Resolves yes if it is announced before the end of 2023 that a majority of Chinese chip companies will be blocked from making leading-edge chips at TSMC, similar to how Huawei was blocked from making new orders starting in 2020.
1,000
3.00I think this should probably resolve yes already? The way the export controls work is that exports of advanced chips to China from TSMC are under US export controls, even if the designs come from Chinese firms. https://arstechnica.com/tech-policy/2022/11/chinese-chip-designers-slow-down-processors-to-dodge-us-sanctions/amp/
https://www.lawfareblog.com/right-time-chip-export-controls
This ban applies to all foreign-made GPUs as a result of a modified extraterritorial jurisdictional rule called the foreign direct product rule, or FDPR for short. The new FDPR subjects any such chip made directly from American technology or software, or produced, even in part, from U.S.-made semiconductor production equipment to U.S. jurisdiction.
Because all semiconductor fabrication facilities use at least some U.S.-made equipment, every such GPU on the planet is now subject to U.S. controls.
@Radicalia Wow I'm surprised this wasn't bigger news when the sanctions were announced. I was anticipating a ban on leading edge process nodes but the legal threshold is 600 GB/S of interconnect bandwidth. For reference, Nvidia HPC GPUs max out at 600GB/s of NVLink bandwidth plus 64GB/s of PCIe bandwidth so it's slightly over the threshold, and AMD Epyc server CPUs max out at 512GB/s of PCIe bandwidth. Though I don't think this affects many things other than training large AI models and scientific supercomputers for things like computational chemistry and nuclear weapon simulations, this limit won't increase over time and so it will eventually become a significant impediment for those applications, and so I think you're right that this should resolve Yes.
