Part IV — Digital Money

Chapter 26 · Stablecoins: Mechanics and Failure Modes

Chapter 22 sorted stablecoins into one line of a table: private token, sovereign denomination, promise-based par. This chapter opens the promise and looks inside. What exactly backs a three-hundred-billion-dollar market, who verifies it, what a run feels like when the reserves are parked at a failing bank — and why the two instruments that broke, broke for exactly the reasons the taxonomy predicted. The session's tool is the reserve audit: the same scepticism a bank examiner brings, aimed at a whitepaper.

📖 Chapter 26 of 35 ⏱ ~65 min read 🎓 Session 35 · CILO 3 Core 65% Stable 15% Volatile 20% 📝 Quiz · 30 MCQs

🎯 Learning Objectives

1. Anatomy of a Stablecoin

Strip the marketing and every stablecoin is four components bolted together. The issuer — a company, a protocol, or nobody — owes the holder. The reserve — somewhere, something — is what the issuer claims backs the token. The redemption mechanism is the promise to convert tokens into the reference asset at par. The secondary market — exchanges and pools — is where most holders actually exit, at a floating price. A useful first characterisation: a fiat-backed stablecoin is a money-market fund that trades like a token — it takes in dollars, invests the float, and promises a dollar back — minus the regulation, the audit, and (until recently) the interest paid to the holder. The issuer keeps the float's yield. That is the entire business model, and it explains almost everything issuers do.

Four designs cover the field. Fiat-backed (USDT, USDC, PYUSD): reserves in cash and near-cash dollar assets; the dominant family — together the top two hold over four-fifths of the market. Crypto-overcollateralised (DAI, now USDS): locked crypto collateral worth more than the debt, liquidated automatically if it falls — no bank counterparty, but exposure to crypto prices and to the collateral itself (DAI once held USDC). Algorithmic (UST, historically): no reserves at all; the peg maintained by a mint-burn arbitrage against a sister token — the family that produced the genre's defining collapse. Synthetic (USDe): a dollar position built from derivatives hedges — backing of a kind, but nothing a holder can knock on a door and demand. The taxonomy of Chapter 22, turned sideways: the rows differ in who holds what, and who can fail you.

2. How the Peg Holds — and Breaks

Two channels keep a token at one dollar, and they work in tandem. The redemption channel is the anchor: whenever the token trades below par, anyone can buy it cheap and redeem it with the issuer for a full dollar — riskless profit that pushes the price back up. Above par, the mirror trade: mint new tokens with the issuer at one dollar and sell them dear. The arbitrage channel is the secondary market doing the same work at scale. So long as redemption works, arbitrage keeps the price within basis points. The peg is not magic; it is a subsidy to arbitrageurs paid out of the reserve — which is why every peg question is finally a reserve question.

The break points are equally mechanical. First, redemption gates: if the issuer suspends, delays, or quietly throttles conversion, the anchor chain snaps and the secondary price finds its own level. Second, reserve impairment: if the market doubts the reserve covers the tokens — assets frozen, uninsured deposit stuck at a failing bank, loans to affiliates — it prices the doubt immediately; rating agencies call a move beyond three per cent from peg in a day a depeg. Third, the timing mismatch: tokens trade around the clock, but reserves settle in banking hours and Treasury markets — over a weekend, doubt has two days to compound before a single bond can be sold. And underneath all three sits the run mechanic every banker will recognise: redemption is first-come-first-served, so the rational response to doubt is to be first — and a rush of redemptions forces fire sales that impair the reserve further, validating the doubt. The stablecoin run is the bank run, rebuilt on a blockchain, with no lender of last resort.

📝
The missing sentence in every whitepaper: a bank survives a run on central bank support and deposit insurance; a stablecoin issuer has neither. Its run defences are: excess reserves, high-quality liquid assets, and credibility. That short list is the whole audit.

3. Reserve Quality: the Audit Lens

Five questions, in descending order of how often they are ignored. What is in the bag? Composition runs from Fed deposits and overnight repos (money today) through Treasury bills (money this week) to money-market fund shares (money this week, if the fund gates) to secured loans and corporate paper (money someday) to gold and bitcoin (money at the market's pleasure). Who holds it? Custody and counterparty: cash at a bank is an unsecured exposure to that bank — the lesson of March 2023, when $3.3 billion of USDC's reserves sat uninsured at Silicon Valley Bank. Who verifies, and how? The industry standard is the attestation — an accounting firm's assurance opinion that a snapshot of reserve figures fairly reflects records, usually monthly or quarterly. An attestation is not an audit: it is point-in-time, it tests reporting not solvency, and it typically says nothing about whether the assets are encumbered. How fast does it liquidate? Duration and market depth under stress — a fire sale of even Treasuries moves prices at scale. Who profits? The float's yield belongs to the issuer — Tether reported over thirteen billion dollars of profit for 2024, largely reserve income on a portfolio that includes more than a hundred billion dollars of US Treasuries. When the profit motive pulls toward yield — longer paper, riskier lending — the audit question is whether the reserve drifted to earn it.

💡
Pro Tip: Read an attestation the way a banker reads a balance sheet: look for the lines a newcomer skips — loans to affiliates, encumbered assets, "other investments," anything priced by the issuer itself. The credibility of a stablecoin lives in the footnotes, and the exercise at the end of this chapter makes you write the footnote review.

4. The Failure Gallery

Three failures, three different mechanisms — and three different lessons, because one of them recovered.

EventMechanismOutcomeLesson
Terra UST, May 2022Algorithmic: peg rested on mint-burn arbitrage against sister token LUNA; Anchor paid ~20% yield funded by subsidyDeath spiral in days; roughly $40 billion of value destroyed; LUNA hyperinflated to zeroNo reserve, no peg: a design that must be believed, and dies when it is doubted
USDC–SVB, March 2023$3.3bn of reserves stranded as uninsured deposits at a failed bank; weekend timing, redemption impossible until banks openedDepegged to roughly $0.87; recovered to par within days after the US authorities guaranteed all SVB deposits; USDC still shed ~$20bn and half its market shareBanking counterparty risk and the timing mismatch — and the state's backstop, extended to a stablecoin's reserve by accident
Tether disclosures, 2017-21Reserves at periods held commercial paper and loans rather than the Treasuries marketed; regulators found misstatementsNY AG settlement $18.5m (2021) and CFTC penalty $41m (2021); portfolio later rebuilt around T-bills and repos with quarterly BDO attestationsThe gap between the brochure reserve and the actual reserve is the oldest failure mode — and regulation now writes the reserve list

Note what the gallery teaches by contrast. UST failed because there was nothing to run to. USDC survived because its reserves were real, liquid, and — decisively — because the United States chose to make SVB's depositors whole; the Fed's own staff study of the episode records the secondary contagion too: DAI, holding USDC in its stability modules, depegged alongside an asset it did not issue. Contagion in stablecoin land travels through reserve overlap, not branding. And Tether's chapter teaches the slow failure: not a depeg but a credibility debt, repaid in fines first and rebuild later. Three failure modes — design, counterparty, disclosure — and every audit in Section 7 checks for all three.

⚠ Volatile content — Reviewed: November 2026 · Next review: February 2027

5. The Market Today

The scale snapshot, late October 2026 (DeFiLlama and tracker data; supply figures move daily — treat as a reading, not a constant):

MetricValueReading
Total stablecoin market~$300–305 billionPeak $315.5 bn in May 2026; broadly flat over twelve months
USDT (Tether)~$180–183 bn, ~60% shareLargest; heaviest use offshore and in emerging markets
USDC (Circle)~$71–72 bn, ~24% shareThe regulated-issuer benchmark; contracting in late 2026
Others in the top tenUSDS/DAI, USDe, USD1, USDG, PYUSD, RLUSDEach $2–7 bn; a long, thin tail
Where they runEthereum ~50% of supply; Tron ~34%Tron's share is the emerging-market retail fingerprint
Cross-border transfer value$220 bn+ (Jul 2025–Jun 2026, +77%)The usage that worries every central bank in Chapter 25's world

Two readings matter for this course. The stability reading: a third of a trillion dollars of dollar-claims now circulates on public blockchains, and after May 2022 the big two have held their pegs through bank failures and crypto crashes — the reserve model, at scale, has worked so far. The sovereignty reading: that third of a trillion is dollar liabilities issued by private companies, growing fastest exactly where local currencies are weakest — which is why the policy chapter (Chapter 29's fork) treats stablecoins as a monetary question before a consumer one.

📋 Stable content — Reviewed: November 2026 · Next review: May 2027

6. The Law: GENIUS, MiCA, India's Fence

Regulation converged fast, on the audit's exact terms. The United States passed the GENIUS Act (Public Law 119-27, signed 18 July 2025): payment stablecoins must be issued by permitted issuers — federal or state-chartered — holding reserves at at least one-for-one in a closed list: coins and currency, Fed accounts, insured demand deposits (within regulator-set limits), Treasury bills of 93 days or less, overnight repos, and government money-market funds. Issuers must publish a redemption policy and monthly reports of reserve composition, with examination and audit duties; three years after enactment it becomes unlawful to offer a non-permitted payment stablecoin in the United States. The OCC's implementing rule takes effect in January 2027. Read the list again: it is Section 3's audit lens written into statute — the 93-day maturity cap is the liquidity question, the insured-deposit limits are the SVB lesson, monthly disclosure is the Tether lesson.

The EU's MiCA regime (fully applicable since the end of 2024) reaches the same place through its e-money token and asset-referenced token categories: par-claiming single-currency tokens face reserve, redemption and disclosure requirements, with authorised-issuer status and strict rules on the interest incentive. And India, as Chapter 2 of Part IV established, has fenced rather than adopted: no dedicated stablecoin law, tokens inside the VDA perimeter for tax and anti-money-laundering, no recognition as payment instruments, and an RBI position that treats private dollar tokens as a sovereignty risk, not a product. Three regimes, one spectrum — regulate tightly, regulate as e-money, exclude — and the audit method works unchanged across all three.

7. The Auditor's Method

Assemble the chapter into the checklist you will run in the hands-on — and in any future diligence of a token, a fund, or a fintech balance sheet. Ten questions, three verdicts.

  1. Who is the issuer, and what law holds them to the promise?
  2. List every reserve line. Tier it: today / this week / someday / at the market's pleasure.
  3. Which lines are unsecured exposures — to a bank, an affiliate, the issuer itself?
  4. Attestation or audit? Point-in-time or continuous? Who signs it?
  5. What is the redemption policy — gates, minimums, fees, jurisdiction of the holder?
  6. Run the arithmetic: a 30 per cent redemption day, with haircuts. Covered, or not?
  7. Where does the float's yield go — and has the reserve drifted toward yield?
  8. Whose failure breaks you: banks, custodians, other stablecoins in the reserve?
  9. What does the secondary market's worst week tell you about who exits first?
  10. Verdict: credible (audited, liquid, unencumbered, redeemable), conditional (real but concentrated or attested-only), or unauditable (design or disclosure makes the question unanswerable).

The method's spirit is the course's spirit: instruments do not get the benefit of the doubt; they get the questions. Chapter 27 turns to the assets that make no par promise at all — bitcoin and its kin — where the same scepticism produces a different verdict: not "is the reserve real?" but "is the mechanism what the claim says it is?"

🛠 Hands-On: Reserve Audit Exercise

Three attestation snapshots, anonymised but realistic. Audit all three; verdict each.

Reserve lineIssuer A ($140 bn)Issuer B ($72 bn)Issuer C ($18 bn)
Cash and bank deposits5%11%0%
T-bills ≤ 93 days45%4%0%
Overnight reverse repos20%0%0%
Government MMF shares15%85%0%
Secured loans to affiliates8%0%0%
Gold and bitcoin7%0%0%
Sister-token swap rights0%0%100% (algorithmic)
VerificationQuarterly attestationMonthly attestationProtocol code, public
Stated excess reserves5%2%n/a

Steps

  1. Tier each issuer's reserves: today / this week / someday / market's pleasure. State the percentage in the top two tiers combined.
  2. Run the 30 per cent redemption day. Assume haircuts: cash and repos 0%, T-bills 1%, MMF 1%, affiliate loans 50%, gold and bitcoin 30%, sister-token rights 100%. Which issuer covers a $42 bn (A), $21.6 bn (B) run day without selling the "someday" tier?
  3. The attestation critique: three weaknesses of each issuer's verification, in one sentence each.
  4. Contagion: which two issuers could break the third? Name the channel.
  5. Verdict for each — credible, conditional, or unauditable — with the single change that would upgrade it.
View Solution / Walkthrough

1. Tiers. Issuer A: today = cash 5% + repos 20% = 25%; this week = T-bills 45% + MMF 15% = 60%; top-two total 85%; someday = affiliate loans 8%; market's pleasure = gold/BTC 7%. Issuer B: today = 11%; this week = MMF 85% + T-bills 4% = 89%; top-two total 100% — the cleanest table. Issuer C: 0% in every reserve tier; its "backing" is a swap right against a volatile sister token — market's pleasure, 100%.

2. The 30 per cent day. Haircut values: A: cash 7 + repos 28 + T-bills 55.8 (×0.99) + MMF 20.8 (×0.99) + affiliate 5.6 (×0.5) + gold/BTC 6.4 (×0.7) ≈ $123.6 bn usable vs $140 bn face — the $42 bn run is coverable 2.9× over even after haircuts, but only if the someday tier is not sold first; with top-two tiers alone ($123.6−5.6−6.4 = $111.6 bn usable) it is still covered. B: 7.9 + 61.1 (×0.99) + 2.8 ≈ $71.8 bn usable vs $72 bn — the $21.6 bn run is covered 3.3× with no awkward assets at all. C: usable reserve ≈ $0 bn vs a $5.4 bn run — the arithmetic cannot even be performed; the design answers a different question. Conclusion: A and B survive the arithmetic; C was never doing arithmetic.

3. Attestation critiques. A: quarterly (not monthly) point-in-time snapshot; attestation tests records, not encumbrance of the affiliate loans; the affiliate line prices issuer credit at issuer discretion. B: attestation still not an audit — no going-concern or valuation opinion; cash line is an uninsured bank exposure (SVB lesson) until the GENIUS limits bite; MMF shares can gate. C: public code is transparent but transparency of a mechanism is not a reserve; the sister token's value is endogenous — the thing being verified depends on the promise holding.

4. Contagion. If A depegs, B is largely insulated (no A tokens in reserve) but market sentiment repricing all attestations hits B's redemption queue. The sharper historical channel: a bank failure breaks both A and B through their cash lines (SVB, March 2023); and DAI-style issuers holding B's token inherit B's depeg through their own reserves. C needs no help — but C's collapse (the UST template) burns whoever holds C, including other protocols that accepted it as collateral.

5. Verdicts. A: conditional — large, mostly liquid, but attested-only, with affiliate and crypto lines; upgrade = full statutory audit and encumbrance disclosure (which is, roughly, what the GENIUS Act now compels). B: credible — 100% top-two-tier, monthly attestation, stated excess; upgrade = insured-custody proof and a full audit. C: unauditable — there is nothing to audit; no upgrade exists short of becoming a different product (which is what the algorithmic family's extinction demonstrated).

Key Takeaways

1

A stablecoin is four pillars — issuer, reserve, redemption, secondary market — and a money-market fund's economics: the issuer keeps the float's yield, which is why reserves drift.

2

The peg is arbitrage subsidised by redemption; every peg question is finally a reserve question, and the run is a bank run with no lender of last resort.

3

Three failure modes, all seen: design (UST — no reserve, no peg), counterparty (USDC–SVB — uninsured bank deposit, weekend timing), disclosure (Tether's fines).

4

An attestation is not an audit — point-in-time, records-only, silent on encumbrance. The GENIUS Act writes the audit lens into statute: 1:1, a closed asset list, 93-day cap, monthly disclosure.

5

~$300 billion of private dollar claims now rides public blockchains — stable at scale so far, and a sovereignty question precisely where local money is weakest.

Test Your Understanding

Choose one best answer per question. The reserve table in the hands-on is fair data for any item. Aim for 24/30 before moving to Chapter 27.

1. The four components every stablecoin is built from:

2. A fiat-backed stablecoin's core economics, in one line:

3. The stablecoin family with no reserves at all:

4. The redemption channel holds the peg below par because:

5. A depeg, by the working definition used in this chapter:

6. The timing mismatch in stablecoin design:

7. The stablecoin run mechanic, compressed:

8. What the reserve line 'cash at banks' really is:

9. An attestation, precisely defined:

10. The scale of the float business — Tether's 2024 profit:

11. Terra UST, May 2022 — the mechanism and the loss:

12. The USDC–SVB episode ended how?

13. The contagion channel the USDC depeg revealed:

14. Tether's disclosure record produced, in 2021:

15. The failure gallery's three modes, in order of the case studies:

16. The total stablecoin market, late October 2026:

17. The market-share pair, late 2026:

18. The chain-level fact that reads as an emerging-market fingerprint:

19. The GENIUS Act became law on:

20. Under the GENIUS Act, permitted reserves include all EXCEPT:

21. Reading the GENIUS reserve list as codified audit lessons, the 93-day cap is:

22. GENIUS Act disclosure cadence, and the OCC rule date:

23. The EU route to the same destination:

24. India's position on stablecoins, restated from Part IV's opening:

25. The adoption statistic that most worries central banks:

26. In the audit tiering, Treasury bills and MMF shares sit in:

27. In the hands-on's three issuers, the one with 100 per cent of reserves in the top two tiers:

28. The audit verdict for Issuer C:

29. Which single counterparty failure breaks both Issuer A and Issuer B at once?

30. Chapter 27's question, for assets that make no par promise: