Japan's $2.3 trillion gamble, China's $3 AI, and the new map of capital flows

Tokyo framed it as a national mission. On 3 July 2026, the minister in charge of Japan's growth strategy said a 370 trillion yen ($2.3 trillion) investment plan — the country's largest ever — was designed to reignite so-called "animal spirits" in an economy that has spent three decades drifting between stagnation and half-measures. The figure, carried by Nikkei Asia, lands in a global week dominated by two other data points from the Asian corridor: Chinese AI inference tokens now selling for $2–$3 per million output, an 80 percent discount to comparable US models, and a record $60bn of foreign capital chasing Japanese equities in the first half of 2026.

The thesis this publication draws from those three threads is straightforward: the post-2020 narrative of US financial and technological primacy is being repriced, on a monthly basis, by capital and compute flows across the Pacific. None of the three data points is decisive on its own. Read together, they describe a coherent rerouting.

The size of the Japanese bet

The 370 trillion yen ($2.3 trillion) figure covers a five-year horizon and bundles private-sector co-investment with public capital, Nikkei Asia reported. The framing matters: Tokyo is not announcing a fiscal stimulus of that magnitude from the central balance sheet. It is signalling where it wants coordinated state, bank and corporate capital to land — semiconductors, decarbonisation, AI infrastructure, biotech, defence-adjacent industries, and the supply chain hardening that has become a regional obsession since 2022.

That the language of "animal spirits" is being deployed is itself telling. The phrase belongs to the Keynesian lexicon of the 1930s, and its rehabilitation in Tokyo in 2026 is a quiet admission that monetary policy has done what it can. With the Bank of Japan normalising rates and yen volatility still a constraint on corporate planning, the political class has decided that the demand-side answer is investment, not liquidity. Whether the private sector actually deploys the matching capital will be the test of the next twelve months.

The other half of the Asia trade

Foreign investors bought a net 9.7 trillion yen ($60bn) more Japanese stock than they sold in the first six months of 2026 — the largest first-half inflow on record, according to Nikkei Asia. The composition of that buying matters: it is not a retail-trade phenomenon driven by household savings. It is institutional capital pricing in two things — corporate-governance reform under the Tokyo Stock Exchange's continued pressure on price-to-book ratios, and a yen exchange rate that, even after the recent recovery, still leaves Japanese assets cheap by global standards.

The flows are not zero-sum against China. They are an expression of a portfolio that wants Asian exposure and is choosing Japan for the first time in a decade as its anchor — partly because Chinese onshore equities remain accessible but geopolitically awkward, and partly because Japan's reform story finally has measurable evidence behind it. The capital is not abandoning Greater China. It is diversifying out of it.

The price war underneath the AI race

The most consequential of the three data points is also the quietest. According to reporting referenced by Unusual Whales, certain Chinese AI models now price output at $2 to $3 per million tokens, against roughly $15 for comparable US models — a 5x gap. UBS has flagged this dynamic as a structural threat to US hyperscaler economics. The point is not that Chinese models are necessarily better on benchmarks. The point is that at inference, on the price-sensitive workloads that dominate enterprise adoption — customer service automation, document summarisation, code generation, internal search — the marginal buyer is rationalising compute spend the way they rationalised cloud spend in 2014. Whoever is cheapest at the unit level captures the long tail.

This is the same playbook that delivered Chinese automakers a record month in Europe. Nikkei Asia reported on 2 July that Chinese carmakers surpassed Japanese brands in Europe's passenger car market for the first time in May, despite EU tariffs on Chinese electric vehicles. The mechanism is identical to the AI story: scale, vertical integration, and willingness to compete on margin where Western peers will not. Beijing's industrial policy did not invent this capability, but it financed the capital expenditure that made it cheap.

What the consensus misses

The Western framing of these three stories tends to treat each in isolation. Japan's package is read as stimulus; the equity flows as a yen trade; the AI price gap as a subsidy artefact. Each of those readings has a kernel of truth, but each underweights the same underlying fact: a global capital and compute order that, five years ago, priced American assets at a structural premium is now pricing them at a premium that has to be earned, quarter by quarter, on output and unit economics.

The counter-narrative — that Chinese pricing is subsidy-distorted and that flows will reverse if US rate policy shifts — has weight. The AI discount is enabled by domestic Chinese chip capacity that benefits from state-supported capex, and Japanese equity flows are partially a function of the yen's safe-haven correlation during specific risk-off episodes. Neither of these counter-narratives refutes the underlying story. They bound it. The question is not whether the rerouting reverses; the question is how far it gets before it does.

Stakes and the next six months

The winners if the trajectory continues are Japanese reform-oriented corporates, Chinese AI labs serving the cost-sensitive inference market, and Chinese OEMs with European distribution. The losers are US hyperscalers whose pricing power has, until now, been treated as gravity, and Japanese automakers still treating their domestic market as a protected annuity. The time horizon that matters is six to twelve months — long enough for the H1 2026 flow data to be ratified by Q3 corporate guidance, and short enough that the AI price gap is still contestable rather than structural.

What remains uncertain is whether the Japanese private sector actually matches the public signal with deployment. The 370 trillion yen plan is a coordination device, not a cheque. Until buyback announcements and capex guidance begin to corroborate the political signal, the inflow can mean a long-term reweighting or a temporary hunt for yield. The AI price gap, meanwhile, depends on access to advanced fabrication, which is itself a function of export-control politics not visible in the unit economics. Both stories are real. Both are also more conditional than their current framing admits.

This piece is framed by Monexus as a single arc across three data points — Tokyo's investment plan, the H1 equity flows, and the AI pricing gap — rather than as three separate Asia stories. The wire services reported each in isolation.

Wire provenance

This editorial synthesis draws on the following public wire/social posts:

• https://t.me/s/NikkeiAsia
• https://t.me/s/NikkeiAsia
• https://t.me/s/NikkeiAsia
• https://unusualwhales.com/news/fda-approves-philip-morris-zyn-reduced-risk