At a newly built factory in southwestern Japan, a robot makes a mistake — and then corrects it.
At Yaskawa Electric’s new 20 billion yen ($126 million) factory building in Kitakyushu, one of the company’s robots fails to fasten a screw. Instead of stopping and waiting for human intervention, it reverses its tool, removes the part, adjusts its angle and tries again. The second attempt succeeds.
The episode offers a practical illustration of the technology at the centre of Prime Minister Sanae Takaichi’s economic strategy: “physical AI”, in which artificial intelligence is integrated with machines capable of interacting with the physical world.
The factory has about 110 robots, including 35 powered by artificial intelligence and performing work that previously required people. Together with a new control system, the machines have more than doubled productivity, according to Yumie Kubota, a director and general manager of Yaskawa’s AI robot division, who spoke to Nikkei Asia. The screw-fastening process can now be completed almost 100% of the time.
For Takaichi’s government, such developments represent more than advances in manufacturing. They form part of a much larger economic wager: that targeted government support can stimulate domestic investment, raise productivity and enable Japan to grow faster than its enormous public debt.
The strategy, known as the Honebuto no Hoshin, or “big-boned policy”, seeks to mobilise more than 370 trillion yen ($2.3 trillion) in public and private investment in priority sectors by fiscal 2040. The documents, however, do not provide a complete breakdown between public and private investment or specify how much the government will ultimately have to spend.
The scale of the ambition has already unsettled financial markets. Long-term Japanese government bond yields surged to levels not seen in decades after a draft of the policy was released at the end of June. Investors questioned whether the promised economic growth would materialise before higher borrowing and interest costs place additional pressure on public finances. The concerns have also contributed to downward pressure on the yen.
The government’s objective is to put Japan’s debt-to-GDP ratio, currently about 190%, on a sustained downward trajectory by 2040. Achieving that goal depends heavily on faster economic growth. The strategy assumes average potential growth will rise to 1.8% during 2036-2040, from only 0.4% between 2021 and 2025.
That would require a dramatic improvement in productivity. Growth in total factor productivity, which measures the additional output generated through more efficient use of labour and capital, would have to more than double from its 2021-2025 average of 0.5%.
The government has identified 17 sectors and 62 products and technologies as priorities, including artificial intelligence, semiconductors, biotechnology, defence, energy and shipbuilding. At the heart of the plan is an “AI transformation”, with physical AI expected to play a role across all 17 sectors. The objective is to create an economy capable of generating greater value even as Japan’s population declines.
Technology companies are already developing components of that vision. SoftBank Corp. is working on AI-RAN, which would equip mobile base stations with graphics processing units to provide robots with additional computing power through low-latency networks. The technology could allow industrial machines to run newer AI models without frequent hardware upgrades.
The government is also supporting Noetra, a corporate consortium developing a Japanese foundational model designed to understand space and the physical world. State support is expected to reach about 1 trillion yen through fiscal 2030.
Ren Ito, an AI specialist at the Japan Research Institute, considers the government-led effort both necessary and welcome. Japan has engineers and companies capable of developing AI models, he argues, but has lacked sufficient investment and a common platform capable of bringing those efforts together.
The strategy reflects a broader diagnosis of Japan’s prolonged economic stagnation. According to Takuji Aida, chief economist for Credit Agricole Securities in Japan and a member of the cabinet’s growth strategy council, the central problem is chronic underinvestment rather than simply the country’s shrinking population.
Aida argues that investment can become self-reinforcing. As one company spends to develop new markets or technologies, competitors may follow, generating further investment and demand. “The impact of the initial investment expansion should be significant” for productivity growth, he said.
But whether the investment will materialise at the scale envisioned remains uncertain. Keiji Kanda, chief economist at the Daiwa Institute of Research, questions both the amount of investment the strategy will generate and whether the government can meet its productivity target. Japan has attempted similar objectives before. The government of former Prime Minister Shinzo Abe targeted total factor productivity growth of 1.8% by the early 2020s, but the average was just 0.4% between 2020 and 2023.
There are also questions about the consequences for employment. Saisuke Sakai, a senior economist at Mizuho Research Institute, argues that greater attention should be paid to which sectors will adopt AI, how automation will affect employment and how surplus workers will be redeployed.
Some critics challenge the strategy at a more fundamental level. Richard Katz, author of “The Contest for Japan’s Economic Future,” describes it as a return to the industrial policies of the 1950s and 1960s. Japan’s experience with initiatives such as Japan Display, Elpida Memory and the Cool Japan Fund, he argues, offers little reason for confidence.
For Katz, the central difficulty is not necessarily developing new technologies but converting them into successful businesses. Japan, he says, suffers from the “valley of death” separating laboratory development from commercialisation.
Supporters reject the comparison with earlier industrial policy. Aida describes the 62 priority areas as a diversified portfolio of sectors with credible “winning paths”, while cabinet adviser Masahiko Hosokawa says the selections were made following consultations with industry and academia.
Hosokawa, a former Ministry of Economy, Trade and Industry official and now a professor at Meisei University, also emphasises public procurement. Government demand in areas such as disaster response, construction and defence could create early markets for emerging technologies, encouraging companies to invest. “When markets can be expected, suppliers are willing to invest,” he says. “Simply telling them to invest is not enough.”
Artificial intelligence may provide Japan with an opportunity precisely because the technological race is not yet settled in every field. Ito says Japan has been “overwhelmingly outpaced” by the United States and China in large language models, but argues that physical AI has yet to establish a dominant architecture. That leaves Japanese manufacturers with an opportunity to compete from an existing position of industrial strength.
SoftBank Corp. President Junichi Miyakawa has described the opportunity as Japan’s “first and last real chance” to secure a position in the AI race.
Yet capital alone may not determine the outcome. Aniket Shah, a managing director at Jefferies and an expert on industrial policy, argues that investment funding is generally not the binding constraint. Permitting, siting rules, labour constraints and regulations can be equally decisive in determining whether companies actually invest.
The stakes are particularly high because the government’s fiscal strategy depends on the assumption that investment-led growth will eventually stabilise Japan’s debt burden. A July survey of 50 prominent economists conducted by Nikkei and the Japan Center for Economic Research found that 70% believed the government was unlikely to achieve a sustained decline in the debt-to-GDP ratio.
The Honebuto documents allow for a temporary deterioration in the primary balance when strategic investment requires it, representing a shift from previous governments’ emphasis on balancing revenue and spending excluding interest payments.
Takaichi is instead betting that faster economic growth will cause gross domestic product to expand more rapidly than debt. The government says its contribution will be financed through regular revenue, spending reviews and bridge bonds, although the precise amounts remain undecided.
Financial markets, however, have already signalled that patience may not be unlimited. The sharp increase in bond yields following publication of the draft policy demonstrated investor concern about the fiscal implications. Kanda argues that, even while pursuing investment-led growth, the government should at least maintain a positive primary balance to demonstrate what he describes as “responsible fiscal policy”.
At Yaskawa Electric, the companies that are expected to deliver the physical AI transformation have their own view of what government support should accomplish.
Kubota welcomes the government’s emphasis on physical AI as a “tailwind” for the industry. But she says discussions she attended, including meetings convened by METI, focused heavily on supporting manufacturers developing the technology.
For Yaskawa, that is not the main obstacle. The company already plans to invest 250 billion yen over the next four years, including 120 billion yen in physical AI. Developing and selling such systems, Kubota argues, is part of the normal business of manufacturers.
Government assistance, she says, should instead reach companies willing to become early adopters in sectors where there are few precedents and financial returns remain uncertain. Those companies, she argues, are assuming the greater risk.
“Support the companies willing to be the first to try it,” Kubota said. “Then we can do business as usual.”
That distinction captures the central question facing Takaichi’s $2.3 trillion industrial-policy gamble. Japan possesses advanced manufacturers, engineers and technologies, but turning those strengths into sustained economy-wide productivity growth will depend on whether government investment can create markets and encourage private companies to take risks without leaving taxpayers and financial markets carrying an unsustainable burden.
The robots in Kitakyushu demonstrate that productivity gains are technically possible. The harder question is whether Japan can reproduce that success across an economy large enough to transform growth prospects — and quickly enough to tame a debt burden already standing at roughly twice the size of the economy.
