China just threw down the gauntlet in the global supercomputing race. The country announced its LineShine supercomputer on April 24, a machine designed to achieve over 2 exaflops of sustained performance using entirely domestic CPUs, with full deployment planned by the end of 2025.
If that target holds, LineShine would leapfrog the US’s El Capitan, currently the world’s fastest verified supercomputer at 1.742 exaflops on the TOP500 benchmark list. In English: China is trying to build a machine that can perform more than two quintillion calculations per second, all without a single foreign chip.
The numbers behind the rivalry
El Capitan, deployed in 2024 at a cost of $600M, represents the third US exascale system. It packs over 11 million cores across more than 11,000 nodes and hit a peak performance of 2.79 exaflops. Its verified sustained score of 1.742 exaflops displaced the previous champion, Frontier, which had achieved 1.353 exaflops.
China’s current verified top supercomputer, by contrast, sits at 487.94 petaflops. That’s roughly a quarter of one exaflop, meaning the jump to LineShine’s claimed 2-plus exaflops would represent more than a fourfold increase over anything China has publicly demonstrated.
China has long been suspected of running exascale-class machines that it simply doesn’t submit for independent benchmarking. The country withdrew from the TOP500 rankings years ago, making it difficult to compare capabilities directly.
The most notable technical detail is the architecture. LineShine is built on a CPU-only design, deliberately avoiding the GPU-accelerated approach that powers El Capitan and most modern high-performance systems.
Why CPU-only matters: the export control angle
Washington has spent the past several years tightening the screws on advanced chip exports to China, particularly targeting high-end GPUs from companies like Nvidia and AMD. By going CPU-only with fully domestic components, China is building LineShine as an export-control-proof machine. Every layer of the stack, from hardware to software, is designed to operate independently of foreign technology.
The timeline is aggressive. El Capitan took years of development and $600M in investment before going operational in 2024. China is claiming it can deploy LineShine by the end of 2025, which would mean going from announcement to activation in roughly eight months.
What this means for crypto and decentralized computing
There’s also the AI angle. Supercomputers like El Capitan and LineShine are the training grounds for the next generation of AI models. As AI-driven crypto sectors expand, including everything from on-chain AI agents to decentralized inference networks, the underlying hardware race shapes what’s possible.
Until LineShine submits to independent benchmarking on the TOP500 or an equivalent list, its claimed 2-plus exaflops remain just that: a claim. China’s history of avoiding public benchmarks means the global computing community will likely treat this announcement with healthy skepticism until hard numbers appear.
