Three Arrows Capital didn’t collapse because Luna crashed. Luna crashed in May 2022 and cost the fund about $200 million, real money, but not fund ending money for a firm that had managed close to $10 billion. What actually ended 3AC was a credit structure built two years earlier, one that had nothing to do with Terra, and that turned an ordinary bad month into an unrecoverable one. Tracing the cash backward through Teneo’s liquidator affidavits, filed in 3AC’s bankruptcy proceedings and reported in detail by Decrypt and Benzinga, gets you a different story than the one everyone already knows.
Start with the trade that made them famous, not the one that killed them
By late 2020, 3AC had built the largest position of any institutional holder in Grayscale’s Bitcoin Trust, GBTC. The trade was mechanical and, for a while, close to free money. Grayscale would accept Bitcoin and issue GBTC shares in return. Genesis, the crypto lender owned by the same parent company as Grayscale, would lend 3AC the Bitcoin to put into the trust in the first place. GBTC then traded on the public market at a persistent premium to the actual Bitcoin backing it, sometimes over 20%, because for years it was one of the only regulated ways for certain investors to get exposure at all.
3AC didn’t just buy GBTC. It borrowed the Bitcoin from Genesis to create the shares, then pledged the resulting GBTC shares back to Genesis as collateral for still more loans, and used that fresh capital to build an even larger position. Genesis was a willing and structurally motivated counterparty throughout; the trade generated fees and volume for its own affiliated trust. By the time the position was fully built, Genesis alone had extended 3AC roughly $2.36 billion, with a margin requirement of just 80%, meaning the loan was undercollateralized relative to its own stated terms before anything even went wrong.
The flaw sat inside the six month lockup on newly issued GBTC shares, a rule of the underlying trust structure, not a choice 3AC made. As long as the premium held, the lockup was irrelevant. It didn’t hold. Starting in 2021, GBTC flipped from trading at a premium to a persistent discount to net asset value, and it stayed there. The position that had generated the fund’s reputation was now underwater, illiquid for six months at a stretch, and still serving as collateral for billions in loans that didn’t care what the shares were actually worth on any given day.
The pivot, and the part that actually broke
Losing the GBTC trade’s edge didn’t bankrupt 3AC by itself. What it did was push the fund toward a new source of return to cover the same debt load: leveraged bets on Layer 1 tokens riding the 2021 bull market, Solana, Avalanche, Near, and Luna among them. This is the part of the story that gets told as “3AC bet big on Luna and got wiped out.” It’s true as far as it goes. It also skips the fact that the fund was already running a multibillion dollar undercollateralized loan against a position it couldn’t sell, before Luna ever entered the picture.
When Terra’s UST lost its peg in May 2022, 3AC’s Luna exposure went to close to zero within days, for a loss the fund’s own affidavits put at around $200 million. On a $10 billion peak book, that’s a real wound, not a fatal one on its own. What made it fatal was that 3AC had no ability to raise cash anywhere else. The GBTC position was still locked up and still worth less than the loans against it. Every other altcoin bet had cratered in the same broad selloff Luna helped trigger. There was no dry powder left, because the dry powder had already been spent defending a trade that had already failed a year earlier.
The actual forcing event, the one that put 3AC into a BVI courtroom rather than a difficult quarter, was Voyager Digital calling in a loan. Voyager had lent 3AC roughly $661 million in a mix of stablecoins and Bitcoin, per Benzinga’s creditor list, and when it demanded repayment in June 2022, 3AC couldn’t produce it. Genesis had already been unable to get satisfaction on margin calls of its own by that point either, according to the liquidators’ filings. A fund with $10 billion in assets under management eighteen months earlier couldn’t meet a $661 million call. That gap, not the Luna headline, is the number that actually explains the bankruptcy.
What happened once the courtroom got involved
A British Virgin Islands court ordered 3AC into liquidation on June 27, 2022, appointing Teneo’s Russell Crumpler and Christopher Farmer as joint liquidators, according to filings later summarized by Ogier and Erskine Chambers. Su Zhu and Kyle Davies did not show up to help. Teneo’s own filings describe founders who were, in their words, providing only cursory disclosures and were believed to be somewhere between Bali and the UAE, jurisdictions the liquidators specifically noted as difficult for enforcing foreign court orders. When ordinary legal service failed, Teneo got court approval in the US and Singapore to serve subpoenas over Twitter and email, because Davies, evading process servers in person, was still posting from the same accounts the whole time. I don’t know of many bankruptcy proceedings where the debtors were legally served through a platform they were using to promote their next venture.
That next venture, as it turned out, was a marketplace called Open Exchange, launched in 2023, for trading illiquid claims against bankrupt crypto firms, including FTX, Celsius, Voyager, and 3AC itself. Zhu and Davies built a business around the tradeable wreckage of the fund they’d just run into the ground, while its liquidators were still trying to locate their assets. I’m not sure there’s a cleaner single detail in this entire case for what a lack of consequences looks like in practice.
Consequences eventually arrived, just slowly. Singapore’s Monetary Authority banned both founders from regulated financial activity for nine years in September 2023, CryptoSlate reported. That same month, Su Zhu was arrested at Singapore’s Changi Airport attempting to leave the country and was sentenced to four months for contempt of the liquidation order; Kyle Davies received an identical sentence in absentia and was never taken into custody. In December 2023, BVI and Singapore courts froze up to $1.14 billion in assets belonging to Zhu, Davies, and Davies’ wife, Kelly Chen, who had separately filed her own $66 million creditor claim against the very estate her husband helped deplete. Zhu, for his part, filed a $5 million claim against his own former fund. None of that is a joke I’m making up for effect. It’s in the creditor filings, as reported by Bloomberg Law.
Where the money actually stands
As of Teneo’s December 2023 report to creditors, summarized by The Block, the liquidators estimated a 45.74% recovery rate against roughly $2.7 billion in expected recognized claims, out of more than $3 billion in claims filed overall. In March 2024, the BVI court approved an initial $100 million interim distribution to creditors, per Lexology’s coverage of the order, with a portion of the estate’s value locked in illiquid tokens on multi year vesting schedules stretching into 2026. I haven’t been able to confirm a more recent total distribution figure than that March 2024 order, and Teneo’s periodic creditor reports aren’t all public, so treat 46% as the last verified estimate rather than a current one.
Where I’d be wrong
The claim this piece rests on is that the GBTC and Genesis credit structure, not the Luna loss, was the actual load bearing failure, and that Luna simply removed the fund’s last remaining cushion at the exact moment it had none left to spare. That’s a reading of the sequence in the liquidators’ own affidavits, not a certainty. If it turns out 3AC’s Luna position was leveraged well beyond the $200 million loss figure reported, through derivatives or additional borrowed exposure not yet public, then Luna itself becomes the primary cause rather than the final push, and this piece has the emphasis backward. Corrections and sourcing standards for this piece follow How This Research Works.
The 3AC failure is also a worked example of two broader frameworks: map contagion through collateral, credit and shared service providers, then separate the asset from the legal and operational claim on it.
