China reclaimed the United States' coveted computing crown on Tuesday, intensifying a furious technical rivalry with ramifications for geopolitics, science, and national security.
A team of researchers used a set of common criteria for supercomputers to declare LineShine, a huge computing machine in Shenzhen, China, the fastest in the world. In addition to its speed, the system was unique in that it didn't use graphics processing units, which are specialized chips used by most high-end supercomputers to analyze large amounts of data. Instead, it used regular microprocessors.
According to Jack Dongarra, an organizer of the so-called Top500 list of the world's most potent supercomputers, that fundamental design may indicate a better way to combine artificial intelligence with conventional scientific work.The innovative device was recently examined at the Shenzhen Cloud Computing Center by Dr. Dongarra, a computer science and electrical engineering professor at the University of Tennessee. El Capitan, a system at California's Lawrence Livermore National Laboratory that has led a twice-yearly rating of supercomputer performance since November 2024, outperformed LineShine by more than 20%. Since 2017, China had not ranked a machine at the top of the list.
Regarding LineShine, Dr. Dongarra remarked, "It's an impressive system." "They improved us by creating a system that doesn't depend on GPUs."The competition for technological supremacy between China and the US is intensified by the powerful supercomputer. Leading AI models have been created by US IT companies including OpenAI, Anthropic, and Google, while Nvidia has emerged as the world's leading provider of AI hardware. China has attempted to innovate in a variety of ways. Last year, the Chinese start-up DeepSeek unveiled a state-of-the-art AI model that used a very small percentage of specialist AI chips.
President Trump has levied tariffs and occasionally put restrictions on the sale of AI chips in order to keep China from catching up. However, China's use of CPUs—standard microprocessors—instead of GPUs to build an ultrafast supercomputer raises the possibility of overcoming those obstacles.

Jimmy Goodrich, a senior scholar at the University of California Institute on Global Conflict and Cooperation, stated that "the US government should have stronger controls on the export and manufacturing of CPUs for the China market." "The present regulations have a loophole."
Since the 1960s, the largest scientific machines, known as supercomputers, have been employed for jobs including developing nuclear weapons, cracking codes, and modeling the climate. They usually express integers using 64 bits of data and employ high-precision mathematics.In comparison, commercial AI systems from firms like Google and OpenAI can operate even more quickly.

They rely on what are known as four-bit and eight-bit numbers, which enable the computers to perform numerous simpler computations simultaneously, and they can employ approximations for tasks like detecting photos or choosing the next word in a sentence.
Mr. Goodrich stated, "It is noteworthy and impressive what China has done here, but they can't hold a candle to these massive AI supercomputers that have been built by American AI labs."
The primary purchasers of some of the biggest supercomputers are US national labs, who are keen to deploy AI to speed up certain parts of their scientific research. Thus, in addition to the 64-bit version, they are using more of these less accurate computations.

Foreign systems have occasionally risen to the top, despite the fact that US corporations have traditionally dominated the ranks of the very largest supercomputers. For instance, a system in Japan topped the list between 2020 and 2022.
According to Addison Snell, an analyst at Intersect360 Research, a company that monitors the industry, "there is a lot of talk that America is the only country capable of these systems." "Then you discover that other businesses are also capable."
The Department of Energy and other US agencies have consistently pushed for increased funding for supercomputers because to powerful systems from China and Japan. The Trump administration launched Genesis Mission in November with the goal of utilizing supercomputers at US government labs and private businesses to accelerate scientific and artificial intelligence development.

In the recent competition for supercomputers, GPUs—mainly created by Nvidia and Advanced Micro Devices—have been an essential tool. These chips are excellent at performing multiple tasks at once, such as matrix multiplication, which is essential to many AI activities, and so-called vector calculations, which are utilized in research.
According to Dr. Dongarra, China was forced "to invest in developing architectures and technology to effectively have supercomputers that are at the same level as the US's highest-performing systems" when US officials restricted exports of certain machines used to manufacture the most sophisticated semiconductors and restricted access to GPUs and other powerful chips.

Unlike most high-end systems, China's LineShine system does not divide the conventional functions of microprocessors and GPUs. Rather, it incorporates GPU-style operations with specialized hardware that speeds up vector and matrix computations. Ninety hardware cabinets contain chips with a total of almost 14 million processing cores, or tiny electronic brains, that are equipped with this capability.
These chips are a unique design based on a set of instructions licensed by Arm Holdings, a British business under the management of SoftBank, a Japanese conglomerate. Although Arm's technology is best recognized for powering smartphones, Nvidia, Amazon, Qualcomm, and other companies have recently modified it for use in data centers.

Arm has been doing business in China for a long time. According to a business spokesperson, "Arm operates globally, including in China, in compliance with applicable export control laws and regulations."
According to Dr. Dongarra, LineShine's creators, who are seasoned Chinese supercomputer experts, have not revealed information regarding the manufacturer of the chips or the degree of chip production technology employed.
Although Chinese laboratories had not yet submitted test results, he and other experts had long believed that China had systems capable of a No. 1 rating.
Mr. Snell remarked, "I'm not surprised that there is a Chinese machine capable of being No. 1." "It's surprising that they desired recognition."

Dr. Dongarra, who produced a thorough report on the new system, stated that the designers believed it was OK to submit tests for the Top500 ranking because the system had been created without government financing while in China.
According to Dr. Dongarra, the Shenzhen scientists have also submitted 14 applications for the Gordon Bell Prize, which encourages the solution of complex scientific problems, in an effort to gain recognition for the new machine. In addition to three systems for a related climate science prize, three systems are finalists for that award.
According to Dr. Dongarra's study, LineShine has been utilized for projects such as a complicated simulation of the human brain and a sophisticated simulation of Earth with elements of the atmosphere, ocean, land, and ice.