This is Hamamoto from TIMEWELL.
AI for Science is the effort to make AI "a tool for scientific research itself." It is not about summarizing papers or streamlining clerical work; it is about embedding AI into the core research process — forming hypotheses, designing experiments, finding laws in data — and changing the speed of discovery by an order of magnitude. Just as AlphaFold's protein-structure prediction led to a Nobel Prize in Chemistry, science worldwide is already moving in this direction.
Japan, too, has begun moving in earnest. On March 31, 2026, MEXT formulated the "Basic Strategy for Advancing AI for Science"[^1], setting out to achieve a "Revival of Science" over the five years from FY2026 to FY2030 — the period of the 7th Science, Technology and Innovation Basic Plan. This article draws on primary sources to lay out the substance of the strategy, along with SPReAD 1000, the open-call program that is immediately relevant to any researcher.
MEXT's strategy: reading it through four pillars
Cross-referencing the strategy with the budget materials, the program organizes into four pillars[^1][^2].
| Pillar | Content |
|---|---|
| AI foundation models and AI agents | Development of Japan-originated, cutting-edge foundation models for scientific research, and of AI agents that advance research autonomously |
| Next-generation AI-driven labs | Development of lab systems that automate the planning, execution, and analysis of experiments with AI and robotics |
| Computing infrastructure | Development and buildout of the supercomputers essential to foundation-model development, including Fugaku NEXT and HPCI systems |
| Research-data infrastructure | Strengthening the infrastructure that supports the storage, management, and distribution of ever-growing research data |
It is worth grasping the scale as well. The FY2026 budget request set aside roughly 35.5 billion yen for advancing innovative research through the use of AI[^3]. An implementation framework combining the FY2025 supplementary budget with the FY2026 initial budget is in place[^2] — a design that does not end with a single year of proof-of-concept.
The pillar I personally find most compelling is the "AI-driven lab." Generative AI proposes hypotheses, robots run the experiments, AI analyzes the results and designs the next experiment, and human researchers concentrate on setting the questions and making judgments. The very structure we have supported in corporate AI adoption — "hand the work to AI so people can focus on judgment" — is now taking the same shape at the frontier of scientific research.
Interested in leveraging AI?
Download our service materials. Feel free to reach out for a consultation.
SPReAD 1000: 1,000 projects open to researchers in every field
Carrying the strategy's broad base is SPReAD 1000 (formal name: the AI for Science Emerging Challenge Research Creation Program)[^4]. As the name suggests, with an extraordinary number of awards — roughly 1,000 projects — the program aims to rapidly grow the ranks of AI for Science practitioners across research sites nationwide.
| Item | Content |
|---|---|
| Eligibility | Researchers affiliated with domestic universities, technical colleges, inter-university research institutes, incorporated administrative agencies, etc. (students may also apply) |
| Support amount | Direct costs of up to 5 million yen per project (indirect costs of 30% to be provided separately) |
| Number of awards | Roughly 1,000 projects across the two open calls combined |
| Eligible uses | Computing resources, data acquisition and usage fees, API costs, equipment, data-management personnel costs, and more |
| Review | An agile, challenge-oriented review method that includes random sampling |
The 2026 schedule ran the first call from April 17 to May 18 and the second from June 2 to July 3, with each moving to award notification within a few months — a fast design[^4]. The review method is what stands out: it incorporates a mechanism for selecting from qualifying proposals through a method that includes random sampling. The idea is not to over-filter emerging challenges through a conventional, heavyweight review. As a piece of institutional design, I think it is a bold choice.
That it is open to every field also matters. Not only life sciences and materials science but all fields — including the humanities and social sciences — are eligible, and the very question of "how do I embed AI into my research" is itself what the program supports.
What researchers and companies should each do
For researchers. The 2026 SPReAD 1000 calls have closed, but the roughly 1,000 awardees will begin their research one after another from here. Since the strategy is a five-year framework, there is value in taking stock now of which stages of your own research could change with AI, ahead of the next open-call opportunity. A scale of 5 million yen is plenty to cover API costs and computing resources and validate an idea on a small scale.
For companies. AI for Science is not just a matter for universities. AI-driven labs need equipment makers and robotics firms; foundation-model development needs those who can provide computing resources and prepare data. If anything, the points of contact for industry-academia collaboration are increasing. And above all, embedding AI into your own R&D division runs in exactly the same direction as this national strategy. If you are an R&D-focused company, I recommend thinking about "where in our lab's experiment planning, data analysis, and literature review should we introduce AI" on the same timeline as the national effort.
At TIMEWELL, we are involved in this "research x AI" frontier through ZEROCK for Deeptech, which supports patent analysis and paper searches in deep-tech fields, and through WARP, our hands-on support for AI adoption including R&D divisions. If you are considering embedding AI into your own research and development process, please feel free to reach out.
Summary
- AI for Science is the effort to embed AI into the core research process to accelerate discovery. MEXT formulated its strategy on March 31, 2026, framing it as a "Revival of Science."
- The program's pillars are foundation-model and AI-agent development, AI-driven labs, computing infrastructure such as Fugaku NEXT, and research-data infrastructure — four in all. The budget request was roughly 35.5 billion yen.
- SPReAD 1000 supports roughly 1,000 projects by researchers across all fields, with 5 million yen each. It features a challenge-oriented review method that includes random sampling.
- Researchers should take stock of their research stages ahead of the next call; companies should advance industry-academia collaboration and embed AI into their own R&D — all on the same timeline.
References
[^1]: MEXT, "Basic Strategy for Advancing AI for Science" (March 31, 2026) [^2]: AI for Science Promotion Committee, "On the FY2025 Supplementary Budget and FY2026 Initial Budget Proposal Related to AI for Science" (February 9, 2026) [^3]: Nikkei, "MEXT to request 35.5 billion yen for 'innovative research through AI,' boosting competitiveness" (August 2025) [^4]: MEXT, "SPReAD 1000 — Unleashing the Potential of Research with AI" (AI for Science Emerging Challenge Research Creation Program)
We also referred to MEXT, "On the Progress of the 'Basic Strategy for Advancing AI for Science'" (May 21, 2026, Cabinet Office CSTP materials).