Circle's Q1 numbers landed yesterday.

USDC in circulation reached $77 billion, up 28% year on year. Onchain transaction volume hit $21.5 trillion in the quarter, up 263%. Revenue $694 million, up 20%. Adjusted EBITDA $151 million, up 24%.

By any measure these are extraordinary numbers for a company that only listed twelve months ago.

That is not the big story.

The big story is Arc.

Arc is Circle’s own blockchain, currently in testnet but preparing to scale.

Circle raised $222 million in a token presale for its new institutional Layer 1 blockchain at a $3 billion fully diluted valuation. a16z crypto led with $75 million.

The rest of the cap table reads like an institutional ‘who’s who.’

BlackRock. Apollo. Intercontinental Exchange. Standard Chartered Ventures. Janus Henderson. Marshall Wace. Ark Invest. SBI Group.

These are not crypto-native funds taking speculative bets. These are the institutions that price the global cost of capital.

The architecture matters and is elegantly designed: Arc uses USDC as its native gas token, not a volatile crypto asset. Features include sub-second finality, EVM compatibility and optional quantum-resistant wallets at mainnet launch.

Circle holds 25% of initial supply, which means it captures gas revenue and staking yield as the network grows.

Why this matters

It matters because Circle has now moved up the stack.

They are no longer just a stablecoin issuer collecting reserve yield (a margin certain to come under pressure.) They are becoming the operator of the settlement layer USDC runs on.

The float economics, the gas economics, and the staking economics now sit inside one company. The buyer list tells you the rest. When BlackRock and Apollo write cheques into infrastructure, they are not betting on the technology. They are positioning for the rails the next decade of institutional settlement will run on.

As we have said since issue 1: stablecoins are an infrastructure story.

On the same day Circle announced the Arc raise, Tether announced QVAC, MOS, WDK, Pears, and a developer grants program with no cap on total payouts.

Tether is building a different future.

QVAC runs AI inference directly on device: no remote servers, no data leaving the phone.

WDK is an open source kit that embeds self-custodial wallets inside any application.

MOS and Pears extend the peer-to-peer stack.

The grants program pays developers in USDT or Bitcoin for completed work on each layer. No equity, no vesting cliffs and no platform lock-in.

Tether did not need a presale to fund this. The Q1 numbers tell you why.

Net profit of $1.04 billion. Excess reserve buffer at an all-time high of $8.23 billion. Reserves remain US Treasury heavy. They are self-funding the entire stack from cash flow.

The contrast with Arc is structural. Circle is building an institutional Layer 1, regulated under GENIUS, settling into the US financial system, with BlackRock and Apollo writing cheques.

Tether is building infrastructure that routes around financial systems entirely. Local AI. Self-custodial wallets. Peer-to-peer payment rails.

With no custodians, no exchanges and no APIs in the middle.

Why this matters

The Tether thesis is that dollar denomination is one thing and dollar rails are another. USDT preserves the first while attacking the second.

The QVAC, WDK, MOS stack is the operational expression of that thesis. They are building infrastructure that makes USDT reachable anywhere in the world without needing permission from any state, any bank, or any platform.

The two largest stablecoin issuers in the world announced their futures on the same day. They are moving in opposite directions. One is becoming a regulated US settlement rail. The other is becoming a sovereign-resistant stack.

Both are growing rapidly but towards different visions of the future.

Rob Nichols, CEO of the American Bankers Association, sent a letter on Sunday to every bank CEO in the country asking for immediate engagement before Thursday's Senate Banking Committee markup of the CLARITY Act.

The fight is over Section 404.

The Tillis-Alsobrooks compromise prohibits passive stablecoin yield but preserves activity-based rewards. The ABA wants the activity-based language tightened.

Their position is that rewards tied to balance and duration are interest by another name. The numbers do not support the rhetoric.

The Treasury Department estimated stablecoin yield could trigger $6.6 trillion in deposit flight. That number is what the ABA's lobbying campaign rests on. A White House Council of Economic Advisers analysis put additional system lending capacity under an aggressive yield ban scenario at around $531 billion. That is 4.4% of Q4 2025 loan volumes.

Having run a listed bank's treasury, the asymmetry is familiar.

Retail deposits are sticky. Operating cash is operationally captive. The flight risk concentrates in corporate treasury balances and institutional cash.

That money already left for money market funds in the last rate cycle. Stablecoin rewards simply give it a new destination. The Crane 100 already pays north of 3.5%. Banks are not losing depositors to stablecoins, they are losing the depositors who price funding rates against MMFs and short-dated Treasuries.

Why this matters

The CLARITY Act fight is being framed as banks versus crypto. It is something else.

It is the US banking lobby defending a deposit franchise that institutional and corporate cash has already partially exited. The Section 404 markup on Thursday determines who gets to compete for the next tranche of flow. It does not determine whether that flow leaves.

The deposit base that funds American lending is being repriced.

Stablecoin yield is the symptom, not the cause.

Tom Cruise running is cinema shorthand for one thing: something has changed suddenly.

The structure that held the protagonist a minute ago no longer holds. The only response left is velocity.

Every financial run looks the same.

THE TREASURER’S WAR ROOM

Every quarter, as a bank treasurer, I modelled three scenarios.

Base. Adverse. Severely adverse.

The numbers carried specific meaning. Base assumed normal market conditions with modest deposit churn. Adverse assumed a confidence event in the sector, name-specific funding pressure, and meaningful outflows over a defined window. Severely adverse assumed the worst case the regulator could plausibly require us to survive.

A run.

Each scenario produced an outflow profile. Each profile demanded a liquidity response. Levers to pull on:

Government bonds were not equal to quasi-government bonds. Quasi-government bonds were not equal to bank paper. The further down the quality stack the asset sat, the wider the bid-offer in the kind of market where you would actually need to sell. In stressed environments, the illiquidity premium rises.

The point of the exercise was not the numbers. The point was the discipline.

You modelled the run so that on the day a run started, the team already knew the playbook. Which depositors would move first and which counterparties would pull lines. Which collateral could be posted at the central bank window, and which could not. How long the liquidity buffer would last. How much time you had before the buffer ran out and decisions stopped being yours to make.

Every bank treasurer in the developed world runs a version of this exercise. Most of them have never experienced a real run first-hand. The exercise still matters because the moment a run starts, the time to prepare for it has already passed.

A story of 3 financial runs across a markets career.

EXHIBIT A: NORTHERN ROCK (2007)

I watched Northern Rock collapse from my trading desk at RBS in London.

The irony was not lost on anyone who survived the next twelve months as the GFC began.

The wholesale market started to seize up in August. Spreads on bank paper widened. Counterparty limits tightened and term funding got harder to roll. The plumbing the entire banking system relied on (short-dated unsecured wholesale) started to malfunction.

Northern Rock's fragility was baked in. They lent long (25 year mortgages) financed mostly with wholesale paper, weeks and months in tenor.

The asset-liability mismatch was the entire business model. When the wholesale market closed, the funding stopped. And when the funding stopped, the queues started.

The images from September 2007 are now historic - depositors lined up around the block at Northern Rock branches across Britain. The Bank of England announced emergency liquidity support and The Chancellor of the Exchequer was forced to guarantee deposits. The run continued anyway because trust had broken.

A bank run is not about the assets. Northern Rock's mortgage book was performing: the borrowers were paying and the collateral was intact. The bank failed because confidence in its ability to fund itself collapsed faster than it could liquidate to meet redemptions.

Confidence and liquidity. The same two variables every time.

EXHIBIT B: Terra / LUNA (2022)

Fifteen years later I watched Terra/Luna implode from my vantage point as Co-CIO of a digital assets fund that was preparing to launch in Australasia.

We had been live testing in the market and achieving good risk-adjusted returns. The edge was discipline: forty years of combined fixed income portfolio management expertise brought into a new asset class and a focus on low-hanging fruit in young, inefficient digital markets. Everybody else was looking for home runs, we were happy with singles and doubles (with some leverage.)

Funding curves. Contango. Arbitrage.

Mature concepts in fixed income but we combined them with a new understanding of crypto-specific risk management: whitelisting addresses, multi-sig wallets, digital custody architecture.

Above all this was knowing what we didn’t know.

Algorithmic stablecoins sat firmly in that category, so we avoided them.

In May 2022, Terra (UST) lost its peg. The mechanism designed to restore parity instead became the accelerant. Selling UST minted Luna. The Luna supply exploded and Luna's price collapsed. The backstop disappeared into its own arithmetic and a death spiral wiped out US$50Bn of market capitalisation.

Watching a financial asset fall 99% in a single day was new to me. Watching it happen three days in a row was something twenty years in trading had not prepared me for.

The fragility of the design was now visible. So was the speed at which a blockchain project could unravel.

Some analysts, myself included, later argued the unwind was as much an attack as a structural failure. A digital retelling of the Soros and Druckenmiller trade against the Bank of England in 1992. The mechanics are the same. Find a system where the defence is rules-based and finite, then push until the rules break.

The contagion that followed reshaped the digital asset landscape. Three Arrows Capital, Celsius, Voyager, and ultimately FTX. The crypto winter that followed paused capital formation across the sector for eighteen months.

I wrote about it here.

It was crypto’s first bank run. The digital Bear Stearns that preceded crypto’s Lehman moment: FTX.

Run Number 2.

EXHIBIT C: SILICON VALLEY BANK (2023)

Silicon Valley Bank held $211.8 billion in assets.

A significant portion of those assets sat in a Hold to Maturity (HTM) book of long dated US Treasury and agency securities, accumulated near the top of the market.

The collateral was real. The borrowers, the US government, were not going to default.

On a mark-to-market basis, the bank was already gone.

The rate cycle had repriced the HTM book by hundreds of basis points. Unrealised losses sat at roughly $15 billion against a tangible common equity base of around $12 billion. The capital had been eaten by duration risk. Nobody had to write the loss down until the bank had to sell.

On 8 March, SVB announced a $1.8 billion loss on the sale of part of its securities portfolio and a planned capital raise. The disclosure was the trigger. The depositor base was too concentrated: venture-backed technology companies all connected by WhatsApp groups, Slack channels, and Twitter and used to moving capital at speed.

Confidence collapsed inside one news cycle.

By 9 March, depositors had filed withdrawal requests for $42 billion. A quarter of the bank's deposits, in a single day.

On the liability side, the bank was funded by hot money. VC treasuries and startup deposits that moved at the speed of gossip. On the asset side, long-dated US Treasuries bought at the pamdemic yield lows (bond price highs) that would recover their par value but not until maturity.

By the morning of 10 March, SVB was in receivership.

Thirty six hours from disclosure to failure.

The mechanics were the same as Northern Rock: long dated assets, short dated liabilities. Confidence event and an asset-liability mismatch.

The speed was new. Northern Rock's run took weeks but SVB's only took a day and a half. The depositors did not need to queue around the block anymore. They moved as a herd on mobile banking apps while reading the news feeds that told them everyone else was moving too.

Run number 3.

WHY STABLECOINS ARE DIFFERENT

A British mortgage bank in 2007.

An algorithmic stablecoin in 2022.

An American tech bank in 2023.

Different eras. Different technologies. Different depositor bases.

But the same arithmetic.

Confidence collapsed and then redemption demand outpaced the system's ability to honour it. The system design was not fit for the speed of redemption requests.

Each of these institutions had something in common. A maturity mismatch.

Northern Rock funded twenty-five year mortgages with weekly wholesale paper. Terra backed a dollar-pegged liability with a reflexive equity token. SVB funded overnight deposits with ten year Treasuries.

Each had leverage. Each had a structural reason it could not meet redemptions at the speed they arrived.

Properly-structured stablecoins do not have these problems.

Why?

THE BANKING COMPARISONS

A bank operates on borrowed money.

Customer deposits sit on the liability side of the balance sheet. Loans, securities, and other assets sit on the other.

The difference between the two is equity. Equity is small. Liabilities are large. The whole structure runs on leverage. Bank executives describe their balance sheets in terms of loan-to-deposit ratios, capital ratios, and leverage ratios for a reason: the composition of the leverage is the business.

The Tier 1 leverage ratio measures Tier 1 capital against total exposure. The lower the ratio, the higher the leverage. Globally significant banks (G-SIBs) operate in a tight range as they have to - regulators have judged that they pose a greater threat to the financial system if they are undercapitalized.

The point is not that banks are imprudent. The point is that the entire model is leveraged by design. Fractional reserve banking creates credit which creates economic activity. Without leverage, banks do not exist as we understand them.

Properly structured stablecoins do not engage in fractional reserve banking. They hold cash and short dated government securities equal to or greater than the value of stablecoins in circulation.

They do not lend the reserves. They do not transform maturity. They do not create credit.

A USDC issued is a dollar held in a reserve account against it. A USDT issued is a dollar (or equivalent short-dated asset) held against it. The arithmetic is conservative by construction and more akin to a money market fund than a bank.

A bank's core vulnerability is its asset-liability mismatch.

A stablecoin's vulnerability is a different risk management riddle to solve.

THE CONSTRAINT OF THE STABLECOIN MODEL

The asset-liability mismatch in a stablecoin is not a maturity mismatch.

A USDC liability lives on a blockchain. The blockchain is open 24 hours a day, seven days a week, every day of the year. There are no banking hours and no weekend close. A holder anywhere in the world can present USDC for redemption at 3am on a Saturday.

The reserves do not work that way.

USDC reserves sit predominantly in short-dated US Treasury bills, repo collateralised by US Treasuries, and bank deposits. These assets settle on traditional rails. T-bill auctions settle T+1. Repo unwinds during business hours. Bank deposits move on Fedwire during business hours. Money market funds redeem same-day if before 2pm Eastern, otherwise T+1.

The collateral is real. The collateral is high quality.

The collateral cannot be reached at 3am on a Saturday.

In normal markets this gap is invisible. Mint and redeem flows during the week are absorbed by issuer working capital and intraday banking liquidity. The gap exists but it does not bind.

In a confidence event it binds immediately.

A USDC redemption demand spiking over a weekend cannot be met from reserves.

The blockchain processes redemption requests in seconds. The US Treasury market does not reopen until Sunday evening Eastern time which creates settlement and timing risk for the stablecoin issuer.

This is the structural vulnerability of the current stablecoin model. Not insolvency. Not fractional reserves. Not maturity transformation.

A timing mismatch between when liabilities can be presented and when assets can be reached.

T ‘now’ liabilities. Against T+1 or T+2 day settlement for assets.

THE ARCHITECTURAL ANSWER

The solve is structural. Move collateral on-chain.

Tokenised US Treasury bills already exist. BlackRock's BUIDL fund issues tokenised exposure to short-dated US Treasuries on Ethereum. Franklin Templeton's BENJI does the same. Ondo Finance's USDY tokenises T-bill yield. Circle's Arc, raised this week, is built around USDC as gas with onchain collateral integration as a design assumption. Tether's WDK and broader stack are designed to let value move peer-to-peer without traditional rails in the middle.

Different vendors. Different jurisdictions. Different technical approaches. The same direction of travel.

When the collateral lives on the same blockchain as the liability, three things change.

1)      Mark-to-market becomes continuous. The price of a tokenised T-bill updates with the market in real time. No end-of-day NAV calculation. No opaque accounting treatment hiding unrealised losses. The collateral is marked at the price someone will actually pay for it, right now. The SVB problem cannot happen if the asset is marked continuously and the market sets the price every block.

2)      Redemption becomes atomic. A holder presenting stablecoin for redemption can be paid directly from on-chain collateral in a single transaction. No banking hours. No settlement delay. No reliance on intraday liquidity facilities or central bank emergency windows.

3)      The timing gap closes. T+0 liabilities meet T+0 accessible assets. The Saturday morning confidence event no longer has sixty hours to mature into a crisis. The redemption is met or it is not, in the same minute the demand arrives.

This is what a 24/7 stablecoin requires to survive a run.

Not faster banking rails. Not better intraday liquidity. Not a Federal Reserve backstop.

Collateral that lives in the same time zone as the liability.

WHAT STOPS THIS BEING DONE NOW?

If the architecture is structural, why is every stablecoin not built this way already?

Four reasons. Each solvable but not fully solved yet.

Custody and legal title. A tokenised T-bill is a digital representation of a real US Treasury sitting somewhere. The somewhere is the legal question. BlackRock's BUIDL fund holds physical T-bills via BNY Mellon as custodian. The token on Ethereum is a claim against that custody arrangement. Tokenisation has not eliminated custody risk. It has rearranged it and added a new technical layer to it. The legal infrastructure for resolving that gap is still being written.

Settlement plumbing on the cash leg. Atomic redemption requires both sides of the trade to settle in the same transaction. The stablecoin side is on-chain. The cash side, if the holder wants dollars rather than another token, requires a banking system that moves dollars at the speed the blockchain moves. That system does not exist at scale.

FedNow runs during specific hours with transaction limits. RTP runs longer but with similar constraints. A 24/7 wholesale dollar settlement layer is what the architecture actually requires. The Federal Reserve has signalled limited appetite to build it.

Liquidity in the secondary market for tokenised T-bills. BUIDL, BENJI, and USDY trade thinly. Daily volumes are a fraction of the underlying T-bill market. A stablecoin issuer holding tokenised T-bills as reserves cannot necessarily liquidate them at scale during a confidence event. The on-chain market for the collateral is shallower than the off-chain market it represents. The architecture is sound.

The liquidity that supports it at scale is not there yet.

Cross-chain settlement. True atomic settlement requires the collateral and the stablecoin to live on the same chain, or on chains connected by reliable bridging. Cross-chain bridging is the largest source of value loss in crypto history. Until cross-chain settlement is engineered to institutional reliability, the atomic redemption promise is constrained to single-chain architectures.

This is why Arc matters. Circle is collapsing the chains.

These are engineering and regulatory problems. They are not theoretical objections to the architecture.

The architecture is right. The build is not finished.

THE IMPLICATION

Northern Rock had assets. The borrowers were paying. The mortgages were performing. The bank failed because it could not reach the assets fast enough to meet the redemptions.

Terra had a backstop. The Luna mechanism was supposed to absorb selling pressure. The mechanism could not operate at the speed the redemptions arrived.

SVB had collateral. The US Treasuries were money good. The bank failed because liquidating them would have crystallised losses that broke its capital, and it could not access them at the speed the depositors moved.

Every run in financial history has been the same problem in a different costume.

Liabilities arriving faster than assets can be reached.

Stablecoins do not have to fail this way. The technology that creates the redemption pressure also creates the architecture that meets it. The same blockchain that lets a holder redeem in a single transaction can hold the collateral that funds the redemption.

When that architecture is in place, stablecoins will not win only because they are faster or cheaper.

They will win because they cannot run.

-Thanks for reading.

Mark McKendry, Stablecoin State

P.S. If you would like to contact our team just reply to this email - we read every response.

If this issue made you think differently about stablecoins, forward it to a CFO, treasurer, or finance leader who should be reading it.

The conversation is just getting started.