Zhihu Series Review: Qianxia Qingfengyue’s Dollar-System Crisis Thesis

Since 2026, Zhihu author Qianxia Qingfengyue has written a series around two questions: “Why hasn’t the US recessed yet?” and “What will the next crisis look like?” Read as one body, the four pieces form a complete argument chain: fiscal capacity is the foundation of all competition → the dollar system is an “imperial tax” → the US tax base sits on asset bubbles → five tricks keep the expansion alive → bubbles come in three tiers → the endgame is a slow crisis. I fact-checked each piece; this is the combined review.

Verdict First

Bottom line: the skeleton is real, the frameworks are worth keeping, but the conclusions keep sliding into narrative. The most valuable piece in the series is the bubble taxonomy in How Scary Is the Next Crisis (valuation/credit/currency); the most aggressive is What Is America Really Anxious About (“one Huang Chao away from collapse”, the capital-controls thesis); the most restrained — and closest to my own view — is the 8/25 closer.

Series score: 6.5/10 — partially correct, real highlights but clear holes.

I. The Series’ Argument Chain

Assembled, the four pieces run:

Fiscal capacity is the foundation of competition: under the surface of industrial/tech/military rivalry sits fiscal capacity.
The dollar system is an “imperial tax”: the US prints → the world produces → dollars flow back → the inflation tax is shifted onto surplus countries; the stronger a surplus country gets, the more it wants out — hence the de-dollarization impulse.
The US fiscal tax base rests on asset bubbles: equities = the anchor of consumption, pensions, corporate financing, and capital-gains tax — a “left foot stepping on right foot to climb to heaven” loop, structurally homologous to China’s old land finance.
Five tricks keep the economy from recessing: 6%+ deficit injection; Yellen’s T-bill duration magic + RRP drain; stablecoin legislation binding issuers to T-bills; Big Tech’s AI capex as “off-balance-sheet QE”; geopolitical turmoil + Japan/Korea bubble harvesting to pull hot money back.
Bubbles come in three tiers: valuation (2000 internet, repairable, leaves infrastructure) → credit (2008 housing, penetrates the credit system, can be absorbed by own-currency credit) → currency (Argentina/Turkey, the true endgame).
The endgame judgment: most likely a 1970s-80s-style stagflation slow crisis (long, high inflation, negative real rates diluting debt); an acute liquidity crisis can’t be ruled out — and if it happens without timely rescue, it’s “the financial market ceases to exist.”

II. Piece-by-Piece Audit

What Tricks Has the US Pulled to Avoid Recession

My verdict: the first four tricks are real; the fifth is narrative.

① Deficit-driven growth: real, the thickest foundation. The deficit sequence (2020 14.9% / 2021 12.4% / 2022 5.5% / 2023 6.3% / 2024 ~6.8%) roughly checks out; net interest formally exceeded defense in FY2025 (FY2024 was close, accounting-dependent), and interest is now the #2 budget line after Social Security, above Medicare. Without this 6%+ purchasing-power injection, none of the other four tricks would exist.
② Yellen’s duration magic + RRP drain: real, the most technically sophisticated. RRP fell from the $2.55T peak (June 2022) to below $400B in 2024; money-market-fund cash went from “parked at the Fed” to “buying T-bills,” simultaneously helping QT and finding buyers for Treasuries. The article’s sharpest point: the pool runs dry — once RRP empties, the marginal buyer becomes bank reserves, directly conflicting with QT, which is exactly the technical reason QT had to slow/stop.
③ Stablecoin legislation: real, but two details are wrong. The GENIUS Act does pin compliant stablecoin reserves to T-bills (1:1, no rehypothecation), but it was signed 2025-07-18 (not June as the article says) and enforced from 2027-01-18; the scale is inflated — the global stablecoin market is ~$250-300B (USDT ~$189B in Q1 2026; USDT+USDC ~93% of the market), and USDT is domiciled in El Salvador and can’t legally serve the US market without Treasury recognition — the real compliant T-bill buyer is essentially USDC alone.
④ AI capex “off-balance-sheet QE”: real but the most expensive. The mechanism holds (Fed shrinks while giants lever up), but the article’s specific numbers (Google’s 20 years of net cash exhausted, Meta FCF collapsing from $54B to $0.78B) couldn’t be independently verified — directionally plausible.
⑤ Geopolitics + Japan/Korea harvesting: the facts are real, the explanation is narrative. The KOSPI crash is confirmed (circuit breakers, won at 17-year lows, foreign outflows), but “Washington engineered the bubble and harvested it” is an unfalsifiable narrative; the real mechanism is leverage unwinding plus risk-off repatriation — no conspiracy needed, just leverage plus panic.

How Scary Is the Next Crisis

My verdict: the best of the three full essays; the framework is the real content.

The bubble taxonomy is useful: valuation bubbles (wealth redistribution among shareholders, don’t penetrate the credit system, leave infrastructure behind); credit bubbles (asset prices become collateral — the 2008 lesson); currency bubbles (printed money that can’t buy things — a mismatch between money and real productive capacity). This classification is the most valuable part of the essay.
The three-anchor bind (stocks/bonds/dollar bound together): true. And it adds a key observation — every past cycle had a “next baton” (internet → housing → QE → fiscal), and now AI is the only baton left; there is no next one.
Two things to wring out: ① “AI as the new anchor of dollar credit” overreaches — AI is a marginal growth narrative, not an anchor; the anchor remains the US institutional/energy/military complex. ② “A perfectly closed logic loop + gold/RMB as Noah’s Ark” is self-serving narrative — self-consistency is not correctness, and it ignores that in the acute phase of a dollar liquidity crisis, gold and EM assets fall together (even gold was sold first in March 2020). The conclusion is actually a bet on “the dollar system damaged after the crisis,” not a crisis-period hedge.
Data corrections: the article says debt passed $40T by 2026-06-30; it actually crossed on 8/18 (June-end was ~$39.4T). The Fed’s $9T balance-sheet peak is right, but it’s now down to $6.76T (-26%) — there is actually room to re-expand.

What Is America Really Anxious About

My verdict: good framework, aggressive conclusion, the piece needing the most caution.

The skeleton is right: the fiscal tax base rests on asset bubbles; the top 10% hold 85-90% of stocks and the top 1% over 50% (matches Fed SCF); the top 10% drive ~50% of consumption; “assets must not fall, debt must expand, deficits must run, liquidity must persist” — policy is handcuffed to the market, consistent with my own view.
Three things to wring out:
- “One Huang Chao away from collapse” — pure drama. Unsustainable ≠ imminent. Japan has run debt above 200% of GDP for 25+ years without collapsing.
- The “inevitability” of de-dollarization is too mechanical — surplus countries are locked in (a prisoner’s dilemma); China is the one that least wants a disorderly dollar collapse (it would destroy its own reserves and export markets). De-dollarization is hedging, not exit.
- The capital-controls thesis is the weakest link in the series: using “in an extreme scenario the US forces you to hold and won’t let you sell” as the basis for allocation is a tail-within-a-tail scenario; and the irony is sharp — the author fears capital controls, yet his own currency, the RMB, is among the world’s most capital-controlled, which he never mentions.
Audit of the two evidence points: ① the private-credit blowup is real (KKR/Cliffwater/Blue Owl/Apollo redemption caps cascaded in early 2026; Morgan Stanley warned of a “Covid-scale shakeout”); ② the UAE swap is misread by the author — the UAE isn’t short dollars, it wants US political backing after the US-Iran war cut off Hormuz (Brad Setser: “a status signal, not financial distress”; OMFIF: “political, not financial”).

The 8/25 Closer: Slow Crisis First, Acute Crisis as Tail

My verdict: the most restrained piece, and the one closest to my own view.

Slow crisis first, acute as tail — exactly my “chronic attrition as the main path, acute as the tail.” Stagflation plus negative real rates diluting debt (financial repression) is the monetization path I’ve described all along.
“Liquidity isn’t the problem — the cost of rescue, inflation, and votes are” — the sharpest line. In the fiat era, deflation can simply be made not to exist; it’s only a question of whether the government is willing to debase.
AI capex sustainability is the core question: $1.3T already spent plus $1T+ expected — who leveres up to generate that demand? With the whole world highly leveraged and yields rising outside China, this monetization gap is the right question.
Four things to wring out: ① the “5x revenue / $10T” arithmetic is a rule of thumb, not a law (look at incremental cash flow over asset life); ② “AI = Ponzi” overreaches (it’s classic overinvestment, not fraud); ③ “penetrated into credit like 2008” — the channel is real, the scale isn’t (2008 was household leverage; this is giant balance sheets plus private credit, far from systemic saturation); ④ “Wall Street is betting on a $10T expansion and going all-in” is motive attribution, unfalsifiable.
An internal tension: in the earlier essay the author said valuation bubbles “leave infrastructure behind, capital loses, society eats the productivity dividend” (self-healing); in this piece he says AI “penetrates credit like 2008” (system-injuring). Which is it? Reality sits in between: AI capex is funded by a mix of giant equity buffers (being burned), private credit, and chipmakers’ own cash flow. That’s precisely why it’s the one variable that could flip the system — and why its direction is the most uncertain.

III. Data Verification Checklist

US federal debt >$40T: confirmed (first crossed 2026-08-18, Reuters); >120% of GDP, first time since WWII.
Net interest: formally exceeded defense in FY2025; now the #2 budget line, above Medicare. ✅
RRP peak $2.55T (June 2022) → below $400B by 2024. ✅
GENIUS Act: signed 2025-07-18, enforced 2027-01-18. ⚠️ (article says 2025-06, wrong)
Stablecoin market: ~$250-300B global; USDT+USDC ~93%. ⚠️ (article inflated)
Fed balance sheet: $9T peak (Apr 2022), now $6.76T. ✅
KOSPI crash: circuit breakers, -5%/-8.29% single days, won at 17-year lows. ✅
Private-credit redemption caps: KKR/Cliffwater/Blue Owl/Apollo. ✅
UAE swap: real event, but political signaling, not financial need. ⚠️ (author misread)
Giant FCF depletion details (Google/Meta numbers): not independently verified.

IV. My Overall Judgment

Where the series overlaps with my own view, ranking high to low: 8/25 closer > the tricks essay > the next-crisis essay > the anxious-America essay.

Where the author and I genuinely agree:

Chronic debt attrition is the main path; acute is the tail (he calls it a slow/stagflation crisis; I call it chronic attrition — the same thing).
Sovereign debt / dollar credibility is the deepest variable.
AI capex is the swing variable that decides the path (the monetization gap is the right question).
Policy room has narrowed; markets are handcuffed to “assets cannot fall.”

Where we diverge:

He runs the tail event as his base case (Huang Chao, capital controls), while I hold “chronic first, tail elevated.”
His conclusions serve his own positioning (“perfectly closed loop” + gold/RMB Noah’s Ark); I’m wary of self-consistent narratives that can’t be falsified.
A slow crisis is not smooth — the slow path is studded with periodic acute spasms (a bond-market tantrum is a miniature acute event). “Chronic first” and “acute can’t be ruled out” aren’t contradictory; they’re two faces of one process.
The thing to actually guard against isn’t “the US seizing your assets” (that capital-controls scenario is extremely unrealistic) — it’s inflation’s chronic erosion of real purchasing power and the higher real rates from fiscal crowding-out. Those are the certain risks.

The series is worth reading. Its value isn’t in the conclusions (which keep sliding into narrative) but in the frameworks — the bubble taxonomy, the fiscal-tax-base thesis, and the AI capex sustainability question are all working analytical tools. Score 6.5/10: real substance, but you have to wring the water out yourself.