China's LineShine Becomes the World's Fastest—Not a "Supercomputer Defeat" for Japan, but the Start of a "Computational Resource War"

China's supercomputer "LineShine" has claimed the top spot in the June 2026 edition of the TOP500, the global high-performance computer ranking. It is installed at the National Supercomputing Center in Shenzhen, China. Its sustained performance is 2.198 exaflops. This surpasses the previous leader, the U.S. Lawrence Livermore National Laboratory's "El Capitan," marking China's return to the world's fastest position for the first time in about nine years.

This news is not merely about "China building a fast supercomputer." For Japan, it is a warning that touches on science and technology, semiconductors, security, AI, energy policy, and industrial competitiveness.

An article from the Xinhua-affiliated Borneo Post positions LineShine's achievement as "evidence that U.S. technological regulations could not halt China's progress." The U.S. has long imposed export restrictions on Chinese supercomputing-related organizations, aiming to curb the growth of China's computational capabilities by limiting access to advanced CPUs, GPUs, semiconductor manufacturing equipment, and AI chips. However, this time, China has achieved the world's top position with a domestically developed CPU-centric architecture, rather than a U.S.-dependent GPU-based design.

Here lies the essence that Japan must not overlook. Amid U.S.-China tensions, China did not merely "evade regulations." Instead of the path closed by regulations, China has been digging another path on a national scale. If the result is LineShine, Japan should not simply conclude with "China is amazing" or "U.S. regulations failed." Japan is being asked to consider how much of its computational infrastructure it can maintain independently, how much to share with allies, and what to keep domestically.

The Shock of LineShine Lies in "Without GPUs"

The most notable aspect of LineShine is that it achieved the world’s top spot with a CPU-centric configuration without using GPUs.

Since the generative AI boom, the global competition for computational resources has been centered around NVIDIA GPUs. The ability to line up tens of thousands of AI-focused GPUs like H100, H200, and B200 has influenced the development power of large AI models. In Japan, securing GPUs has become a major policy issue in establishing a generative AI foundation.

However, LineShine chose a different path from the current mainstream AI data centers. It lined up a large number of CPUs, connected them with a high-speed proprietary network, and completed a system strong in scientific and technological calculations. In the HPL performance measured by TOP500, this surpassed the U.S.'s El Capitan.

This fact provides Japan with two insights.

First, there are multiple winning strategies in the competition for computational power. Even countries or companies that cannot procure a large number of GPUs can achieve world-class performance in specific areas by comprehensively designing architecture, networks, software, memory, and cooling.

Second, relying solely on GPUs makes a nation's computational infrastructure fragile. GPUs are strong in AI, but their supply is limited, prices are high, and they are prone to export restrictions. If Japan is to establish a research and industrial foundation for the AI era, it must consider how to create a "sovereign computational infrastructure" that includes not only GPU clusters but also CPUs, accelerators, networks, storage, and software.

However, It Cannot Be Said That "China Is Also Number One in AI"

While the news of LineShine is flashy, overestimation is also dangerous.

The HPL benchmark of TOP500 is a traditional indicator for scientific and technological calculations, mainly measuring double-precision floating-point operations. It is extremely important for climate simulation, fluid analysis, material science, seismic analysis, drug discovery, and nuclear fusion research, but it does not directly indicate the learning performance of generative AI.

In AI, low-precision operations like FP8, BF16, and INT8, communication efficiency when distributing large models, GPU interconnections, memory bandwidth, and software ecosystems are crucial. Reuters and others report that while LineShine is at the top of traditional HPC rankings, a different perspective is needed for AI-oriented evaluations.

Reactions on social media are also divided on this point. While there are voices praising China's technological independence, there are also many calm observations such as "Being number one in HPL is not the same as being number one in generative AI," "Aren't the private AI clusters not appearing in TOP500 actually larger?" and "Google, Microsoft, Amazon, xAI, and other undisclosed clusters are outside the rankings."

For Japan, it is crucial to correctly understand this distinction. Being number one in supercomputers and being number one in generative AI infrastructure are not the same. However, both are fundamental to national competitiveness. Japan cannot be complacent with "We have Fugaku, so we're fine" or resigned with "We're doomed because we lack GPUs." As the boundary between HPC and AI dissolves, a design philosophy that connects both is necessary.

Reactions on Social Media—The "Sanctions Are Counterproductive" Argument and Questions About "Power Efficiency"

 

Reactions on social media and tech forums can be broadly divided into four categories.

First, there is the view that "U.S. sanctions may have been counterproductive." As a result of the U.S. tightening advanced semiconductor exports to China, China has moved towards combining domestic chips, proprietary networks, and domestic OS instead of relying on foreign-made GPUs or CPUs. The argument is that while sanctions may have caused short-term pain, they may have accelerated self-reliant development in the long term.

Second, there is technical astonishment at "How far can you go with just CPUs?" In tech communities like Reddit and LinkedIn, there is interest in the achievement of surpassing 2 exaflops without GPUs. Users are viewing it not just as political news but as an architectural experiment, focusing on Arm-based CPUs, a large number of cores, proprietary interconnects, memory configurations, and ease of programming.

Third, there is cautiousness with "It's a different issue for AI applications." In the world of generative AI, NVIDIA GPUs and the CUDA ecosystem remain strong. Even if LineShine is strong in scientific and technological calculations, it's a different story as to how competitive it is in training and inference of large language models.

Fourth, there is the reaction that "Power consumption is too high." LineShine's power consumption is said to be about 42.2 megawatts. This is the price of being the world's fastest. As the competition for supercomputers and AI data centers progresses, power, cooling, water resources, location, and power grids become bottlenecks. On social media, there are voices saying, "We should look at performance per watt, not just performance," and "Who will bear the cost of data center power?"

From a Japanese perspective, this power issue is extremely important. Japan has high electricity costs, limited land, and constraints on data center locations. To have AI and HPC infrastructure domestically, discussions must include not only semiconductors but also power policy, renewable energy, nuclear power, power grids, cooling technology, and decentralized data centers.

How Japan Should Update the Success Experience of "Fugaku"

Japan has "Fugaku." Developed by RIKEN and Fujitsu, Fugaku achieved high rankings in TOP500, HPCG, HPL-AI, and Graph500 in the early 2020s, supporting Japan's computational science. It has produced socially understandable results such as droplet simulation of the novel coronavirus, drug discovery, weather, materials, and industrial use.

However, the emergence of LineShine tells Japan that merely extending the success experience of Fugaku is not enough.

Fugaku's strength was not just in rankings. It was easy to run a wide range of scientific and technological applications on a CPU basis, providing versatility for researchers. This overlaps with LineShine's CPU-centric design. In other words, Japan had already demonstrated the value of HPC that is not solely GPU-dependent.

On the other hand, it is now the era of generative AI and AI for Science. The trend of AI supporting literature search, hypothesis generation, simulation, experimental planning, robotic experiments, and data analysis is strengthening. Next-generation supercomputers need to support not only traditional simulations but also AI model training and inference, scientific data integration, and the automation of the entire research process.

RIKEN and Fujitsu's "Fugaku NEXT" becomes extremely important in this context. What Japan should aim for is not just temporarily surpassing China or the U.S. in TOP500. It is to create an environment where Japanese researchers, universities, companies, and startups can use world-class computational resources domestically.

The Implications for Japanese Companies—Not Separating Semiconductors and Cloud

The news of LineShine poses a heavy question for Japanese companies.

Japan has strengths in semiconductor manufacturing equipment, materials, components, precision processing, power supplies, cooling, optical communication, and data center operations. However, in advanced logic semiconductors, GPUs, AI software infrastructure, and large clouds, it heavily relies on the U.S., Taiwan, South Korea, and China.

In Japan, semiconductor policy, AI policy, HPC policy, and cloud policy have tended to be discussed separately. However, what LineShine has shown is that it is no longer the time to consider these separately. Even if you have chips, they are useless without software. Even if you have a cloud, it cannot be expanded without power. Even if you have AI models, they cannot be linked to industry without research data and computational resources.

Japanese companies cannot secure long-term competitiveness by merely buying NVIDIA GPUs and creating AI services. Of course, GPU procurement is important, but relying solely on it will leave them at the mercy of supply constraints and price fluctuations. Japan's winning strategy lies in connecting data from real industries such as manufacturing, materials, robotics, medical, weather, disaster prevention, drug discovery, power, and mobility with high-performance computational infrastructure domestically.

Computational Power as National Security

Supercomputers are research facilities and security assets at the same time.

High-performance computing is used for weather forecasting and drug discovery, but it is also related to nuclear weapon simulation, hypersonic weapons, cryptanalysis, military AI, and satellite data analysis. The U.S. has regulated Chinese supercomputing-related organizations due to this dual-use nature.

Japan is an ally of the U.S. and is also economically deeply connected with China. Therefore, the discussion surrounding LineShine is difficult. Simply viewing China's technological development as an enemy risks damaging scientific and technological cooperation and economic relations. On the other hand, the reality that computational resources are linked to military power cannot be ignored.

Japan's stance should be neither excessive optimism nor excessive fear. It should clearly separate areas where research cooperation is possible from those that need management. In areas like climate change, disaster prevention, infectious diseases, and basic science, international cooperation is highly valuable. However, in AI, advanced semiconductors, cryptography, and satellite analysis with high military applicability, rule-making with allies and domestic management are indispensable.

Three Issues Japan Must Consider Immediately

In light of LineShine's emergence, there are three issues Japan must consider.

First, securing computational resources domestically. It is dangerous to rely entirely on overseas clouds for the computational power needed for AI and scientific research. To handle research data, medical data, industrial data, and security-related data, a reliable HPC and AI infrastructure is needed domestically.

Second, power and data center policy. The global AI competition ultimately becomes a power competition. If large-scale computational infrastructure is to be placed domestically, power supply, transmission, cooling, location, and disaster measures must be integrated. Concepts of decentralizing data centers to rural areas cannot progress without power and communication design.

Third, talent and software. Buying a supercomputer is not the end. To actually extract performance, talent in parallel computing, numerical analysis, AI, compilers, networks, storage, and application optimization is needed. The value of Fugaku has been supported not only by hardware but also by the researchers and software assets that use it.

What Is More Important Than "Being Number One in the World"

LineShine's world number one is indeed big news. However, what is important for Japan is not whether it has been overtaken by China. It is whether Japan can possess a computational infrastructure directly linked to its social issues and industrial competitiveness, rather than just aiming to be number one in the world.

As a disaster-prone country, Japan needs simulations for earthquakes, tsunamis, heavy rains, typhoons, and volcanic eruptions. For an aging society, drug discovery, medical AI, and genome analysis are necessary. For manufacturing, advanced simulations for material development, semiconductor design, batteries, aerospace, and robotics are needed. For energy policy, power grids, nuclear fusion, and renewable energy demand forecasting are required.

All of these cannot progress without computational resources. In other words, supercomputers are not just for researchers but are the foundation of national life and industry.

China's LineShine is being talked about as a symbol of overcoming U.S. regulations. However, the real lesson for Japan is not about the pros and cons of regulations. It is the reality of how significant investment and technological integration occur when a nation seriously views computational infrastructure as a strategic asset.

Japan is a country that belongs to the U.S. technological sphere while neighboring China. Therefore, it is not enough to just watch the victories and defeats of either side. It is necessary to connect Fugaku NEXT, domestic AI infrastructure, semiconductor policy, power infrastructure, and research talent into a single strategy.

The emergence of LineShine is both a declaration of victory for China and a question for Japan.

Will Japan remain on the "user side" in the next computational power competition? Or will it remain on the side that designs, operates, and connects computational infrastructure to industry and science?

There is not much time left to come up with an answer.

---

Source URL

Borneo Post. Confirmed that LineShine reached the top of the TOP500 under U.S. technological regulations, the overall tone of the article, the description of 2.198 exaflops, and claims regarding international cooperation.
https://www.theborneopost.com/2026/06/29/tech-curbs-fail-to-stop-chinas-supercomputing-rise/

TOP500 June 2026 edition ranking. Confirmed LineShine's ranking, Rmax, number of cores, power consumption, and comparison with El Capitan.
https://top500.org/lists/top500/list/2026/06/

TOP500 official news. Confirmed that LineShine became the new number one in the 67th TOP500, becoming the first China-based system to top the list since 2017.
https://top500.org/news/lineshine