Brian Reed: Eliza, hey — I've been sitting with something all morning that I genuinely cannot make sense of and I need you to help me.
Eliza Ward: Let's hear it.
Brian Reed: Celsius told people they'd earn 18% annually on Bitcoin. And people did earn it — for a bit. Then Celsius collapsed, couldn't pay withdrawals, gone. Then BlockFi bankrupt. Then FTX. And now in 2026 the yields are back. Same ballpark. How is that — I mean, is everyone just choosing not to remember?
Eliza Ward: No, actually — and this is the part that reframes everything. It's not memory. The problem is that the yield felt real because it was real, for whoever got out early. Celsius was collecting actual interest, lending to borrowers like Three Arrows Capital, and passing it on.
Brian Reed: So the first depositor who got paid — they weren't being lied to?
Eliza Ward: Not in any simple sense. Which is — okay, that's the scarier version of this story. The fraud-adjacent chaos at FTX, Alameda Research's balance sheet, those were knowable failures in retrospect. Celsius was something different: a structure that worked until it didn't, and nobody outside could see the maturity mismatch building.
Brian Reed: The part I don't get is — if the structure is the problem, why does fixing it feel like it's still four years away? The BIS only published a call for prudential regulation in April of this year.
Eliza Ward: That's the question. Because new entrants aren't waiting for Basel to catch up.
Brian Reed: Okay but — new entrants aside — let me try to make the yield mechanism concrete, because I think that's actually the piece we skipped. Imagine someone hands their savings to a friend, the friend lends it to a day-trader who's on a six-month winning streak, and at the end of the month you get a cut. While the trader keeps winning, that money is absolutely real. You received it. You spent it.
Eliza Ward: Right — that's what Celsius was. Deposits in, lent to borrowers like Alameda Research, Alameda used leverage to turn maybe a dollar of equity into ten dollars of market position, market goes up, they pay interest, Celsius passes it on.
Brian Reed: So if the yield was real, what exactly was the lie?
Eliza Ward: There wasn't one — not about the yield. That's the thing. The part that wasn't disclosed, and couldn't be independently checked, was that Celsius was fractional reserve. Same architecture as a bank. Fraction of deposits liquid, rest deployed. But without deposit insurance. Without capital buffers. Without any stress test.
Brian Reed: Hold on. So the whole thing depended on the trader never having a bad month.
Eliza Ward: Exactly. And — wait, this is the sharper version — when Alameda did have bad months in 2022, the margin calls came in, collateral got sold, prices dropped further, more margin calls. Cascade liquidation. BlockFi gets dragged in. FTX gets dragged in. And suddenly everyone wants their deposits back at once from platforms holding maybe a fraction of that in liquid reserves.
Brian Reed: So the structure was — I mean, it's bank-like liabilities, but none of the bank-like safety nets.
Eliza Ward: That's the narrowest claim the evidence supports, yeah. The information asymmetry is what sealed it — Celsius held the full picture of who they'd lent to, at what leverage, on what terms. Depositors had the yield number and nothing else.
Brian Reed: And if Alameda's leverage ratio had been public — if any of that was auditable — does the math still blow up the same way?
Eliza Ward: Probably blows up the same way — but slower, and with fewer people inside the building when it does. That's the distinction that actually matters. Because imagine someone — a nurse, end of a night shift, November 2021 — logging into Celsius on her phone to check her balance. She can see her yield. She can see the number going up. What she cannot see: whether Celsius has rehypothecated her Bitcoin. Pledged it as collateral for Celsius's own borrowing. The same coins showing in her account are possibly collateralizing a loan she has no knowledge of. There's no audit. No regulator required one. The only signal available to her is the yield, and the yield is real.
Brian Reed: Rehypothecation — so her coins are doing double duty.
Eliza Ward: Multiple claims on the same underlying asset. And she has no independent means to verify it's happening. Not the reserve ratio, not the loan counterparty identities, not the collateral quality. None of it. That's not a sophistication failure — that's a structural information asymmetry. The architecture hid it.
Brian Reed: Hang on — couldn't a sophisticated user have at least known the risk in the abstract? Like, read enough to suspect it?
Eliza Ward: Known in the abstract, yes. Verified independently — no. That gap is everything. Suspecting rehypothecation is possible and proving your specific coins aren't rehypothecated are completely different things. And the moral hazard underneath it is — I mean, this is the part that's almost elegant in a terrible way — Celsius and their borrowers took the upside. Depositors took the downside. Opacity is what made that arrangement stable for as long as it was.
Brian Reed: So the nurse on her phone was making the only rational choice the information allowed.
Eliza Ward: Exactly — and it's still happening. Tether, $184 billion in circulation as of February 2026, no full published audit of its backing assets. Paolo Ardoino runs the thing and the reserve composition is still not independently verified. That's not legacy opacity. That's live infrastructure the entire dollar-denominated side of crypto is sitting on right now.
Brian Reed: Wait — $184 billion and still no audit?
Eliza Ward: None that resolves the core question. And — it gets worse before it gets better — the cascade that took down Celsius, then BlockFi, then FTX in sequence in 2022 runs directly through that same opacity problem. The margin calls, the forced liquidations, the maturity mismatch hitting all at once — that's the mechanism we haven't fully pulled apart yet.
Brian Reed: Pull it apart then — because the thing I keep getting stuck on is the sequence. It's not that one platform failed. It's that Alameda getting margin-called in 2022 somehow ends with BlockFi in bankruptcy court. What's actually connecting those?
Eliza Ward: Forced selling. Alameda faces margin calls, has to liquidate collateral fast — and that selling pushes prices down further. Which triggers margin calls on the next leveraged position. Which means more forced selling. You can't stop it once it starts.
Brian Reed: Each default accelerates the next one.
Eliza Ward: Right — and meanwhile Celsius, BlockFi, they owe withdrawals on demand. Today. But their assets are locked in loans. Weeks, months out. That maturity mismatch is — I mean, it's fine in a slow market. It is instantly fatal when everyone wants out at once.
Brian Reed: So picture someone — say, a grad student in São Paulo, December 2022, watching crypto prices crater on her laptop, trying to pull her savings out of BlockFi. And BlockFi literally cannot send the money because the money is already deployed somewhere illiquid.
Eliza Ward: And that ends in bankruptcy court. That's the actual legal endpoint — BlockFi couldn't meet customer obligations, formal proceedings. FTX is the sharper version: Alameda's leveraged positions were entangled with FTX customer funds directly, and that entanglement wasn't disclosed to anyone.
Brian Reed: Wait — so FTX customers weren't just exposed to market risk. They were exposed to Alameda's specific bets.
Eliza Ward: Undisclosed. That's the word. Not a market failure — a concealment of the balance sheet. Which is actually what makes the BIS paper from April 2026 so pointed. Entity-based, activity-based prudential rules. Capital buffers. Liquidity requirements. Stress testing. Because none of that existed for any of these platforms in 2022.
Brian Reed: And four years later it's still a proposal, not a requirement. The structure that blew up Celsius, BlockFi, FTX — it's architecturally intact.
Eliza Ward: Architecturally intact — that's the exact phrase. And the reason it stays intact is almost boring once you see it: the economics of rebuilding are better than the economics of waiting for BIS to finalize capital buffer percentages. Perpetual futures, synthetic leverage — the risk didn't evaporate after 2022, it just moved into instruments that look different enough to attract a new cohort.
Brian Reed: And the thing that actually breaks the cycle — I mean, a traditional bank that tried to offer 18% would be shut down before it opened. Capital requirements, stress tests, the whole prudential framework makes it structurally impossible. Crypto lenders just never had that ceiling.
Eliza Ward: That's the BIS paper's actual argument. Not that people are reckless — that the absence of those requirements is what makes rebuilding rational. Until capital buffers exist for crypto intermediaries, someone will always run the structure again for whoever missed the last collapse.
Brian Reed: The lesson isn't about greed. It's just — what happens when you run a bank without the rules that make banking survivable.
Eliza Ward: Yeah. That's the one. I'll take that.
