The Price of Water

The price of water—money—is the interest rate. The US 10-year yield is now 4.68%; the 30-year was 5.25%, a step above 1 year ago; the 30-year has settled in the 5% range. That much is familiar. This report starts with reading one yield as two components.

The nominal yield is the sum of two parts: Real yields (money's true rental cost after inflation) and inflation expectations (the market's inflation outlook, inferred from inflation-linked bond prices and often called BEI). Which component rises changes the story. Inflation concerns lift BEI; more expensive money itself lifts real yields.

Now split it. Over the past 1 year, the 10-year nominal yield rose from 4.43% to 4.68% by+25bp (bp = 0.01pp). Inside that change: real yields rose from 2.04% to 2.41%, +37bp, while breakeven inflation (BEI) fell from 2.39% to 2.27%, −12bp. Oil rose with war and refining margins reached extraordinary levels, yet BEI fell. The market's answer is clear: money itself, rather than inflation, is getting more expensive.

This matters because the real yield is the discount rate converting distant cash flows into today's value. Higher real yields hit assets with distant earnings, such as AI and growth stocks, first. This explains the second week of August, when memory prices accelerated (export prices +168% year on year) while semiconductor shares alone broke down. An issue of discount rates, not earnings.

And this discount rate has a threshold. Expected stock returns start with the earnings yield (company earnings ÷ share price, the inverse of P/E). The S&P 500's earnings yield is now about 3.3% (P/E 30x, trailing realized earnings over 1 year, observed in our gauges). In a world where 10-year Treasuries pay 4.7%, the return for taking risk is below the return without it. The gap is called the equity risk premium (ERP). It first turned negative not now, but in late 2024, when reports called it the first time since the dot-com bubble. What is new is the depth: a trough near −0.4pp in March deepened in August to−1.35%p (observed using realized earnings; analysts' next-year earnings estimates give roughly 0, so the ruler matters). Only one bridge justifies negative ERP: the promise of rapidly growing earnings, the growth story, currently AI. This creates a dangerous combination. If the price of money rises further or the story falters, the shift from stocks to bonds comes as a fault-line break rather than a gradual move. The precedents are 2000 (−49% from the peak) and Q4 2018 (−20%). If #10's proof-of-fruit pillar falters while money is expensive, that is the worst arrangement.

There is another way to watch this threshold: observe the money moving directly. Short Treasuries already offer nearly 4% without risk (1-year 3.98%, 6-month 3.95%, observed). Funds parking cash in short Treasuries and deposit-like assets, money market funds (MMFs), hold a record USD 7.93 trillion (8/12 week, ICI), growing by USD 18.3 billion in that week alone. Which queue does money join: stocks' promise of growth or bonds' fixed return in the 4% range? If ERP is the threshold, the growth rate of MMF assets is the sound of footsteps moving toward it.

The question narrows: why are real yields rising?

So far: The main rise in the price of money came from real yields, not expected inflation. This section asks why.

Money's price follows supply and demand. What is happening now is not shrinking supply, but surging demand. More hands suddenly want long-term borrowing. Two hands queue at the same well.

The first is the US government. Already the world's largest borrower, it faces cooling Treasury-auction demand. In March, weak 2-, 5- and 7-year auctions came in succession. Dealers—the banks supporting auctions—took 24% of the supply, twice normal. The Treasury froze rather than expanded issuance: no increase in medium/long coupon issuance through year-end. Instead, it rolls the difference into bills (discount debt maturing in less than 1 year). Dealers calculate that continuing this creates a USD 1.45 trillion financing gap in 2027–28. This postpones, rather than resolves, the problem.

The second hand is new: AI. Data-center construction is now funded through bond markets, not just company cash. Global AI-related debt issuance this year is forecast at USD 570 billion, with USD 236 billion already issued by end-May, four times last year's pace. Demand is already cooling: in February, hyperscaler (large cloud-company) bonds drew orders at 5x issuance; by July, below 2x. Outside public markets, another~USD 800 billion is expected from private credit. It does not appear in filings; we count only what is visible.

Government and AI compete for the same pool of long-term funding. As US stocks absorbed global equity capital (#10's absorption mechanism), US debt now absorbs global bond capital. The price of water rises because more mouths drink, not because less water exists.

So far: Competition for long-term funding pushes real yields higher. The buffer that absorbed that pressure is now gone.

The financial system has a reservoir for surplus cash: reverse repo (RRP), the facility holding spare market cash overnight at the Fed. While full, it let markets absorb large Treasury issuance without strain. 2 years ago, it held USD 2 trillion. Now its balance is USD 450 million, effectively empty.

Once that reservoir empties, pressure reaches bank reserves: banks' settlement cash at the Fed, the financial system's blood pressure. Over the past 4 weeks, reserves fell USD 152.8 billion, crossing our 4-week −USD 150 billion boundary for the first time. Treasury funding is now draining not surplus cash, but the system's blood directly.

This liquidity pressure coexisted with a pause in long-term yields in the same week (August's second week: 30-year 5.24→5.21%). Tension has not disappeared. The next auction will distinguish the possibilities; gauge ③ watches that.

So far: Demand has surged and buffers are exhausted. Someone now has to name a price. The twist is: both sides avoid doing so.

The Treasury avoids it through issuance structure. Asking markets for a long-term rate risks an expensive answer, so it freezes coupon issuance and rolls funding through bills, avoiding the curve's long end. The Fed's avoidance is political. With USD 38 trillion of debt, a rate rise directly hits government interest costs (1pp ≈ USD 380 billion a year); in a midterm-election year, the decision becomes a battle. The two institutions each hold one end of the curve, while the middle—long-term rates—is vacant. The 10-year–30-year spread weighs that vacancy. A wider gap means a higher premium for the distant future: term premium.

Four possible futures, in one table:

We suspect S4 is the baseline path. The components observed throughout this series fit S4: weak-dollar coordination (#11), bill-heavy issuance (this edition), institutionalized interest-free dollar tokens (stablecoins, interest-free borrowing from the government's perspective), and tariffs (accepting higher import prices). Holding rates while prices rise lowers real rates and quietly erodes debt. The US halved its war debt that way in the 1940s. Inflation is a tax paid by bondholders and depositors.

This path has a deadline. If nobody sets the price, markets eventually will. Bond investors who lose faith in fiscal policy push long yields higher: the bond vigilantes. Capital could not cross borders in the 1940s; now it can. The vigilantes arrive faster. This report's spine: when nobody offers to set the price, bond vigilantes do.

So far: Nobody officially names the price of water. Markets do not wait. Private price tags are already appearing in three places.

First, gold. Textbooks say gold moves opposite to real yields: an asset paying no interest should lose appeal as money's rental cost rises. The observed 1-year correlation, −0.61, agreed. Over the past month, however, real yields rose and gold rose with them. Gold exceeded the regression model by +9.6pp, a top-4% deviation over 1 year. The buyers explain it: Western ETFs, driven by interest-rate logic, sold USD 8.9 billion in June alone, while central banks bought 289 tonnes in Q2 alone throughout the price decline. Gold moved from rate-sensitive sellers to buyers avoiding sanctions and the inflation tax. Gold's owners are changing, and those hands do not fear real yields.

Second, the dollar's new wrapper. Stablecoins—tokens pegged 1:1 to the dollar—have shrunk from their May peak by more than USD 15 billion, the largest drop since Terra collapsed. The cause is rules, not panic: GENIUS prohibited interest payments, pushing yield-seeking money into tokenized Treasuries, interest-paying Treasury assets sold as tokens. What looks like a crypto retreat is actually dollar savings shifting from interest-free to interest-bearing wrappers. Either wrapper holds short-term US Treasuries. From the Treasury's perspective, either wins.

Third, the periphery. In August's second week, US long yields paused for three days (5.24→5.21%). Over those three days, foreigners bought Korean stocks worth KRW +8 trillion, the largest inflow in 5 weeks, precisely as money's price paused. Reversing the reading, renewed rate rises send money out just as quickly. Peripheral asset prices are now a function of US real yields. This is the footing beneath the KRW 8 trillion in #12 §08.

So far: So far, US money prices. But the world's largest water exporter is Japan, home to #12's waterwheel: overseas earnings paid to domestic shareholders. Prices are rising at that source too.

Japan's 30-year yield is 4.002%. The 4% notch on #12's gauge ⑤ triggered two days before publication. The waterwheel's condition fits one inequality: cost of borrowing <earnings yield. More precisely, after-tax borrowing costs must be below the earnings yield (the inverse of P/E) for debt-funded buybacks and returns to work. Kinden borrowed JPY 230 billion from banks to buy its shares; US companies issued bonds for buybacks over 15 years. Both rested on this inequality. And in the US, the inequality has already reversed: funding in the 5% range versus earnings yields of 3–4%. The decline in total US shareholder returns in #08 was arithmetic, not morality. Japan's inequality is still open (funding 1–2%, earnings yield 6–12%). For now.

A larger story follows. In S4's world of holding rates despite inflation, the biggest bill goes to depositors. Nominal interest stays unchanged while prices rise, quietly cutting purchasing power. Of Japanese households' JPY 2,385 trillion in financial assets, 47.2% is cash and deposits. When that money chooses movement over erosion, with domestic yields now supplying a reason, the largest shift on the global capital map begins. Sent to the US, it fuels the absorption mechanism. Brought home through repatriation, it becomes the largest selling pressure on US long bonds. Which direction is Q4's biggest question.

So far: Private price tags have appeared; official pricing is postponed. The dates ending that delay are on the calendar.

September — CLARITY vote. The US crypto market-structure bill faces Senate cloture (ending debate, requiring 60 votes). Its stablecoin-rewards clause switches §05's wrapper flows: permission means stablecoin re-expansion; a continued ban entrenches the move to tokenized Treasuries. October — Refunding. The Treasury's quarterly refunding announcement (QRA): keep coupons frozen, or increase issuance facing the USD 1.45 trillion gap? Keeping the freeze formalizes S4. Meanwhile, monthly 10-year and 30-year auctions provide the earliest vigilante signals: tails (yields above expectations) and indirect allotments (foreign demand).

This timetable sits on #10's broader calendar: liquidity stress overlaps with proof of fruit, AI monetization, in the Q4 2026–2027 window. This report adds that we enter that window with buffers exhausted (§03).