China’s National Supercomputing Center has launched LineShine, an exascale supercomputer built entirely from domestically produced chips that, if its performance claims hold up, would dethrone the US’s El Capitan as the world’s most powerful machine.
The system, also known as Lingsheng, targets 2 exaflops of sustained performance. For context, El Capitan currently holds the top spot at 1.8 exaflops on the Linpack benchmark, the standard yardstick for measuring supercomputer muscle. In English: LineShine is designed to do two quintillion calculations per second, roughly 11% faster than the best America has to offer.
All Huawei, all the time
LineShine runs on 47,000 Huawei LX2 Armv9 CPUs, each packing 304 cores. That is a staggering amount of silicon, all of it Chinese-made.
The full architecture spans 92 compute cabinets with a total storage bandwidth of 10 TB/s and 650 petabytes of capacity. Each node delivers 1.6 terabits per second of bandwidth, giving the system the interconnect fabric needed for both AI training and traditional high-performance computing workloads.
Full deployment is expected by the end of 2025. Huang Xiaohui, who has been associated with the project’s vision, has described the goal as achieving “full-stack independence” in computing technologies.
No independent Linpack benchmark data has been published for LineShine. El Capitan’s 1.8 exaflops figure is verified and publicly listed in the TOP500 rankings. LineShine’s 2-exaflop target is just that: a target. Until independent benchmarks confirm the number, the claim sits in a different category than El Capitan’s verified result.
China has historically been reluctant to submit its most advanced systems to the TOP500 list, which makes direct comparisons difficult and fuels skepticism among Western researchers.
Why now: export controls and the chip cold war
The US has spent the last few years tightening export controls on advanced semiconductors headed to China, specifically targeting the kinds of GPUs and AI accelerators that power cutting-edge computing. Washington’s restrictions have hit Nvidia, AMD, and other chipmakers, limiting what they can sell to Chinese entities.
Huawei’s LX2 processors, built on the Armv9 architecture, represent China’s domestic alternative effort in silicon form. LineShine is essentially a proof of concept for an entire computing ecosystem that does not rely on American technology at any point in the supply chain.
What this means for crypto and AI compute
Supercomputers at the exascale level bring the conversation about quantum and post-quantum cryptography closer to practical relevance. Bitcoin’s SHA-256 and Ethereum’s Keccak-256 are not at immediate risk from classical supercomputers, but the pace of advancement compresses timelines.
The verification question remains the most important near-term variable. If LineShine submits to the TOP500 and confirms its 2-exaflop claim, it reshapes the supercomputing hierarchy in a single stroke. If it does not, the announcement remains a statement of ambition rather than achievement, and the US retains its verified lead with El Capitan.