What works and the next five years

Tether next to sovereign states

Tether Limited is a private issuer, domiciled in El Salvador since January 2025, that mints USDT against a reserve of cash and short US government paper. Circle Internet Financial is the listed issuer of USDC, and it is building Arc, the Circle Payments Network and USYC as revenue lines beyond reserve interest.

Tether's Treasury exposure in the knowledge-base series ran from about 98.5 bn USD on 31 March 2025 to 141 bn USD on 31 March 2026. In the Q1 2026 attestation, dated 30 April 2026 and prepared by BDO, Tether called itself the 17th-largest holder of US debt globally. Other sources, on other dates, put the same company at 19th. No confirmed September 2026 ranking was found. Both ranks stay here. Neither is treated as the live figure.

Treasury International Capital data, TIC, is the US Treasury's survey of who holds US securities. Germany's holding on that survey, relayed via FXStreet on 19 May 2025, was 111.4 bn USD. Tether's 127–141 bn USD range in the same comparison sits between Germany and South Korea and Saudi Arabia, both of which have sat below Tether since October 2025, and just under Brazil at 135 bn USD. A later reading, the Q2 2026 attestation (BDO, 31 July 2026), put 114.96 bn USD in T-bills plus 18.63 bn USD in reverse repo, about 133.6 bn USD. That is a different date, not a number to average with 141.

Three disagreeing estimates exist for issuers' share of the T-bill market, and they are not reconciled. Circle and Tether combined were 1.7% of stock on December 2024 data. The Treasury Borrowing Advisory Committee put the share at about 2.5% in October 2024. MIT DCI put daily Treasury turnover through these issuers at about 20%, against that same 1.7% of stock. Stock and flow are different objects. A 1.7% holder that is 20% of daily turnover is a trading footprint, not a stock-of-debt footprint.

A systemic threshold is the point at which a supervisor treats an issuer as large enough that its failure would transmit into the rest of the money system, and steps in as co-regulator. The Bank of England's UK design uses that word for tokens HM Treasury designates; as of 4 September 2026 no issuer, including USDC and USDT used in the UK, had that status. The BIS measured a related market channel directly. A 3.5 bn USD stablecoin inflow cut 3-month T-bill yields by 0.71 bp the same day, up to 4 bp over ten days (BIS Working Papers No. 1270). Outflows push yields up 2–3× more than inflows push them down, because the issuer then has to sell paper quickly.

Four uses that survive a subsidy test

The common feature of the four uses below is that the user pays a market price and comes back, not because of points or an airdrop.

Dollar access for people who cannot hold a dollar account is the oldest documented product-market fit in this industry. Felix Pago, a remittance firm running a USD to USDC to peso sandwich over WhatsApp, had processed more than 8 bn USD and served more than 6 million people in 11 Latin American countries by September 2026, at under 1% against 5–7% on traditional channels. LemFi, a payments firm with a USDT back end and an investment from Tether, handled more than 1 bn USD of monthly payment volume to August 2026, compressing the FX spread to 1.5–2.5% above mid-market against 7–12% on traditional channels into Africa. USDT was about 74% of centralised-exchange volume in 2026 against about 59% of supply (CoinDesk Data/Artemis, 2026). The gap is consistent with some of the supply sitting in wallets as savings rather than spinning in the order book.

The second use is the unit of account on exchanges and in DeFi. Across six lending protocols (Aave V3, Morpho Blue, Spark, Compound V3, Fluid, Euler V2) combined TVL on 4 September 2026 was about 36.5 bn USD (author's calculation, DefiLlama API), with stablecoins dominating the debt side, not only collateral. Hyperliquid held 6.99 bn USD of stablecoins as margin for perpetual futures that day, 97.9% of it USDC, more USDC than Base and Arbitrum combined.

The third is B2B and cross-border settlement in narrow corridors. McKinsey and Artemis Analytics put real stablecoin payments in 2025 at 390 bn USD, of which B2B was 226 bn USD, up 733% year on year (February 2026). The London to Lagos corridor fell from 7.8% all-in in 2023 to 2–3% in 2026. Circle Payments Network had four live corridors with four financial institutions on 10 August 2025 (Circle Form 424B4, SEC, August 2025).

The fourth is institutional collateral. BlackRock's BUIDL, a tokenized Treasury fund, was accepted as off-exchange collateral on Binance from 14 November 2025. Circle's USYC supply on BNB Chain had grown to 1.84 bn USD by March 2026. Deribit and Crypto.com accepted BUIDL as derivatives margin in the first quarter of 2026.

What joins the four is the absence of a cheap alternative. Where that alternative is already free and instant, SEPA Instant in Europe is the type case.

Six tests, applied by the author to twelve constructions, sort that paying demand from a supply ranking.

A compound annual growth rate, CAGR, is the constant yearly rate that takes a start value to an end value over a stated number of years. Forecasts speak CAGR. The six tests do not. They ask what remains when a named factor is removed.

1. The rate test: does the product still make sense if the federal funds rate is 0%.
2. The incentive test: what demand remains when the airdrop, reward or token emission stops.
3. The 50 mn USD liquidity test: can 50 mn USD exit in a day without moving the price more than a few basis points. Kaiko's figure for a 0.1% move in USDC on price-discovery venues was about 38 mn USD sold (21 April 2026).
4. The crisis test: how the token behaved in the last documented stress, not in a hypothetical.
5. The paying-customer test: is there a user who pays a market price today and would return without a subsidy.
6. The 2027-regime test: does the construction survive the payment-stablecoin rules that actually come into force under the GENIUS Act.

USDT and USDC each pass five of six. USDT fails the 2027-regime column (MiCA has already pushed it off licensed EEA venues; GENIUS-shaped USAT is still a separate token). USDC's incomplete mark is the crisis column, which is the Silicon Valley Bank weekend in chapter 1, not a fail on construction. USDe passed the crisis print on venues where the on-chain peg held, and failed the rate and funding-survival tests: sUSDe's seven-day APY ran from 17.2% in April 2024 to about 3.7–4.1% in 2026. PYUSD fails the customer test. Supply tracks the reward, not transactional use. Algorithmic fails all six. It is the only category in this material with a documented collapse to nothing, Terra's UST from above 18 bn USD in early May 2022 to 0.12 USD on 16 May.

Supply and the six tests are different axes. crvUSD at 0.29 bn USD and GHO at 0.57 bn USD pass three of six, the same as the whole tokenized-Treasury class at 16.17 bn USD. PYUSD at 3.01 bn USD passes zero, matching algorithmic at 0.67 bn USD. RLUSD at 2.40 bn USD passes four, more than USDe at 4.27 bn USD. If a product is being built on a named dollar token, the six tests are a better two-year predictor than the supply league table.

Who earns cash is a separate ledger. Tether netted more than 10 bn USD in 2025, with distribution costs near zero on the scale of Circle's Coinbase share. Circle's net margin in the knowledge-base synthesis is 5–7%. Distribution has sat at 58–62% of Circle's reserve revenue since the third quarter of 2025. Same gross T-bill yield, different residual.

The 2026 contraction is the incentive test run in public. USD-pegged supply peaked at 322.4 bn USD on 17 May 2026 and then fell 3.28% into September, the first contraction in this record that did not sit next to a crash. USDT was −2.0% year to date on 4 September 2026 and USDC −1.1%. USDe was −32.1% and PYUSD −16.4%. Tokenized Treasuries and newer fiat-backed names grew over the same window (USDY, USDG, RLUSD, BUIDL, USYC).

Four stories that have not been proven

Mass merchant payments remain a story. PYUSD has access to 439 million active PayPal accounts at end-2025, and 3.00 bn USD of supply on 4 September 2026. Visa's USDC settlement with partner banks from 16 December 2025 is an interbank layer. The cardholder does not see a different swipe. Card chargebacks cost merchants 33.79 bn USD globally in 2025. A merchant who takes a stablecoin takes the risk the card network currently carries, because a push payment has no chargeback.

Yield as a durable product edge is the second story. The White House Council of Economic Advisers, in April 2026, put the modelled effect of a yield ban on bank lending at 531 bn USD even if the market grew sixfold, about 4.4% of total loans (AssetWhisper, 9 May 2026). Wrapper tokens, sUSDe beside USDe, sUSDS beside USDS, still sit in an open regulatory question as of September 2026. Chapter 3 defines the wrapper.

Most bank and corporate stablecoins announced in 2025 and 2026 are the third story. A consortium of 21 banks, announced on 1 September 2026, had disclosed neither a company name nor a token nor a chain by 4 September 2026. That is an intention, not a product. BlackRock's BUIDL is the instrument that wins in the author's bear path, as capital rotates from payment tokens into yield-bearing funds.

Algorithmic and uncollateralized designs are the fourth. Four years after Terra, DefiLlama still classified 23 tokens as algorithmic, 0.67 bn USD against 335 USD-pegged tokens on 4 September 2026, about 0.2% of the market. The design is extinct in this material. Chapter 2 keeps the two Terra loss figures. This chapter only needs the category's documented zero.

Transfer volume is not payments

Headline on-chain transfer volume is the number the industry puts in a slide. The Polish edition's working pair for 2025 is 33–35 tn USD (BIS, Hernández de Cos, April 2026; Visa/Artemis, 2026). The knowledge-base extremes sit in a 28–62 tn USD range. Real-economy payments sit at about 390 bn USD (McKinsey/Artemis, February 2026) or 350–550 bn USD (BCG/Allium, January 2026). Artemis's "Ground Up" survey sits at 122 bn USD run-rate (October 2025). Any pairing of the 28–62 tn USD extremes with the 350–550 bn USD payment range puts real payments at 0.56%–1.96% of transfer volume, a 50–177× gap. After dropping exchange flows, the Polish funnel leaves about 21 tn USD, then 390–550 bn USD, of which 226 bn USD is B2B.

UPI, India's interbank rail built by NPCI, processed 228.3 billion transactions worth about 3.4 tn USD in 2025 (NPCI, July 2026). Pix, Brazil's central-bank rail, processed about 80 billion transactions worth about 6.5 tn USD (Banco Central do Brasil, October 2026). Each, in one country, is larger than the entire global "real" stablecoin payment estimate. Pix alone is about 17× larger. That is a scale check, not a same-category contest, and not proof that stablecoins have failed. UPI and Pix count domestic payments on rails that already exist at both ends. Stablecoins still earn their keep on corridors where that rail is missing.

Three horizons, labelled as scenarios

The starting point for every path below is 311.2 bn USD of USD-pegged supply on 4 September 2026 (DefiLlama API), after a peak of 322.43 bn USD on 17 May 2026 in the author's series. Competing May 2026 peaks sit in chapter 2.

These are the author's compounding paths, per period, not a linear extrapolation of 2024–2025. The 2033 column is an illustrative point on the same paths. None of the nine cells is a prediction.

Each horizon is dominated by a different variable.

2026–2027 is dominated by whether GENIUS Act rules finalise on time. The Act was signed on 18 July 2025. The one-year statutory rulemaking deadline, 18 July 2026, passed with zero final rules from any primary regulator. As of 10 September 2026 the statutory effective date defaults to the 18 January 2027 backstop. In the bear path those rules miss January 2027 and the 21-bank consortium slips past the first half of 2027. In the bull path a new bank token takes 3–5% in its first year, on the pace USD1 ran from 125 mn USD to 4.6 bn USD in 14 months. The checkable signal is a token name and a chain for the consortium before the end of 2026.

2028–2030 is dominated by whether B2B payments scale past the first SWIFT-migration wave. In the bear path, growth pays single-digit rates after that wave, and tokenized deposits take most of the corporate demand. In the bull path, B2B breaks G10-to-G10 corridors through interoperability with those deposits. The bear signal is year-on-year B2B payment growth falling below 50% in two successive annual reports. The base signal is Circle Payments Network corridors above 50, against four in August 2025.

After 2030 the dominant variable is whether the legal line between a payment stablecoin and a tokenized deposit holds or blurs. In the bear path, banks win through deposits that interoperate with each other (The Clearing House). In the bull path that legal line starts to crack. The checkable signal is a change in the GENIUS Act or a successor that formally allows yield on payment stablecoins, which is banned today. Wrapper yield remains an open question until that change, or a ruling, arrives.

Standard Chartered's 2 tn USD by end-2028 (24 February 2026) sits close to the author's 2030 bull of 2,076 bn USD. Citi's 1.9 tn USD base for 2030 (25 September 2025) sits well above the author's 949 bn USD base. The author's base is lower because it assumes the yield ban on payment stablecoins holds for the whole horizon.

The tokenized-Treasury side of the same 2030 argument, scored on turnover rather than supply, sits in What works, what does not, and the next five years.