Japan unveils "third pole" AI strategy, targets 30% of global AI robotics market by 2040

On March 10, the Council for Japan's Growth Strategy, convened by the cabinet under Japan's Prime Minister, adopted a "winning strategy" framework that names artificial intelligence and AI robotics as one of 17 priority sectors and sets a public target of more than 30% global market share in AI robots by 2040, according to materials released by the Prime Minister's Office and reporting by Reuters and the Japan Times. The same package raises Japan's annual domestically produced semiconductor sales target to 40 trillion yen — roughly five times current levels — and explicitly frames the country as a "third pole" in AI between the United States and China.

The council, formally relaunched under the current cabinet in late 2025, is the successor to the New Capitalism Realization Council and serves as the Prime Minister's top forum for industrial policy. Tuesday's session was its third under the new chair and ran through 61 specific products and technologies — from edge AI inference accelerators to humanoid robot actuators — that the government will prioritise for direct subsidies, procurement and tax incentives. The Yomiuri Shimbun reported that the AI robotics subgroup, jointly led by METI and the Cabinet Office, had been working since December 2025 to map where Japan retains a manufacturing edge that can plausibly translate into AI-era leadership.

According to the official summary published by the Prime Minister's Office, the Prime Minister's Office instructed the relevant ministers to "swiftly advance the formulation of a Public-Private Investment Roadmap" and to clarify, for each of the priority products, the "scale and timing of induced domestic investment." The Office also explicitly requested that the Minister of Economy, Trade and Industry and the Minister of Defense cooperate to translate emerging technological seeds into defense procurement. The kantei summary does not include a verbatim ministerial quote about the "third pole" framing; that wording appears in the council's own documents and is repeated by the Yomiuri Shimbun and the Japan Times in their coverage of the meeting.

The numerical envelope is sizable. Reuters cites a council target to lift Japan-produced semiconductor sales from roughly 8 trillion yen today to 40 trillion yen by 2040, while the Yomiuri Shimbun puts the AI-robotics market objective at over 30% global share — a level that would imply Japanese suppliers capturing on the order of 20 trillion yen annually in 2040 prices on the council's own demand assumptions. NEDO's existing five-year, 1 trillion yen ($6.3 billion) AI fund, opened to applications in fiscal 2026, is positioned as the first major instrument of this strategy. By comparison, the U.S. CHIPS Act and the EU Chips Act both target only 20% global semiconductor share for their respective blocs by 2030.

Industry reaction was guarded. Foreign-affairs analysts at Nikkei Asia argued that Japan's industrial structure — capital-heavy, slow on talent mobility — remains the main implementation risk for any "third pole" ambition, regardless of how much capital the council steers. Reuters' equity desk noted that listed semiconductor names rallied modestly on the 40 trillion yen figure. Reporting from the Yomiuri Shimbun and the Asahi Shimbun has separately linked the framework to a planned Physical AI joint venture among SoftBank, NEC, Sony Group and Honda intended to apply for support under the NEDO program — a corporate vehicle widely read as the operational spearhead of the council's "third pole" framing.

For us at Enpo Sekai, the value of this announcement is less the headline market-share number than the underlying admission that Japan is choosing to compete above the silicon and below the application — at the foundation-model and physical-control layer. As a Tokyo-based studio building character engines, voice and persona for desktop and consumer products, we sit one layer above where the council is concentrating capital. A credible "third pole" with locally hosted compute and Japanese-trained foundation models would, over time, lower our cost of running character agents on local-first desktop builds, and would give our B2B engine licensing customers a domestic alternative they can defend in front of their procurement and legal teams.

We will be watching three things over the next twelve months: (1) whether NEDO's 1 trillion yen AI fund actually disburses to non-mega-cap recipients, since on past form Japanese industrial funds tilt heavily toward the largest existing players; (2) what licensing terms the SoftBank-NEC-Sony-Honda physical AI venture eventually attaches to its model — open weights, commercial license, or closed API — because that determines whether mid-sized character studios like us can build on it without rebuilding our entire stack; (3) how the council reconciles its energy assumptions, given that 30% global share in AI robotics implies a domestic compute footprint Japan cannot serve under current grid constraints.