USDT is not exactly a dollar — and other things your balance hides

You open the exchange, see a number ending in USDT, and your brain immediately converts it to dollars. Or to euros, if that's where the rent comes from. The number feels solid. It isn't.

Between that USDT figure and the fiat amount you're picturing, three different markets sit stacked on top of each other, each one running its own price for the same notional thing. They mostly agree. Sometimes they don't. And when they don't, the gap can be a full percent or more — small enough to ignore on most days, large enough to ruin a withdrawal at the wrong moment.

This post is the unglamorous tour of how that stack actually works. What USDT is pegged to (the answer is more complicated than "the dollar"). When it has actually broken and by how much. How your balance gets converted to your home currency, and why two trackers showing the same wallet can disagree by a percent. None of this is a reason to panic — stablecoins work. But the fiat number on your screen is always a translation, and translations have errors.

What USDT actually is

Tether — the company that issues USDT — runs a fairly simple promise. For every USDT in circulation, they claim to hold one US dollar of equivalent reserves. Hold a USDT, and you can in principle redeem it directly with Tether for one dollar.

The word doing all the work in that sentence is "in principle." Direct redemption from Tether requires a verified corporate account with at least $100,000 minimum, a verification process that takes weeks, and a 0.1% fee. The vast majority of USDT holders never redeem directly. They sell it on an exchange to someone else, or they swap it for a different stablecoin, or they spot-trade it into BTC and out the other side.

What that means in practice: USDT's price is not enforced by Tether. It's enforced by arbitrage. If USDT trades at $0.98 anywhere, someone with the relationship to redeem at $1.00 buys $98m of USDT and gets $100m of dollars back, and the price snaps to par. If USDT trades at $1.02, someone deposits $100m at Tether, mints $100m of USDT, sells it for $102m, and the price snaps back the other way.

That arbitrage usually works. When it doesn't — when redemption gets temporarily restricted, or when the people with the relationships are too scared to use them, or when the supply of arbitrageurs runs out — the peg drifts. Sometimes a lot.

The depeg history, plotted

Most charts of USDT show a flat line at $1 with occasional pixel-thin wobbles. That's because tracker apps round to two decimals and the typical deviation is $0.001 to $0.005. Zoom in to four decimals over the last seven years and the chart looks different.

The three obvious dots on that chart:

October 2018, USDT to $0.85. The Bitfinex/Tether relationship was under intense scrutiny, banking partners were dropping the company, and a brief panic pushed USDT down to roughly 85 cents on Kraken before arbitrage dragged it back. Most of the move resolved within 48 hours.

March 2020, USDT to $1.05. This one was the opposite direction. During the COVID liquidity crisis, dollars became scarce in crypto, and USDT — being the easiest dollar-substitute on most exchanges — traded at a premium. People paid $1.05 to hold $1 of synthetic dollars because they couldn't quickly get real ones onto exchanges. The premium lasted about two weeks.

May 2022, USDT to $0.95. The collapse of Terra/Luna and its UST stablecoin caused contagion fears for every other stablecoin. USDT briefly hit roughly 95 cents on Curve and several centralised exchanges before redemptions and arbitrage closed the gap. It took about a week to fully reset to par.

A useful comparison: in March 2023, USDC — generally considered the cleanest, most reserve-transparent stablecoin — depegged to roughly $0.87 over a weekend after the Silicon Valley Bank collapse stranded $3.3 billion of Circle's reserves. USDT was largely unaffected. Different stablecoins, different vulnerabilities. (CoinDesk's coverage is the cleanest write-up if you want details.)

The honest takeaway: stablecoin pegs hold most of the time, fail occasionally, and recover within days to weeks when they do. They are not zero-risk fiat substitutes; they're dollar-equivalents with operational risk attached.

Now layer the fiat conversion

USDT being not-quite-a-dollar is only the first layer. If your home currency isn't USD, there's a whole second layer.

Most exchanges and trackers display your balance in your local currency by computing:

balance_in_local = balance_in_USDT × (USDT/USD rate) × (USD/local_FX rate)

Each of those rates carries its own error.

The USDT/USD rate is the one we just discussed — usually within 0.1% of par, occasionally further.

The USD/local FX rate is messier than people think. Apps don't all use the same source. Some use the European Central Bank's daily reference rate, fixed once at 16:00 CET. Some use a real-time aggregator like Open Exchange Rates or Frankfurter. Some use Coinbase's internal rate. Some use Yahoo Finance, which itself aggregates from several feeds. These sources can disagree by 0.2% to 0.5% at any given moment, especially during off-hours when the underlying interbank market is thin.

Which means two portfolio trackers showing the same wallet can show balance-in-EUR differing by 0.3 to 0.7 percent without either being wrong. They're using different reference rates and possibly stale ones.

There's also a third layer that most trackers hide: the direction of conversion. A USDT-to-EUR conversion via an exchange will execute as USDT → USD → EUR through actual market orders, with two spreads stacked. The number your tracker shows is always optimistic relative to the rate you'd actually receive if you tried to extract the funds.

Build it yourself

The interactive below lets you set each layer's deviation and watch the final number drift:

Default values are sensible: USDT at par, USD/EUR around 0.93 (roughly the 2026 average), 0.10% exchange spread per leg. The naive estimate at the bottom shows what someone using a simple "multiply by FX rate" tracker would compute. The actual delivered amount is always lower, even when nothing has gone wrong, because of the spread.

Push USDT/USD down to $0.95 — the May-2022 scenario — and the gap on a $10k position widens to about $50, plus another $10 from the spreads. Small numbers in dollars; meaningful at higher balances; substantial during a stablecoin event.

When this matters and when it doesn't

For most people, most of the time, none of this matters. You hold a balance, you trade against it, you eventually convert some out — whatever rate you get when you actually convert is the only one that's real. The displayed number in between was always an estimate.

Where it does matter:

Withdrawals during stress. If you're trying to off-ramp USDT to fiat during a depeg event, you eat the gap. A stablecoin trading at $0.96 means a 4% haircut between displayed balance and delivered fiat — and exchange withdrawal queues during these moments tend to lengthen, so you also get to wait a day for the privilege.

Tracking PnL across currencies. If you funded an account with EUR, traded it in USDT pairs, and want to know your real PnL in EUR, the FX rate you use for the conversion changes the answer materially. A 1% favourable FX move can flatter results that otherwise underwhelm. A 1% unfavourable FX move can hide a profitable strategy.

Tax reporting. Most jurisdictions require fiat-equivalent values at the moment of each transaction. The reference rate you use isn't optional; it's prescribed by your tax authority and may not match the rate your tracker shows. Worth checking before you assume the export from your portfolio app is filing-ready.

Cross-stablecoin arbitrage. If you hold balances in both USDT and USDC, and one depegs while the other holds, your actual purchasing power in those balances is no longer 1:1. Trackers that report "stablecoin holdings" as a single number flatten this real difference into a fiction.

The other stablecoins

USDT isn't the only one of these. The major options as of 2026:

• USDC (Circle). Reserve-transparent, US-bank-held, audited monthly. Cleanest collateral story, but as March 2023 demonstrated, exposure to a single failing bank can crater the peg fast.
• DAI (MakerDAO). Crypto-collateralised, partially backed by USDC, partially by ETH and other on-chain assets. More resilient to traditional banking failures, more vulnerable to crypto-specific liquidations.
• FDUSD (First Digital). Hong Kong issued, much smaller market cap. Rose with Binance promotion in 2023-24.
• PYUSD (PayPal). Issued via Paxos. Limited adoption outside PayPal's own rails.
• TUSD (TrueUSD). Has had reserve transparency questions raised over the years; market cap sits well below USDT and USDC.

Each one has different reserve composition, different audit cadence, and different operational risks. They're not interchangeable in a stress event. A diversified stablecoin allocation is, mathematically, a slightly different bet than holding any single one — though most retail traders never make that decision consciously.

For market data on what we're trading against on the bot side — actual BTC, ETH, SOL, XRP price moves rather than the unit-of-account they're priced in — the BTC and ETH correlation post shows what coordinated stress looks like in real-time.

Reading your real balance

A short checklist for being honest with yourself about what your numbers mean:

1. The displayed balance assumes USDT = $1. It doesn't.
2. The fiat-equivalent number uses some FX rate — which one is rarely shown. Check.
3. Trackers showing the same wallet can disagree by 0.3 to 0.7 percent. They're not necessarily wrong, just using different references.
4. The withdrawal rate is always worse than the displayed rate by at least one spread, often two.
5. During depeg events, the gap between displayed and deliverable widens fast. Withdrawal queues get long.
6. Holding USDT is not the same risk profile as holding USD. It's similar enough that day-to-day it doesn't matter; different enough that the rare events do.

Nothing here changes the case for using stablecoins. They're useful, they mostly work, and the alternative — full fiat round-trips for every trade — is its own kind of expensive. The point is just that the number on screen is always a translation. Translations have errors. Most days they're tiny. Some days they aren't.

If the layered-conversion problem made you curious about how prices themselves get reported across markets, the magnetic levels post covers a different version of the same theme — what looks like a clean number is almost never as clean as the display suggests.