Japan is preparing a major push to build domestic AI capacity, with a reported funding plan that would back a new company tasked with training a top-tier base model.
📌 Key Takeaways
- Japan is preparing ¥1 trillion in support spread across five years.
- A new company backed by around 10 firms, including SoftBank, is planned.
- The funding window is expected to start in fiscal 2026, beginning April 2026.
- Roughly 100 engineers from SoftBank and Preferred Networks may join.
- Access to chips and large-scale compute will likely decide the outcome.
What The ¥1 Trillion Plan Looks Like So Far
A report citing a source close to the matter says Japan will provide roughly ¥1 trillion (about US$6.34 billion) in support over five years for a planned new domestic AI company.
That company is expected to be formed by around 10 firms, including SoftBank, with the stated goal of developing the country’s largest base AI model via public-private cooperation.
The same source says the support scheme would begin in fiscal 2026, which starts in April 2026, and the new entity could include about 100 engineers drawn from SoftBank and Preferred Networks.
Why Japan Keeps Coming Back To “AI Sovereignty”
The logic is simple: base models sit under everything, from consumer assistants to government workflows, and they shape what a country can build, audit, and control.
Japan’s broader AI policy discussions have explicitly framed overdependence on foreign AI as a national security concern, alongside risks like disinformation and cyberattacks.
There’s also a usage gap that makes the urgency feel political as much as technical: government-cited data says 27% of people in Japan used generative AI in fiscal 2024, versus nearly 69% in the U.S. and about 81% in China.
The Real Bottleneck Is Chips And Compute
Training a serious base model is not just a research problem, it’s a supply-chain problem, and the report explicitly points to competition over semiconductors as a key constraint.
Japan has already been building policy scaffolding around this, including programs focused on subsidizing access to compute for foundation model development.
“METI will launch GENIAC to provide support in securing computational resources necessary for the development of foundation models.” — Ministry of Economy, Trade and Industry (METI), Japan
If the new funding plan moves forward, it effectively tries to do at national scale what individual AI labs do privately: lock down compute, secure chips, and keep iteration cycles fast enough to stay competitive.
Where Preferred Networks Likely Fits In
Preferred Networks showing up in the staffing detail matters because it already positions itself across the stack, including AI chips, large-scale infrastructure, and a Japan-made foundation model.
In a prior funding announcement, Preferred Networks said it planned to strengthen development of AI semiconductors and its Japan-made generative AI foundation model PLaMo, alongside computing infrastructure.
That kind of capability lines up with the reported goal here: not just building a model, but ensuring Japan can source the ingredients needed to train and run one at scale.
“Preferred Networks’ innovative projects, including its in-house development of next-generation AI semiconductors, will play a key role in and outside of Japan.” — Yoshitaka Kitao, Representative Director, Chairman, President & CEO, SBI Holdings, Inc.
What’s Still Unknown And What To Watch Next
The report is specific on the headline numbers, but several practical details remain unclear, including the new company’s structure, how the ¥1 trillion would be distributed, and what milestones trigger funding.
It’s also not yet known whether the planned “largest base model” is intended to be widely accessible, restricted to strategic sectors, or paired with rules around evaluation, safety testing, and third-party auditing.
The most telling signals will be boring ones: procurement commitments for compute, clear chip supply arrangements, and whether the program produces a model that is actually adopted at home, not just announced.
Conclusion
If the plan holds, Japan is trying to compress years of fragmented effort into a single, funded push: build a domestic base model, staff it with proven engineers, and reduce dependence on foreign AI.
The bet is large, but the success criteria are even larger, and they hinge less on ambition and more on execution, compute access, and whether real users in Japan choose the resulting model.
📈 Latest AI News
22nd December 2025
- The Fastest Way To Make ChatGPT Sound More Professional
- 13 y/o Expelled After Defending Herself Over AI-Generated Nude Images
- Expedition 33 Disqualified From Indie Game Awards Over AI Use
- How to Check If a Video Was Made With AI Using Gemini
- OpenAI Launches ChatGPT App Store: Browse, Connect & Use Apps
For the recent AI News, visit our site.
If you liked this article, be sure to follow us on X/Twitter and also LinkedIn for more exclusive content.
