Water Does Not Ask About Fruit

The US total shareholder payout yield (dividends + share buybacks ÷ market cap) fell from its 2018 level of 5.78% to a 2025 level of 3.06%, nearly halving.

The catch-up is clearest on the dividend axis, where all three countries use the same measure. In 2018, US and Japanese dividend yields started together (1.98% vs 1.97%). They crossed in 2021. Now Japan (1.94%) is above the US (1.28%). Korea (1.40%) is above the US too. Japan caught up to the second decimal place on the payout-ratio axis (#07). The throne lowered itself more than Japan leaped.

A yield, however, is a fraction. Did it fall because the numerator (money returned) stalled, or because the denominator (market cap) surged? One descending line cannot tell us. #07 did not open the fraction. This report does.

Open the fraction. The absolute amount returned by US companies rose from $1,460bn in 2018 to a 2025 total of $2,110bn, up +45%(+5.4% annually). Over the same period, market cap rose from $25,200bn to $68,900bn, +173%(+15.4% annually). The payout yield halved because share prices rose three times as fast, rather than because companies stopped returning money.

So the falling line in #07 was partly an optical illusion. That was yesterday's story.

This year, the numerator stalled. In 2025 capex rose +20% and share buybacks +9%. In Q1 2026: capex +39% vs share buybacks +1%. Goldman Sachs forecasts capex +33% against buybacks +3% for full-year 2026. In 1 year, the scissors opened wide.

Individual companies make it clearer. Alphabet cut share buybacks to $0 in Q2 2026. 1 year earlier, the same quarter saw $13.2bn. It was the first halt in 10 years. In the same quarter, it went further: $49.6bn of new equity issuance and $20.3bn of senior debt raised roughly $70bn for AI infrastructure. Long-term debt doubled from $46.5bn to $98bn in a single quarter.

That brings us to this report's question. What happens next to US shareholder payout ratios? The ratio is ultimately a function of what remains after capex is subtracted from free cash flow. So the question becomes: can capex fall?

Three locks make that difficult.

Another pressure is approaching: the depreciation wall. Depreciation on $434bn of capex over the preceding 4 quarters will fully reach income statements from the second half of 2026 through 2027. Free cash flow tightens again. If capex cannot fall and depreciation rises, only one adjustable variable remains: shareholder payouts.

One turning point may already be behind us. Our pipeline's market-cap-weighted net issuance rate(change in share count × price ÷ market cap) showed net cancellation, with buybacks outweighing issuance and stock-compensation dilution, in 2022 at −5.6% and in 2024 at −2.0%. But the sign turned positive in 2025, at +0.03%. Buybacks no longer beat dilution. The measure fluctuates by year (also +0.1 in 2023), so the 2026 value will decide the verdict. Its direction matches Alphabet's case.

Did less water reach the tree announcing that its fruit was drying up? The reverse. Over the latest 12 months (2025-06–2026-05), foreigners' net buying of US stocks was $901.8bn — versus US Treasury net buying of $282.1bn over the same period, 3.2x as large. America's main channel for raising money abroad is now stocks rather than Treasuries.

In holdings terms, foreigners own US stocks worth $19,860bn(end-2025-06, +17.7% yoy), and their share of the US stock market is a record 18%. Apollo's Slok described the picture in one phrase: “the mirror image of the trade deficit.” (translated) Dollars received by countries selling goods to the US are recycled into US assets.

Korea sits inside this structure. KSD settlement measurements (our pipeline #55) show Korean retail net buying of foreign-currency stocks briefly turned to net selling in April–May this year (−$0.5bn, −$0.9bn), before reigniting at +$4.6bn in July, when Korean stocks slumped and the won strengthened. It was the third-largest month of the latest 14. Meanwhile, custody holdings fell from $204.2bn (KRW 301 trillion) at end-May to $177.3bn (KRW 256 trillion) in mid-July (media figures). Holdings shrank even while investors bought — losses exceeded the purchases.

Japan lies on the same waterway. Commentators repeatedly say much of the household saving mobilized by NISA flows into US funds, but we have no official allocation measurement. The statement remains qualitative.

Funding with no interest and no maturity. If things go well, pay with a growth story instead of dividends. If prices collapse, external liabilities evaporate from the books. That is how equity funding works. Is the combination of less fruit and record water merely a coincidence?

Official statistics for this year's Q1 captured the mechanism in action. The US ran a current-account deficit of $226.8bn that quarter. Yet its net external liabilities (NIIP) improved by $600bn. How? The value of US assets held by foreigners fell, erasing $1,180bn from the liability side of the books. A quarter in which the deficit is real while liabilities disappear from foreign investors' accounts: this is the valuation effect.

There is a precedent: 1998–2001. Foreign buying near the dot-com peak and surging capital-gains taxes helped the US run fiscal surpluses for 4 consecutive years. The debt ratio fell from 49% to 33%. Foreign capital that entered near the peak — mainly European — absorbed losses in the 2000–02 crash. This machine has worked once before.

If the hypothesis is right, Korean households' $177.3bn forms part of that absorption layer. Chart 6's “bought more, held less” happens in exactly that position.

Inflows of $901.8bn cannot repay debt growth of $2,880bn. Redefine a solution as stabilizing the debt ratio rather than repayment, and a threshold appears: the ratio stops rising when nominal growth exceeds 5.9%. It is now 4.7%. Only one candidate can plausibly close the gap of 1.2 percentage points: the growth contribution of AI investment financed by global capital (estimated +0.5–1.0 percentage points). Add capital-gains taxes at the bull-market peak (~$300bn, 15% of the deficit).

The conclusion is conditional. If AI investment produces productivity gains, this structure really eases the debt problem. Otherwise, only exported losses remain — protecting external accounts, but not public finances. The verdict moves to the next section's two pillars.

The machine needs both pillars to stand. The first is water — liquidity must keep absorbing capex funding. The second is proof of fruit — earnings must show that AI makes money. The problem is that the pillars run on different timetables.

This report also measures the base of the water pillar: cash pools absorbing the Treasury's tilt toward bills. MMF $8,400.6bn(+$149.8bn over three months), dollar stablecoins of $307.2bn(Tether alone is among the world's top 20 US Treasury holders, though supply has stalled at −$2.3bn over the latest 30 days); 4-week bill bid-to-cover is 2.68(20-auction average 2.86, not below the low). With the Fed's reverse-repo buffer exhausted, only growth in these two pools can absorb October's TGA rebuilding instead of reserves. No stress signal yet. The baseline is set.

The two stresses overlap in Q4 2026–2027. If a crisis arrives, that window is likely, with a starting point in credit rather than share prices. Off-balance-sheet obligations of $1,650bn are invisible, and surprises emerge where we cannot see. The first crack will appear in the absorption of new bonds — spreads and subscription ratios.

Combine this with the §05 structure and the conclusion twists. Even in a crisis, the allocation of losses is already determined: foreign equity holders ($19,860bn) absorb stock losses, bond and SPV investors absorb credit losses, and data centers remain on US soil. The baseline is therefore a manageable crisis. Only one path makes it unmanageable: outside capital bearing those losses begins leaving dollar assets altogether.

When will the water that flowed without asking about fruit begin to ask? That is the next page of this map.