Japan's ¥370 Trillion Bet: Inside the $2.3 Trillion, 14-Year Plan to Make AI and Semiconductors the Spine of the Economy

On June 24, 2026, after a policy advisory panel meeting, the government of Prime Minister Sanae Takaichi released the documents behind a long-term economic vision that is striking for its sheer scale. The headline figure is more than ¥370 trillion — roughly $2.3 trillion — of combined public and private investment over the 14 years ending in March 2041. Of that, ¥101.6 trillion, just under a third of the whole, is earmarked specifically for artificial intelligence and semiconductors. Slightly less than half of the total is meant to come from government funds, with the remainder from private capital.

The semiconductor ambition is the part to fix in your mind, because it is concrete and measurable. Japan today generates roughly ¥8 trillion in annual domestic chip sales. The plan aims to push that to ¥40 trillion — about $254 billion — by 2040. That is a fivefold increase, and it is not a number conjured from nothing: it sits on top of Rapidus's 2-nanometer foundry effort in Hokkaido, TSMC's operating fabs in Kumamoto, and a decade of METI subsidy programs that have already pulled tens of billions of dollars of leading-edge capacity onto Japanese soil.

And it is not only chips. The same vision names defense, space development, advanced manufacturing, and shipbuilding as priority sectors. Read together, the message is that Tokyo intends to treat advanced industrial capacity — the ability to make the hardest physical things, at scale, at home — as a matter of national strategy rather than market accident. For a country that spent the 2000s and 2010s ceding ground in consumer electronics and memory, this is a deliberate attempt to re-anchor the economy on what Japan still does better than almost anyone: precision manufacturing and the machines, materials, and chemistry that feed it.

Japan's announcement does not land in a vacuum — it lands in the middle of a global race to own compute as national infrastructure. Days earlier, reporting detailed a Chinese plan to spend around 2 trillion yuan (about $295 billion) over five years building a connected network of AI data centers, with state carriers operating most of them and a mandate to source at least 80% of the technology — AI chips included — from domestic suppliers such as Huawei, effectively designing Nvidia out. Fold in power-grid integration and estimates of the total climb toward $740 billion. The United States, for its part, continues to pour private capital into hyperscale buildouts at a pace no government program can match.

Against that backdrop, the strategic logic of Tokyo's bet is unmistakable. When a handful of countries can throttle who gets the most advanced chips, the ability to make them at home stops being an economic nicety and becomes a sovereignty requirement — the same way energy and food security are. The same week underlined why the stakes are real at the application layer too: the global AI-assistant market fragmented, with ChatGPT slipping below 50% share for the first time as Google's Gemini and Anthropic's Claude gained ground. The model layer is consolidating into a multipolar contest, and the hardware beneath it is being nationalized country by country.

What makes Japan's version distinctive is that it does not try to out-spend China or out-hyperscale the US. It plays to a different edge: not the biggest data centers, but the deepest manufacturing stack. Japan already dominates large swaths of the semiconductor supply chain that rarely make headlines — photoresists and specialty chemicals, silicon wafers, deposition and etch equipment, precision optics, and the thermal and power components that every AI rack on Earth depends on. A plan that pours ¥101.6 trillion into that base is not chasing a trend; it is reinforcing a position the country already holds, at the exact moment the world has discovered it cannot do without it.

A ¥370 trillion headline is easy to applaud and easy to overrate. The honest read is that the ambition is sound and the execution is where this will be won or lost. Start with the financing: just under half of the total is government money; the rest must come from private capital that has to actually be committed, not merely hoped for. Industrial-policy history is littered with grand multi-decade targets where the public tranche arrived and the private match never fully did. A 14-year horizon also outlives multiple electoral cycles and budget negotiations — continuity of intent is itself a risk.

Then there are the structural headwinds unique to Japan. A shrinking, aging workforce makes staffing fabs, construction, and the surrounding ecosystem genuinely hard; advanced manufacturing competes for the same scarce engineers as every other priority sector named in the plan. Energy is the other constraint — leading-edge fabs and AI compute are voracious power consumers, and Japan's electricity mix and grid will have to be part of the answer for the chip target to be physical rather than notional. None of these are reasons the plan fails; they are the variables that determine whether ¥40 trillion in 2040 chip sales is a forecast or a slogan.

And there is the gap our readers know well from enterprise AI: the distance between adopting and deploying. Across global firms, AI pilots are easy to announce and stubbornly hard to put into production — the same pattern that has quietly stalled many corporate rollouts. Industrial policy has its own version of this gap. Money allocated is not capacity built; capacity built is not competitive output; competitive output is not durable market share. The plan's credibility over the next 14 years will rest not on the size of the number but on a chain of unglamorous milestones — yield curves, build-out schedules, offtake agreements, and annual budget follow-through — each of which is checkable, and none of which a press release can fake.

For operators who build, supply, or sell into Japan, the most useful way to read this plan is not as a political announcement but as a 14-year demand curve with a government underwrite behind it. A credible, long-dated commitment of this size lowers the risk of long-horizon bets on Japan. It tells a foreign equipment maker, a materials supplier, a software firm, or a localization partner that the buyer at the other end of the table has both intent and a budget that outlasts the next quarter. That is exactly the kind of visibility that makes multi-year investment decisions defensible to a board.

The concrete opportunity sits in the supply chain rather than the headline fab. Fivefold growth in domestic chip output pulls demand through thermal management, factory automation and robotics, specialty materials and chemicals, cleanroom and facilities engineering, test and metrology, and the unglamorous software that runs all of it. Foreign firms with genuine capability in those layers have a window to plug into a subsidized buildout — provided they localize properly. Japan rewards suppliers who show up in Japanese, navigate its certification and procurement norms, and commit to the relationship depth its industrial buyers expect. Localization is not a translation line item here; it is the entry ticket.

This is the work Medusa Japan exists to do — the bridge between global capability and Japanese execution. Our advice to clients is unchanged by the size of the number: do not let a $2.3 trillion headline become either hype or noise. Track the checkable milestones — Rapidus yield progress, fab build-out schedules, annual METI budget allocations, and the named priority sectors — and map your own offering onto where the spending actually lands first. The decision-makers who do well over the next decade will be the ones positioned to supply this buildout, in the buyer's language and on the buyer's terms, rather than the ones who watched it from the sidelines and called it too big to be real.