Ray Dalio · How Countries Go Broke · gauge readings
Big Debt Cycle Monitor
Dalio says the late stage of a debt cycle is measurable, and that
almost nobody measures it. These are his three gauges plus his market-action
markers, read against live data — including where the tape disagrees with the
narrative.
Data as of 2026-08-28Fiscal: US TreasuryRates: US Treasury & 財務省Korea: 한국은행 · 한국부동산원
3 of the three gauges are elevated. Composite
stage: severe.
Gauge 1 · Debt serviceSevereInterest is 20.1% of federal revenue.
Gauge 2 · Supply vs demandElevated30-year at 5.22%, curve +88bp.
Gauge 3 · MonetizationElevatedFed balance sheet $6.73T, +1.8% over 6m.
GAUGE 1
Debt service relative to revenue
Dalio's plaque metaphor: the slowest-moving gauge, and the one that sets the
ceiling on everything else.
Interest ÷ receipts20.1%~$1.08T on ~$5.38T revenue, annualized
Debt ÷ receipts6.00×Held by the public; 7.45× counting intragovernmental
Outlays ÷ receipts140%$2.16T deficit
Avg rate on the debt3.45%1.28pp below the market 10-year
The repricing pipeline
What Treasury pays on its stock of debt, against what the market charges today
Average rate paidMarket rate todayCommitted repricing gap
3.45%avg paid
4.73%10-year
5.22%30-year
2.9%5.8%
Roughly $10T of principal rolls every year. Each maturity that rolls
moves the average rate paid toward the market rate — the gap is arithmetic
already committed, not a forecast.
The average rate paid has risen in 6 of the last 7 months, toward a market 10-year of 4.73%. The 1.28pp gap is committed future interest expense, not a forecast.
GAUGE 2
Selling relative to demand
Dalio's market signature for this stage is rates rising led by the long
end. That is a testable claim, so the tracker tests it — on twelve months
and on 2 years, because the two horizons currently disagree.
US Treasury curve, three dates
2024-08-28 · 2025-08-28 · 2026-08-28, percent
2026-08-282025-08-282024-08-28
Read left to right, the curve un-inverted. Two years ago the
2-year paid 3.83% against the 30-year's 4.13% — barely any
reward for lending thirty years instead of two. Today the 30-year pays
5.22% and the 2-year 4.34%. Over 2 years the long end
rose +109bp against the 2-year's +51bp;
over the last twelve, +34bp against
+72bp.
30-year5.22%+34bp over 12m · +109bp over 2y
2-year4.34%+72bp over 12m · +51bp over 2y
30y − 2y+88bp+30bp 2y ago · +126bp a year ago
Long end led, 2yYesBy +58bp · -38bp over 12m
Dalio's marker for this gauge is rates rising led by the long end, and the answer depends on where you start the clock. Across 2 years the 30-year rose +109bp while the 2-year rose 51bp — the long end led by 58bp and the curve went +30bp -> +88bp. The marker IS confirmed on that horizon. Over the last 12 months it is not: the 30-year rose +34bp against the 2-year's +72bp and the spread narrowed 126bp -> 88bp. Read across the full window that is the debt-cycle claim; read across the last year it is a hiking cycle ending. Both are on the chart above, so neither can be quoted alone.
GAUGE 3
Central-bank monetization
The gauge that decides which exit the cycle takes — and the only one of the
three whose reading is a slope rather than a level. Scored over
6 months, because the weekly H.4.1 print swings on repo and
Treasury-account operations and would flip this gauge in both directions while
the trend did one thing.
Elevated
Fed balance sheet$6.73T-24.9% versus the 2022-04 peak
Change over 6 months+1.8%-14.8bn on the week — noise at this scale
Since the trough+2.7%Bottomed $6.55T in 2025-11, 9 months ago
Change over 12 months+0.11TFrom $6.62T
Federal Reserve total assets
Month-end, 2021-09 to 2026-08
The shape is the reading, and the shape changed. Quantitative
tightening took $2.41T off the balance
sheet between 2022-04 and 2025-11; since then the
line has turned and risen +2.7%. Nine months of slope,
not one week of print.
Quantitative tightening has ended. The balance sheet bottomed at $6.55T in 2025-11, 9 months ago, and is +2.7% since — +1.8% over the last 6 months. Read week to week this looks like noise; read as a slope it is a turn, which is why the gauge is scored on months.
What it is not, yet: the balance sheet is still -24.9% against its 2022-04 peak, and growth of 1.8% over 6 months is far below the pace of net issuance it would have to absorb to be monetizing the deficit. This is the gauge leaving contained, not arriving at the end state — it escalates when expansion passes 5% over the same window.
MARKER
How long is the fuse
Dalio's third market-action tell is that a Treasury which cannot sell
duration starts funding itself short instead. It is the tell that usually comes
first, because it is a decision the issuer makes rather than a price the market
sets — and it is the one most easily hidden, because the obvious measure can be
held flat while the thing it measures changes underneath.
Elevated
Average maturity70.0mo-0.8mo over 5 years · -1.9mo over 12m
Maturing within 1 year33.3%~$10.5T · from 30.2% 5 years ago
Bills ÷ marketable22.2%+3.8pp over 5 years · TBAC band 15–20%
New coupon issuance84.1mo-9.6mo over 5 years · bills excluded
Maturity structure of the marketable debt
Monthly, 2021-08 to 2026-07 · MSPD and auction results
Average maturity, months (left)New coupon issuance, months (left)Maturing within 1 year, % (right)Bills ÷ marketable, % (right)
The blue line is the number Treasury is judged on and it has
barely moved: 70.0 months against
70.8 five years ago. The red and
gold lines are what happened underneath it. Bills went from
18.4% of marketable debt
to 22.2%, and the share repricing inside twelve months
from 30.2% to
33.3%. Pairing more bills with longer bonds holds an
average still. It does not hold a rollover still.
On the obvious measure the marker is NOT confirmed. Average maturity of marketable debt is 70.0 months against 70.8 months 5 years ago — a move of -0.8 months. Treasury has not, on this number, shortened anything.
The composition underneath it did change. Bills are 22.2% of marketable debt against 18.4% 5 years ago, above the 15-20% band the Treasury Borrowing Advisory Committee has long treated as normal, and 33.3% of the stock now matures within twelve months, from 30.2%. That is the barbell: more bills at the front paired with longer coupons behind them holds the average still while raising the share that reprices inside a year.
Which is why this sits next to Gauge 1 rather than on its own. A third of the marketable debt repricing annually is the mechanism by which the gap between the 3.45% Treasury pays and the 4.73% the market charges becomes interest expense — not over the 5.8 years the average maturity implies, but far sooner for the front third of it.
MARKER
Who is financing it
Gauges 1 and 2 measure the burden and the price. This is the other half of
the same question: who shows up at the auction. Central banks and
sovereign funds buy Treasuries for reserve reasons rather than return ones —
they are the bid that does not negotiate. Private money does negotiate, and
the price it negotiates for is the yield.
Severe
Foreign official$3.78T-114bn over 12 months
Foreign private$5.52T+2.22T over 5 years
Official share40.6%From 42.8% a year ago
Total foreign held$9.30TAs of 2026-06 · TIC lags ~2 months
Foreign holdings of US Treasuries, by holder type
Monthly, 2021-06 to 2026-06
Foreign officialForeign privateTotal
The two lines crossed. Five years ago official holders were the
larger half of foreign ownership; today private money holds
$5.52T against official money's $3.78T.
Total foreign demand grew throughout — this is a change in who, not
in how much.
Change in holdings by jurisdiction
2021-06 to 2026-06 · the twenty largest holders, 80% of the total
Reserve managerCustody / fund domicileMixed
United Kingdom
+406bn
Canada
+292bn
Belgium
+254bn
Cayman Islands
+193bn
France
+163bn
Luxembourg
+132bn
Singapore
+96bn
Norway
+87bn
Taiwan
+63bn
UAE
+54bn
Hong Kong
+37bn
Ireland
+31bn
Germany
+25bn
Saudi Arabia
+15bn
Korea
+4bn
Switzerland
-17bn
India
-34bn
Brazil
-81bn
Japan
-163bn
China
-428bn
−$504bn0+$504bn
Almost everything that fell is a sovereign reserve holder;
almost everything that rose fastest is a custody or fund-domicile centre.
That is the same rotation the chart above measures directly — but read it
with care, because TIC attributes a bond to its custodian's country, not its
owner, so a holding that moves from Tokyo to a London custodian appears here
as Japan selling and the UK buying.
Foreign holdings of Treasuries rose $1.78T over 5 years, so demand from abroad is not the problem. Its composition is. Official holders — central banks and sovereign funds, the bid that does not negotiate on price — went down $438bn over the same window while private holdings rose $2.22T. The official share of foreign holdings is 40.6%, from 42.8% a year ago.
The country table splits the same way. Reserve managers hold $350bn less than five years ago; custody and fund-domicile centres hold $1.02T more. Read that cautiously: TIC attributes a holding to the custodian's country, so part of the shift is the same bonds moving to a different custodian rather than to a different owner. It is a reason to trust the official/private split above, which is measured directly, over any single country's line.
MARKER
The measuring stick
A currency's decline is invisible measured against other currencies that are
declining too. Gold rose 32.0% in dollars over twelve months.
What each currency lost against gold
Twelve months to 2026-08-28
JPY
-30.3%
EUR
-24.6%
USD
-24.2%
KRW
-23.5%
CNY
-19.4%
Measured against each other these currencies barely moved.
Measured against gold they devalued together.
The same assets in two denominators
Asset
Level
12m local
12m in gold
S&P 500
7,711.76
+18.6%
-10.1%
KOSPI
6,788.88
+112.4%
+60.9%
MARKER
The Korean buffer
A US homeowner on a thirty-year fixed mortgage is short the bond: inflation
transfers wealth from the lender to them, which is why American housing absorbs
a rate shock through volume rather than price. Whether a Korean household has
any version of that buffer is a measurable question, and the answer has changed
fast — while the stock of debt the buffer was protecting kept growing.
Fixed share of new mortgages (left)가계신용, 조원 (right)
Two series, two axes, one question. Fixed-rate lending has fallen
from 96.4% of new mortgages in 2024-07
to 31.9%, most of it in the last eight months as the Bank of
Korea began tightening. Over the same window 가계신용 rose
+7.7% to 2,020조원. The
share being repriced is rising against a balance that is also rising. Note that
Korean 고정형 is typically 혼합형 — fixed five years, then floating — so the
buffer was always far shorter-dated than the US thirty-year.
Fixed share, new loans31.9%From 88.8% a year ago
Fixed premium+0.41pp고정 4.76% vs 변동 4.35%
BOK base rate2.75%As of 2026-07
가계신용2,020조+7.7% over 3 years
The premium flipped sign. Through 2025 fixed-rate mortgages
were cheaper than floating and roughly nine in ten new borrowers took
them; fixed now costs +0.41pp more and fewer than one in three
do. Lenders are pricing the rate path before borrowers are — which is what the
withdrawal of a buffer looks like while it is happening.
Seoul apartments: price, 전세, 월세
한국부동산원 monthly indices, rebased to 100 at 2023-06 · through 2026-06
실거래가격지수매매가격지수 (조사)전세가격지수월세통합가격지수
Ranked, and the ranking is the finding: what buyers actually
transacted at rose 29.1% over 3 years, the survey index
21.9%, 전세 18.8% and 월세
12.8%. Prices outran the rent on the same apartments by
3.2pp against 전세 and
9.1pp against 월세. A price
rising faster than the rent it can earn is not being paid for out of income —
it is being paid for out of credit, which is the same buffer question in a
different denominator.
실거래가격지수+29.1%Over 3 years, 서울 아파트
매매가격지수+21.9%Survey lags transactions by 7.2pp
전세가격지수+18.8%3.2pp behind 매매
월세통합가격지수+12.8%6.0pp behind 전세
MARKER
Japan, the same measurement
Japan carries roughly 215% debt-to-GDP on the assumption that yields stay near
zero. That assumption is being withdrawn at the long end first, which is exactly
the shape Gauge 2 looks for — so it is worth reading on the same axes as the US
curve rather than as a single 10-year print.
JGB curve, three dates
2024-08-27 · 2025-08-27 · 2026-08-27, percent
2026-08-272025-08-272024-08-27
Two years ago the 2-year JGB paid 0.37% and the
30-year 2.08%. Today they pay 1.70% and
4.04%. The 30-year has risen
+196bp over 2 years and
+90bp over twelve months — a bond market that spent a
generation at zero repricing duration in public. Source: 財務省 daily par
yields, the issuer's own numbers.
30-year JGB4.04%+90bp over 12 months
10-year JGB2.90%+126bp over 12 months
30y − 2y+234bpFrom +171bp 2 years ago
As of2026-08-27Daily, 財務省
Dalio's Q7 asks what happens when the country that proved
large debts can be carried cheaply stops being able to. The US 30-year is
5.22% and the JGB 30-year is 4.04%. The gap between
them has closed to 118bp from
205bp 2 years ago — which is the
hedged-yield arithmetic that has historically sent Japanese capital abroad,
running in reverse.