In July, Japan’s Ministry of Internal Affairs and Communications released the 2025 edition of the Information and Communications White Paper. Since then, in a country once known for “building its nation on technology,” various sectors of society have been engaged in serious reflection over how Japan’s artificial intelligence (AI) research and applications have fallen behind those of the world’s leading countries, such as the United States and China.
In fact, Japan did not get a late start in AI research. Japanese media have even described the country as a “pioneer” during the first AI boom of the 1950s to 1970s, playing a “foundational role.” So why has Japan gradually lost pace in the global AI race since then? What systemic factors are holding it back? And how do different sectors in Japan view the future of their country’s AI development?
Japan’s AI research entered a blank period after the 1990s
According to Japan’s Ministry of Education, Culture, Sports, Science and Technology, Japanese researchers made significant contributions to AI development as early as the first AI boom from the 1950s to the 1970s. In the 1960s, Japanese universities began working on natural language processing, speech recognition, and image processing. By the 1970s, the scope of research expanded further to include Japanese-language information processing.
In the 1980s, computer systems lacked the ability to automatically collect and accumulate information. The process of inputting knowledge into expert systems was extremely cumbersome, which limited their practical use. To solve this problem, Japan launched a national project in 1982 called the “Fifth Generation Computer Systems Project,” with a budget of about 54 billion yen (roughly 220 million U.S. dollars at the time). The project, led by Japan’s Ministry of International Trade and Industry, aimed to develop within 10 years a “supercomputer” capable of thinking, understanding speech, reading text, reasoning, and learning.
Japan’s initiative triggered a sense of crisis in Europe and the United States, leading them to launch their own competitive projects. However, because a large amount of knowledge had to be manually entered into the system and the management of such a massive project was extremely difficult, this ambitious plan ultimately failed to deliver the expected results. It was terminated in 1992, after which Japan’s AI research entered a long period of stagnation.
Professor Mitsuru Ishizuka of Waseda University pointed out that in the course of internet development, technology grows stronger through use, evolving continuously. This “the more it’s used, the stronger it gets” cycle was the key to the internet’s success. The Fifth Generation Computer Project did produce some results, but since they were not widely applied, that success cycle never materialized.
From the 1990s to the early 2000s, while the United States steadily advanced in AI, Japan took a different path. The spread of the internet sparked a global revolution in information technology, but many Japanese policymakers and business leaders failed to fully grasp the importance of this transformation. In 1995, the United States experienced an internet boom with the IPO of Netscape, followed by the rise of Amazon, Google, and other companies. In contrast, Japan remained dominated by a mindset that “manufacturing is central,” with software and services undervalued. Large Japanese electronics manufacturers focused on digital consumer electronics and embedded systems, while overlooking the opportunities of building platform businesses through the internet.
Realizing it was falling behind in the internet revolution, the Japanese government enacted the Basic IT Law in 2001 and launched the e-Japan Strategy, aiming to make the country “the world’s most advanced IT nation” by 2005. Yet the core problem remained unresolved: although infrastructure was improved, it did not give rise to innovative services or companies that could fully leverage it. Japan remained stuck in a state of “shell without substance.” During this period, the failure of the Fifth Generation Computer Project also cast a long shadow, as AI research budgets were drastically cut, driving many researchers to shift to other fields. While American institutions such as Stanford and MIT continued to push forward with AI research, Japan’s AI research was almost at a standstill. At the corporate level, major Japanese IT companies like NTT, Fujitsu, and NEC prioritized improving the efficiency of existing businesses and developing new hardware over AI technologies. This blank period planted the seeds for Japan’s decline in AI competitiveness later on.
It was only in recent years, with breakthroughs represented by deep learning, that AI began attracting renewed attention in Japan. This was reflected in the release of the Artificial Intelligence Technology Strategy in 2017, which focused on technology and industrial roadmaps; the Artificial Intelligence Strategy 2019, which emphasized broad applications of AI; and the revised Artificial Intelligence Strategy 2022, released in 2022.
Shortage of “digital labor,” high legal barriers, and conservative capital
“Doraemon’s homeland embraces new technology in theory, but in practice has failed to truly put it into action,” The Economist reported. In fact, Japan has seen AI applications in various fields: construction companies use it to improve on-site efficiency; convenience stores have introduced AI clerks; and tech investors are showing strong interest in areas where AI intersects with hardware—a field where Japanese companies have traditional strengths. However, Masayoshi Son, chairman of SoftBank Group and a prominent Japanese tech investor, criticized in October last year that Japanese firms are only focusing on small-scale systems instead of striving to build AI giants. As a result, Japan has missed opportunities in AI development.
“It seems like there are new AI breakthroughs from the U.S. or China almost every day… yet Japan, long known as a nation that built its strength on technology, is struggling to keep pace,” The Japan Times reported in March this year. Between 2019 and 2023, Japan released no large-scale models, while the U.S. launched 182 and China 30. It was not until 2024 that Japan produced its first unicorn company, SakanaAI. Local models such as OpenCalm and Rinna performed worse than OpenAI’s outdated GPT-3.5. At top-tier conferences such as NeurIPS, the share of papers with Japanese lead authors fell from 7% in 2012 to just 2% in 2023. Japan only completed the process of phasing out 3.5-inch floppy disks last year, even though these disks had largely disappeared from the Chinese market around 2010.
According to the 2025 edition of the Information and Communications White Paper published by Japan’s Ministry of Internal Affairs and Communications, in fiscal year 2024 only 26.7% of Japanese citizens had used generative AI, compared to 81% in China and 68.8% in the U.S. Among companies, only 49.7% had established AI usage policies—far lower than 84.8% in the U.S. and 76.4% in Germany. When asked why they do not use generative AI, over 40% of respondents said it was “unnecessary in daily life or work,” while nearly 40% said they “don’t know how to use it.”
Japanese society has been widely reflecting on the deeper reasons behind its lag in AI. Population aging and declining birthrates are the most frequently cited factors. On one hand, the labor shortage is pushing demand for AI applications; on the other, limits in talent supply and social acceptance are constraining AI development. Nikkei noted that population decline directly leads to a shortage of the “digital labor force” needed for AI progress.
However, some analysts argue that Japan’s low AI adoption rate cannot be fully attributed to demographic issues. According to the Information and Communications White Paper, even among young Japanese aged 20 to 29, the AI usage rate was only 44.7%. For those aged 30 to 39, the rate was just 23.8%, which was even lower than the 29.6% seen among the 40 to 49 age group.
This has drawn attention to the lack of AI education in Japan. As The Japan Times reported, Japan has generally shown an open attitude toward AI technology. A survey conducted in June last year by French polling firm Ipsos, covering 32 countries, found that Japan had the lowest proportion of respondents saying “AI products and services make me feel uneasy.” Yet this enthusiasm is mixed with skepticism: only 25% of Japanese respondents believed AI would improve the economy within the next 3 to 5 years, compared with 72% in China. One reason is unfamiliarity—Japan was the only country surveyed where a majority of respondents admitted they did not have sufficient understanding of AI.
By the end of 2024, only 12 universities in Japan had formally established undergraduate faculties or independent departments named after “data science,” and most curricula still relied heavily on traditional statistics. Japan produces about 2,800 AI-related master’s graduates annually—just one-ninth of the U.S.—and 60% of them specialize in traditional areas such as robotics control. Kazuo Yamazaki, a professor at Keio University, told The Asahi Shimbun: “Our education system is still producing engineers, but AI needs interdisciplinary digital natives.” In 2023, only 18% of Japanese AI startups had teams with interdisciplinary backgrounds, far below Silicon Valley’s 57%. According to Nikkei, about 12% of Japan’s AI researchers were recruited by Silicon Valley companies in 2023, 40% of them being core talent under 35.
“Digital Infrastructure at a Snail’s Pace”—under this headline, Nikkei sharply criticized Japan’s structural challenges in technology adoption. Data from the Ministry of Economy, Trade and Industry show that 91% of Japanese companies spend over 90% of their IT budgets maintaining outdated systems rather than developing new technologies, hindering the rollout of AI.
In addition, high legal and data barriers cannot be overlooked. Japan’s Act on the Protection of Personal Information is among the strictest in the world. Companies must go through three steps—“notification, consent, record”—when handling user data, making it difficult to build efficient training datasets. For example, the ECG-AI system jointly developed by the University of Tokyo Hospital and Fujitsu was proven in clinical trials to predict heart attack risks 48 hours in advance. However, because Japanese law requires strict anonymization of patient data, much of the identifiable information in training data was stripped away, reducing the model’s generalization capability.
On the investment side, Stanford University’s AI Index Report 2024 shows that in 2023, Japan’s private AI venture investment totaled around $700 million, compared with $67.2 billion in the U.S. and $7.8 billion in China. According to a report by the Japan Venture Capital Association last year, less than 10% of Japanese private investment in AI was directed toward foundational models. This conservatism stems from a combination of consumer and corporate risk aversion, resulting in slow commercialization of AI products, long return cycles, and further dampening investor appetite.
“A Mindset Shift Is Urgently Needed”
In Japan, there are also voices with a positive outlook on the country’s AI development. “Japan already has all the necessary conditions to become a leader in AI… but it urgently needs a shift in mindset,” The Japan Times reported. While successive Japanese governments in recent years have consistently voiced strong support for AI, low public trust in government means that entrepreneurs and businesses must take more initiative. The report offered several recommendations: Japanese companies should leverage their historical strengths to turn foreign technological breakthroughs into real products and services while aggressively expanding overseas markets; Japan should increase investment in AI education, drawing lessons from China’s programs for primary and secondary students to cultivate the next generation of talent; and Japan should seize the opportunity created by the tightening of U.S. immigration policies to attract software engineers and researchers from around the world.
“Why Has Japan Been Consistently Losing to China in Robotics?” asked Nikkei in a July article this year. One of the reasons, it argued, is that Japan’s industrial policy lacks a national strategy and long-term vision. The article gave the example that Japan was the first country in the world to propose the concept of a “hydrogen society,” but its approach centered on company-led promotion of passenger cars and reliance on imported hydrogen. By contrast, China treated hydrogen energy as another tool for securing energy security and rapidly became the world’s largest hydrogen producer.
In May this year, Japan passed its first AI-related law, the Act on the Promotion of Research, Development, and Application of AI-related Technologies. Its enactment marked the elevation of AI development to the level of a “national strategic project.” In November last year, the government also announced plans to provide more than 10 trillion yen (about 490 billion RMB) in support to AI and semiconductor-related industries by 2030, aiming to boost private investment confidence through a government-driven approach. Citing an official from the Ministry of Economy, Trade and Industry (METI), Kyodo News reported: “This is the first time the Japanese government has adopted a project-financing approach in industrial policy.”
In addition, METI’s “AI Visual Inspection for Manufacturing” initiative has already been rolled out, with procurement expected to total 2.6 trillion yen between 2025 and 2027. According to Nikkei, the first batch of 120 “hidden champion” companies in the components sector will gain free access to Toyota’s production line data. The government will subsidize 30% of the related costs, while Toyota and other industry giants will cover the remaining 70% through joint procurement.
Whether these moves—legislation, public funding, and joint public-private procurement—will truly enable Japan to break out of its current difficulties in the AI race and achieve both technological leadership and social inclusion remains to be seen.
Leave a comment