I keep a tab open on my desk that refreshes every sixty seconds. It tracks net issuance across the twelve largest stablecoin issuers. For most of the last quarter, only two names moved the number. Tether. Circle. Together they control roughly 85% of the entire stablecoin float โ a concentration level that sits near an all-time high for the asset class.
Two issuers. One asset. Trillions in annual settlement volume moving through a pipe with two ends.
That statistic gets repeated at every conference panel. It almost never gets traded. And that gap โ between the number everyone quotes and the risk almost nobody prices โ is where the interesting part of this story lives.
I'm an exchange market lead in San Francisco. My job for the past two years has been watching the collateral layer that sits underneath every order book I touch. Exchange leads see the wave before it breaks. Right now the wave isn't a price candle. It's structural. This industry spent a decade arguing about decentralization at the application layer while quietly handing the settlement layer to two companies, in two jurisdictions, answering to two regulatory regimes.
In a bull market, that's a footnote. In a bear market, it's the whole story. When people ask me whether their assets are safe right now, they're not asking about their token. They're asking about the dollar they parked in while they wait.
The first thing to understand is that this didn't happen by accident, and it didn't happen because Tether or Circle built better technology. Neither has ever shipped a meaningful technical innovation. USDT and USDC are, functionally, the same products they were in 2020: a promise, a bank account, and an API.
What they built was distribution.
Tether got there first, in 2014, and spent the next six years becoming the default quote currency on every offshore venue that mattered. By the time the 2020 DeFi summer hit, USDT was already the rail. I was a junior at Berkeley then, live-tweeting liquidity pool mechanics for 72 hours straight, and the thing I noticed โ the thing everyone noticed who was actually in the Discord channels instead of reading docs โ was that nobody asked what backed the pair. They asked whether the pair existed. USDT/BTC existed. That was enough.
Circle took the other road. USDC launched in 2018 positioned explicitly for institutions and regulators: monthly attestations, a licensed money transmitter footprint across the US, eventually a public listing. For most of its life it traded at a structural discount to USDT's float because it was the boring one.
Then came March 2023. Silicon Valley Bank failed with $3.3 billion of Circle's reserves inside it. USDC printed $0.87 on some venues. The peg broke on a Friday night, and the entire industry got a fresh, brutal education in what a stablecoin actually is: not a coin, but a claim โ a redemption right against a balance sheet you cannot fully see, issued by an entity you cannot fully audit.
The FDIC backstopped the depositors. USDC recovered in a weekend. And here's the counterintuitive part: the depeg didn't hurt Circle's growth. It accelerated it. The episode taught the market that USDC was too big to let fail โ a deposit sitting inside the perimeter of the US banking system.
Meanwhile the bear market did what bear markets do. Capital fled into dollars. Not dollars in a bank. Dollars on-chain, because on-chain dollars settle in seconds and earn yield while they sit. Total stablecoin float ballooned even as token prices bled.
And 85 cents of every one of those dollars landed in the same two places.
There's a detail worth sitting with before we go further. Neither issuer publishes a full audit. Both publish attestations โ a snapshot of the balance sheet signed by an accounting firm at a point in time, with no examination of internal controls, no opinion on the completeness of liabilities, and no continuous assurance between reporting dates. That's not a scandal. It's the standard this asset class accepted. It's also a lower standard than any publicly traded company carrying a fraction of the float.
The reserve mixes differ in ways that matter. Tether's has drifted, by its own disclosures, toward short-dated US Treasuries with residual allocations to gold, bitcoin, and secured loans. Circle's is simpler: Treasuries and cash at a handful of custodians. Simpler is not the same as safer. March 2023 proved that concentrated custody reintroduces the exact single point of failure the reserve ladder was supposed to eliminate.
Now let's do what nobody on the conference circuit does, and actually put the number through the math.
Start with the Herfindahl-Hirschman Index, because it's the standard the rest of finance uses and almost nobody in crypto bothers to apply it.
If USDT sits somewhere around 60% of the float and USDC around 25%, the HHI is 60 squared plus 25 squared, plus a long tail that rounds to nothing. That's over 4,200 on a scale where anything above 2,500 is classified as a highly concentrated market. For reference, the US airline industry โ the one everybody complains about โ has historically run in the 2,500 to 3,000 range. Air traffic control has more competitive structure than the dollar rail of the internet.
HHI is a static measure, though. Let me give you the dynamic version that actually matters on a desk.
When a stablecoin market is concentrated, three things change at once.
First, the collateral base becomes a single point of failure for every DeFi protocol that touches it. Not in the price-goes-down sense. In the mechanical sense. Aave's USDC market, Compound's USDT market, Curve's three-pool, Uniswap's stable pairs, every perp venue's margin engine โ they all reference the same two instruments as their unit of account. When those two are 85% of the collateral, the diversification of a DeFi lending book is largely fictional. Fifty protocols, one underlying.
Second, the mint and burn lever becomes a policy tool rather than a technical function. Tether can mint a billion dollars in a day. It has, repeatedly, during risk-on bursts. What it cannot do is unwind a billion dollars in a day. Redemption is throttled by banking rails, by cut-off times, by correspondent banks that open at 9am New York and close at 5pm. Crypto trades 168 hours a week. The dollar rail behind it trades roughly sixty.
That asymmetry is the real risk vector, and it is almost never modeled. In a stress event, the coin doesn't break. The queue does. You will not see USDT at $0.95 on a screen. You'll see it at $0.9995 with a thirty-hour settlement window, and a borrower somewhere down the chain who can't make a margin call because their collateral is real but not yet redeemable.
Third โ and this is the part I care about most โ the concentration is what makes the peg credible in the first place.
That's not a typo. Fragmentation in a settlement asset is a liquidity disaster. Forty stablecoins fighting for float means forty shallow books, forty market-making operations, forty redemption regimes, and spreads wide enough to eat the yield on any trade you actually want to make. The 85% is a risk. It's also the feature that made dollars on-chain usable at all. Both things are true simultaneously, and any analysis that picks one side is marketing.
Here's where the concentration number gets worse, not better, when you zoom out.
The 15% long tail is dominated by three categories, and I want to walk through how much of each one ultimately roots back into the top two.
Crypto-collateralized stablecoins. DAI, now USDS, is the poster child. Read the collateral composition and you'll find a very large share of the backing is USDC plus tokenized Treasury products. The decentralized stablecoin is holding the centralized one and calling it diversification. That's not a knock on the team โ they solved the governance problem and rented the reserve problem. It's a knock on the label.
Synthetic dollars. Ethena's USDe is the most interesting structure to launch in three years: a delta-neutral book, long spot, short perps, funding rate as the yield source. Genuinely clever. Also genuinely dependent on two things โ a functioning perpetual futures market, and collateral custodied by a small set of providers. When funding rates flip negative for a sustained stretch, the yield stops being revenue and starts being a subsidy paid out of the reserve. Same disease liquidity mining had in 2020, wearing a hedge fund's suit.
Yield-bearing wrappers. Every high-yield stablecoin product is, at the bottom of the stack, a Treasury bill ladder with a marketing budget. That's fine. Treasuries are the cleanest collateral that exists. But it means the end investor's exposure chart converges on the same node no matter which wrapper they picked.
Add it up and the 15% alternative is really a 15% sliver, of which maybe a third has reserves that don't ultimately trace back to Circle, Tether, or a custodian holding the same duration risk. The honest concentration number isn't 85%. At the collateral root, it's closer to 93 to 95%.
And there's one more layer, the one I keep coming back to when people ask why the stablecoin float keeps growing. A meaningful slice of that float is not demand. It's incentive. Every time a lending market or a perp venue subsidizes a USDC or USDT pool โ points programs, boosted yield, whatever the branding department landed on โ the float number gets a little less honest. Incentive-driven deposits are the same disease liquidity mining had in 2020: the APY isn't revenue, it's the project buying a supply figure it can put on a dashboard. Stop paying and you find out in a week exactly how much of the total value locked was ever a user.
I know that shape of error intimately. In March 2025 I put $5,000 of my own money into three autonomous trading agents on a new decentralized exchange. I didn't write the bots โ I managed their social presence and watched the logs like reality television. Two of the three ended the month down. The one that survived had the simplest strategy, the fewest dependencies, and the smallest surface area for things to go wrong. Dependency count is destiny. Every additional promise in your stack is a place where the stack can fail.
Which brings me to the analogy that keeps nagging at me, because it's the same mistake in a different costume. For two years the infrastructure conversation was dominated by dedicated data availability layers โ teams building bespoke consensus and storage networks for rollups that push a few dozen kilobytes a day. The architecture got built for the myth, not the load. The stablecoin market has the inverse version of that problem: everyone is funding the myth of decentralization on top of a base layer more concentrated than any industry they'd complain about over dinner.
Let me get concrete about what a concentration event actually looks like, because I think most people are picturing the wrong failure.
They picture Tether collapsing, USDT going to zero, contagion.
That's not the realistic path. The realistic path is operational.
Step one: a macro shock. Some combination of a Treasury market dislocation, a banking stress event, or a jurisdiction doing something unexpected. It doesn't have to be existential. It just has to be inconvenient.
Step two: the discount. Secondary markets reprice first, because they always do. USDT or USDC trades at $0.9950 on an offshore venue. The fifty-basis-point gap looks like free money to a hundred desks at once.
Step three: the queue. Everyone who wants to close that gap has to redeem at par. Redemption requires an issuer KYC process, a banking rail, and a settlement window measured in hours. Creation happens fast. Redemption happens slow. Mints are effectively permissionless. Burns are gated by a compliance stack designed by people who have never had to close a position at three in the morning.
This is where the compliance architecture bites, and it's worth being precise about who it bites. The KYC gate on the primary market is real. It's also largely decorative as a risk control, because the secondary market has no such gate โ you can hold a meaningful position in USDC across multiple wallets and never once complete institutional onboarding. The compliance cost doesn't land on the person trying to route around it. It lands on the honest user who fills out the paperwork correctly, waits three business days for verification, and then discovers their redemption is queued behind a desk that finished onboarding faster because it had a legal team.
I've watched that dynamic play out at the scale of an exchange. Which brings me to the dinner.
In late 2025, when the US framework finally landed โ the stablecoin law the industry had spent two years lobbying for, plus the EU regime that had been phasing in since mid-2024 โ I did something I'd been wanting to do for months. I hosted a dinner in San Francisco for ten people: developers, two policy folks, and a couple of market makers who've been through more stress events than anyone in the room wanted to admit.
We didn't record a press conference. We ate and argued. And the thing that came out of that table, the thing I typed into my notes before I got to the car, was this: the new rules don't address concentration at all. They formalize it.
Think about what the framework actually does. It requires reserves, disclosures, redemption policies, licensing. Who can comply with that? Entities with reserve assets, audit budgets, and banking relationships. Entities with a compliance department. In other words โ the two companies that were already 85% of the market.
Everyone else gets a permissioned lane, a moratorium, or a trip to a smaller jurisdiction.
I've said before that regulation doesn't move at the speed of a block, and it never will. But there's a corollary from that dinner table worth holding onto: regulation is a moat shaped exactly like the incumbent's balance sheet. The 15% long tail now has to clear a bar designed by people who assumed the long tail already had the resources to clear it. Most of them don't.
That's not an argument against the rules. Unregulated dollar rails at systemically important scale were never going to survive contact with Congress. It's an argument against pretending the rules solve the concentration. They ratify it.
Here's where I'll disagree with nearly everyone I talk to about this.
The consensus read is straightforward: 85% concentration equals systemic risk equals danger, and the fix is decentralized stablecoins. It's the argument in every thread, repeated now with the confidence of people who decided the answer before doing the arithmetic.
My read is different, and I'll be honest that it's a minority position even on my own desk.
The acute risk isn't that Tether or Circle fails. It's that they succeed so completely that they get reclassified. Once a stablecoin issuer is regulated as a utility โ reserves, disclosures, licensing, redemption guarantees โ it stops being a growth asset. Utilities get rate-capped. They get treated like money transmitters, which is exactly what they are. The float keeps growing, the revenue gets squeezed, and the equity story dies. That's not a crash. For anyone holding exposure, it's worse: it's a slow repricing into the return on a T-bill ladder minus operating costs.
Which means the decentralization narrative doesn't get resurrected by idealism. It gets resurrected by economics. When the regulated issuer's spread compresses to nothing, the 15% tail suddenly has a business model that doesn't require being the biggest.
And the second half of the contrarian take: the concentration is not nearly as fragile as the thread implies. Two issuers with a couple hundred billion in combined float, most of it in short-dated Treasuries, is a structure that has survived a bank failure, a depeg, two Congresses, and a full bear market. The scenario that actually breaks it isn't a headline. It's the queue, in a week when the Treasury market can't absorb the sale.
Which is exactly the scenario nobody backtests, because the data to backtest it with is locked inside two private companies.
Here's the bear market lens I keep applying. Total stablecoin float is the only number on my board that reflects actual dollars sitting in the system rather than the market's opinion of a token. When it contracts, capital is leaving crypto. When it grows while prices fall, capital is waiting. It's the closest thing this industry has to an M2. Which is precisely why it matters who issues it.
So if you're reading this in a bear market, asking whether your assets are safe โ here's what I'd actually watch. Not the price of anything.
One: the daily market-share split between USDT and USDC. Not the combined number. The split. If USDC's share climbs while USDT's falls, that's a migration toward the regulated perimeter, and it reprices regulatory risk in both directions. If the combined share climbs past 88%, you're watching a settlement layer consolidate into a single operational dependency.
Two: the skew in the major stable pools. When one leg of a three-asset pool holds 70% of the liquidity, the market is telling you something about redemption confidence before it shows up in price.
Three: the redemption window. If any issuer shortens or lengthens its published settlement time, that's the most honest signal in this entire market. It tells you what they believe their banking rail can absorb. It will never appear on a conference panel. It'll be a quiet line change on a terms page.
The question I keep coming back to isn't whether two companies should control 85% of the dollar plumbing. It's whether an industry that spent a decade promising trustless settlement can admit that its base layer is a promise from a company with a bank account โ and start pricing it that way.
Speed is the pulse of the market. Structure is the heartbeat underneath it. Most people never hear the second one until it skips.