China Defies US Blockade, Unveils Exascale 'Monster' Supercomputer Without Nvidia GPUs

The Chinese government has once again demonstrated that technology blockades by the United States cannot stop their ambitions. The Celestial Empire has just showcased LineShine, a colossal exascale supercomputer built entirely with domestically-made components. The launch delivers a heavy blow to US export controls. LineShine was built without using a single Western-made graphics processing unit (GPU), such as Nvidia or AMD.

The supercomputer, developed by the National Supercomputing Center (NSCC) in Shenzhen, features a highly unique architecture. Instead of relying on GPUs to process artificial intelligence (AI) tasks and scientific simulations, LineShine has been designed exclusively as a CPU-only machine. To compensate for the absence of GPUs, China designed this hardware system on a massively large scale. LineShine is powered by 40,960 LX2 processors based on the Armv9 architecture, specially designed by Huawei. Each of the tens of thousands of LX2 processors is equipped with 304 cores. In total, the supercomputer “monster” aggregates around 2,451,840 CPU cores operating concurrently. In specific AI model training scenarios, the machine’s peak performance is claimed to reach up to 2.16 Exaflops. If this sustained performance is proven in practice, LineShine would theoretically surpass the fastest US supercomputer at present, El Capitan (1.8 Exaflops). Each chip combines integrated High Bandwidth Memory (HBM) of 32 GB with a very fast bandwidth of 4 TB/s, plus an external DDR5 memory of 256 GB. The LineShine supercomputer (also often referred to as Lingsheng) represents Beijing’s direct response to Washington’s tightened embargo. For years, the US has assumed that cutting China off from GPU supply would cap its computational capabilities. However, LineShine’s launch proves otherwise. China has built High-Performance Computing (HPC) infrastructure using domestically-produced silicon technology. This colossal computing system will be aimed at helping Chinese universities and research institutions execute complex workloads, from training large-scale AI models, to molecular simulations and climate modelling.