshit site that you're misinterpreting anyways. Intel might plausibly use outside fabs for their new discrete GPU product line, but they are absolutely not getting out of the fab game in any meaningful way. They are capacity-constrained at the moment and don't really have room to make big GPUs in-house without subtracting from their manufacturing capacity for their well-established CPU product line.
      - blahblah12 7y ago
        That’s definitely not verified. I’ve read that rumor multiple times and each time, it hasn’t happened.
  - fermienrico 7y ago
    Having worked in the semiconductor industry, manufacturing specifically, I can attest that this is a false statement - US technology in semiconductor manufacturing is not simply from running the fabs, but building fabrication equiment that fabs purchase to process and package wafers. The US (Teredyn, Applied Materials, Entegris, KLA Tencor, Lam Research), Netherlands (ASML) and Japan (Daifuku, Hitachi, TEL, Advantest) are the top leaders of the semiconductor equipment [1]. The ones I just listed account for over 90% of the semiconductor equipment. Most of the deep tech in a fab is the equipment. The rest of the stuff is a giant box that's super clean and maintains that clealiness. All of these Fabs[2] buy their equipment from the aforementioned companies. Here is what an example supply chain looks like: NVidia(US) has the IP and "soft" tech, they use software tooling(US) to make layouts, then contracts out the manufacturing to TSMC(Taiwan) who then buys their equipment from KLA/AMAT/LAM/TEL/ASML. TSMC(Taiwan) processes the wafers which are packaged at OSATS(OSATS are a collection of many companies[3]). The final product such as a graphics card is then further made accessible by writing insanely deep software driver stack (CUDA drivers for example) that are all developed by NVIDIA(US). For CPUs, there would be ISA and other "soft" tech involved which is Intel(US), AMD (US), ARM(UK, owned by Softbank(Japan)), Qualcomm, Broadcom, Microchip and Apple (all US, its a long list). I didn't even touch storage/memory. In addition to US, a lot of software and hardware talent and manufacturing comes from the EU, Germany in particular. I used to get on the call with a German company in the late morning and the Germans would be working their ass off at 9pm in Germany. We had an amazing team. It is not uncommon to pay $20k for a motor made in germany. Somewhere in the middle to low end of the supply chain (motors, wires, connectors, may be even steel, copper, precious metals, chemicals) probably has Chinese involvement. You can imagine how a graphics card's supply chain gets very complex and deep. Back to the central point - Semiconductor is a massive industry with complex supply chain. Depends on the level we are talking about. Also Global Foundries is not considered as among top dogs when it comes to semicon tech, although that's also debatable. Edit: added tooling [1] Source of the data (2011, old but still relevant): https://www.vlsiresearch.com/public/cms_pdf_upload/706001v1.0.pdf https://www.vlsiresearch.com/public/cms_pdf_upload/706001v1.... [2] https://en.wikipedia.org/wiki/List_of_semiconductor_fabrication_plants https://en.wikipedia.org/wiki/List_of_semiconductor_fabricat... [3] https://en.wikipedia.org/wiki/Integrated_device_manufacturer https://en.wikipedia.org/wiki/Integrated_device_manufacturer
m0zg 7y ago
There's no guarantee there will ever be "3nm and beyond".
- huonpine 7y ago
  Any marketers know what happens once the 1nm name is taken? Dose it go negative, go to decimals, start counting back up or just a new naming convention?
  - RandomTisk 7y ago
    I wouldn't shed any tears for the industry if some lawsuits got filed against misleading claims regarding transistor feature sizes. That probably won't happen because feature size isn't a spec requirement for anyone per se, but it would be nice to see on principle.
  - noway421 7y ago
    Go down to picometre. 1000pm, etc.
    - wallacoloo 7y ago
      But 1000pm looks larger than 1nm. How about “-3 nm”?
  - brnt 7y ago
    I think we started speaking of nanometers around the 100nm point. Before that we had eg .25, .13 microns cpus. And that is of course where the microcomputer got its name, initially ICs had features of a few microns. Of course, the angstrom being the size that it is (.1nm), makes it unlikely there ever will be another scale. Whatever is next, it won't be silicon.

sremani 7y ago

I understand, but fabs are realm of competition with China and their 2025 plan. I am big fanboy of Robert Lighthizer, and with Phase 1 effectively ripped we are going to see a lot of moving things. May not happen immediately, first it will be slow then sudden.

arcticbull 7y ago
Fabs aren't really a realm of competition with China. Leading fabs are in Germany (GlobalFoundaries in Dresden), Taiwan (TSMC's primary fabs), South Korea (Samsung) and even the US (GlobalFoundaries) and Singapore (GlobalFoundaries).
- wtallis 7y ago
  GlobalFoundries isn't really "leading"; they got out of that race a while ago and are now focusing all their R&D on niche fabrication technologies that aren't in competition for building CPUs, GPUs, or smartphone SoCs. And you didn't mention Intel, which is the only serious competition to TSMC when it comes to advanced fab technology for those kinds of chips.
  - arcticbull 7y ago
    Yeah, I didn't mention Intel, I know they have fabs domestically but they're also moving to TSMC after 10nm [1] so I figured they were getting out of the game. Good call-out re: GF. [1] https://wccftech.com/intel-abandoning-10nm-after-dg1-planning-to-use-tsmcs-6nm-and-3nm-for-next-generation-xe-gpus/ https://wccftech.com/intel-abandoning-10nm-after-dg1-plannin...
    - wtallis 7y ago
      That's a bullshit rumor from a bullshit site that you're misinterpreting anyways. Intel might plausibly use outside fabs for their new discrete GPU product line, but they are absolutely not getting out of the fab game in any meaningful way. They are capacity-constrained at the moment and don't really have room to make big GPUs in-house without subtracting from their manufacturing capacity for their well-established CPU product line.
    - blahblah12 7y ago
      That’s definitely not verified. I’ve read that rumor multiple times and each time, it hasn’t happened.
- fermienrico 7y ago
  Having worked in the semiconductor industry, manufacturing specifically, I can attest that this is a false statement - US technology in semiconductor manufacturing is not simply from running the fabs, but building fabrication equiment that fabs purchase to process and package wafers. The US (Teredyn, Applied Materials, Entegris, KLA Tencor, Lam Research), Netherlands (ASML) and Japan (Daifuku, Hitachi, TEL, Advantest) are the top leaders of the semiconductor equipment [1]. The ones I just listed account for over 90% of the semiconductor equipment. Most of the deep tech in a fab is the equipment. The rest of the stuff is a giant box that's super clean and maintains that clealiness. All of these Fabs[2] buy their equipment from the aforementioned companies. Here is what an example supply chain looks like: NVidia(US) has the IP and "soft" tech, they use software tooling(US) to make layouts, then contracts out the manufacturing to TSMC(Taiwan) who then buys their equipment from KLA/AMAT/LAM/TEL/ASML. TSMC(Taiwan) processes the wafers which are packaged at OSATS(OSATS are a collection of many companies[3]). The final product such as a graphics card is then further made accessible by writing insanely deep software driver stack (CUDA drivers for example) that are all developed by NVIDIA(US). For CPUs, there would be ISA and other "soft" tech involved which is Intel(US), AMD (US), ARM(UK, owned by Softbank(Japan)), Qualcomm, Broadcom, Microchip and Apple (all US, its a long list). I didn't even touch storage/memory. In addition to US, a lot of software and hardware talent and manufacturing comes from the EU, Germany in particular. I used to get on the call with a German company in the late morning and the Germans would be working their ass off at 9pm in Germany. We had an amazing team. It is not uncommon to pay $20k for a motor made in germany. Somewhere in the middle to low end of the supply chain (motors, wires, connectors, may be even steel, copper, precious metals, chemicals) probably has Chinese involvement. You can imagine how a graphics card's supply chain gets very complex and deep. Back to the central point - Semiconductor is a massive industry with complex supply chain. Depends on the level we are talking about. Also Global Foundries is not considered as among top dogs when it comes to semicon tech, although that's also debatable. Edit: added tooling [1] Source of the data (2011, old but still relevant): https://www.vlsiresearch.com/public/cms_pdf_upload/706001v1.0.pdf https://www.vlsiresearch.com/public/cms_pdf_upload/706001v1.... [2] https://en.wikipedia.org/wiki/List_of_semiconductor_fabrication_plants https://en.wikipedia.org/wiki/List_of_semiconductor_fabricat... [3] https://en.wikipedia.org/wiki/Integrated_device_manufacturer https://en.wikipedia.org/wiki/Integrated_device_manufacturer

m0zg 7y ago

There's no guarantee there will ever be "3nm and beyond".

huonpine 7y ago
Any marketers know what happens once the 1nm name is taken? Dose it go negative, go to decimals, start counting back up or just a new naming convention?
- RandomTisk 7y ago
  I wouldn't shed any tears for the industry if some lawsuits got filed against misleading claims regarding transistor feature sizes. That probably won't happen because feature size isn't a spec requirement for anyone per se, but it would be nice to see on principle.
- noway421 7y ago
  Go down to picometre. 1000pm, etc.
  - wallacoloo 7y ago
    But 1000pm looks larger than 1nm. How about “-3 nm”?
- brnt 7y ago
  I think we started speaking of nanometers around the 100nm point. Before that we had eg .25, .13 microns cpus. And that is of course where the microcomputer got its name, initially ICs had features of a few microns. Of course, the angstrom being the size that it is (.1nm), makes it unlikely there ever will be another scale. Whatever is next, it won't be silicon.