BitMEX founder Arthur Hayes long article: The Three Arrows Capital I Know

Author: Arthur Hayes, Founder of BitMEX

Compiled by: GaryMa, Wu Says Blockchain

This current bear market is the third one I have experienced. Although it sometimes feels like a replay, each time brings new insights. Everyone's takeaways are different, but if you tend to listen to mainstream financial media, I can assure you that you have been fed the wrong narrative. Mainstream media serves traditional finance, and they will seize every opportunity to mock cryptocurrency as an economic/social experiment, arrogantly claiming that cryptocurrencies are worthless. Of course, as a staunch believer in the cryptocurrency industry, I will attempt to correct their malicious attacks.

In this article, I will use the legendary story of Three Arrows Capital (3AC) as a lens through which we can better understand the real insights we should gain from the current bear market. Please note that while I personally know Su Zhu and Kyle Davies (the heads of 3AC), I have no knowledge of what happened beyond what has been publicly reported. I intend to leverage my understanding of cryptocurrency, financial services, and common sense, starting from the implosion of TerraUSD and Luna, to identify what I believe led to their collapse.

The collapse of 3AC itself is not particularly noteworthy. A hedge fund that previously executed a boring but stable arbitrage strategy decided to use leverage to accelerate returns, and they paid the price. Using borrowed funds to play UST arbitrage was a death sentence for them.

However, what made the default of 3AC so impactful was that it blew a shark-sized hole in many of the largest centralized cryptocurrency lending businesses. Due to the losses from 3AC's loans, many of these lending businesses have locked up customer funds and are essentially in a state of insolvency. Credit has been withdrawn from the cryptocurrency ecosystem, causing a widespread collapse of Bitcoin, Ethereum, and the entire altcoin market, with no coin escaping unscathed.

But what many media outlets fail to mention is that both centralized and decentralized lending companies/platforms had exposure to 3AC, yet only the centralized lending companies collectively collapsed, while decentralized lending companies liquidated collateral without any operational issues. Using the story of 3AC as a canvas, let me paint you a picture of why Bitcoin and Ethereum stand the test of time and what this means for the future of cryptocurrency.

Hong Kong

Before we delve deeper, let’s quickly review how Su Zhu and Kyle Davies, the heads of 3AC, rose to prominence.

They graduated from university in 2008, the same year as I did, and at some point thereafter, they came to the Asia-Pacific region as employees of traditional financial banks/market makers.

The investment banking scene in Hong Kong, Singapore, and Tokyo is very tight-knit. Although I did not directly meet them until many years later (when we all entered cryptocurrency), we had overlapping social circles. The first time I met Kyle at a restaurant in Singapore, I swore I had seen him at a gathering in Hong Kong before.

I never had any trades with Kyle; he worked at Credit Suisse for a while, but I did trade with Su when he was a market maker at FlowTraders. As the chief market maker for Deutsche Bank's Asia-Pacific ETF business, I would quote buy/sell prices for large ETF product suites on the Hong Kong and Singapore stock exchanges. I often made mistakes and frequently lost money to them. Su was one of the professionals at FlowTraders who kept me on my toes every day. After working at FlowTraders, Su spent some time at Deutsche Bank.

The point of this story is that Su Zhu and Kyle Davies are arbitrageurs. Throughout their banking careers, they were trained to profit from small price discrepancies. Doing this over and over again, the money would grow. They brought this method and mindset to the firm they founded, which started by arbitraging the very inefficient over-the-counter non-deliverable forward (NDF) market.

Now, let’s continue with my days at Citibank.

When I was the chief ETF trader at Citibank, I also got involved in stock index forward trading, including NDFs. Our desk traded stock index forwards for major indices in Hong Kong, Taiwan, India, and South Korea. I was also responsible for the entire China A-share ETF trading book and traded a large volume of equity-linked derivatives.

NDFs are currency forward contracts that do not settle the principal on the maturity date but instead deliver the difference between the market price and the execution price directly.

Taking the USD/KRW (Korean Won) exchange rate as an example. South Korea does not allow the Korean Won to be freely convertible. Suppose I want to sell Korean Won for US dollars in the next 30 days, and I want to lock in the exchange rate today. I would trade on the right-hand side (RHS) of the USD-KRW NDF. If I enter the 30-day NDF at an exchange rate of 1,000 KRW, and the spot rate settles at 1,200 KRW, I would earn [ 200 KRW * nominal USD rate / USDKRW spot rate ]. If my trade size is $1,000,000, that would equate to about $166,666 in profit.

Understanding why NDFs are a necessary part of stock index forward trading is not crucial to this story, so I won’t delve into those bloody details.

The OTC NDF market is enormous, and all major investment banks have dedicated trading desks for these derivatives. As early as 2010, trades were conducted via Bloomberg chat. This meant that while traders had a rough idea of how a particular NDF should be priced, they could not see a consolidated market like they could for spot currencies. Therefore, arbitraging between different NDF trading desks was very easy.

Every day, I would pick a currency and try to arbitrage between two trading desks. Suppose I wanted a nominal 20,000,000 USD 30-day Korean Won NDF. Bank A would quote me 1,000/1,001; Bank B would quote me 1,002/1,003. Do you see the trade?

I buy at 1,001 from Bank A and sell at 1,002 from Bank B, and within 30 days, I would earn 1 KRW per dollar in profit. On a nominal $20 million, assuming the spot rate is 1,000 KRW, the total profit would be $20,000. This is not much for a bank trading desk, but if you do it often enough, it adds up to millions of dollars in risk-free returns each year.

As bankers, Su Zhu and Kyle Davies saw this inefficient market and started a fund with their own money, whose trading strategy was centered on profiting from these mispriced assets. Notably, unless you are a trader at a bank or a very large hedge fund, it is nearly impossible to trade in the NDF market. You must have an ISDA, and the senior management of the bank you are trading with must allow you to trade with them. Yet somehow, 3AC managed to open accounts at several investment banks and arbitraged among them in the Asian NDF market. When Su Zhu and Kyle Davies told me how they got started, I was impressed by their rapid entry into this lucrative market.

This is how 3AC made money over the years. Then, at some point, they discovered cryptocurrency and began trading as cash arbitrageurs. This is the bread and butter of any cryptocurrency arbitrage fund. Historically, the arbitrage space for short positions in perpetual contracts has been larger than for long positions, meaning that if you buy BTC and short the same amount of BTC/USD perpetual contracts, you will earn the basis differential over time.

These currency arbitrage and funding trades are very profitable, but they are also capital-intensive. You must post margin to trade each NDF derivative, and you do not increase your credit line by having the same and opposite positions at another bank. For cryptocurrency perpetual contract funding rate trading, you cannot use any leverage. This means that the growth of your hedge fund's AUM is slow and predictable.

The only way to improve hedge fund performance is to become a directional trader and/or use borrowed funds. 3AC did both.

UST/Luna

Under the temptation of Anchor's 20% yield, the critical thinking of those so-called institutional fund managers was thrown out the window, and they rushed to deposit funds for interest arbitrage.

Let’s return to 3AC.

I previously rambled on about their experiences in Hong Kong because I wanted you to understand how these individuals initially grew their fund. Arbitrage trades like UST are hard to ignore, especially for someone like Su Zhu and Kyle Davies, who have always been staunch advocates of DeFi and correctly bet on the price growth of several other protocols, making a lot of money. We know that 3AC held a large amount of LUNA, but what we do not know is the depth of their UST arbitrage trades. It’s worth noting that 3AC invested $200 million in LFG's February funding, and that investment is now nearly worthless.

The UST arbitrage trade is quite simple:

1. Borrow USD at an annual interest rate below 20%
2. Convert USD to UST
3. Deposit UST into Anchor to earn 20% interest
4. Annualized unrealized profit = 20% - borrowing cost

Using your own capital as a hedge fund severely limits profit potential. As a "cosmic master," when faced with such a lucrative arbitrage trade, one would leverage borrowed money.

Imagine you can borrow $1 billion at a 10% annual interest rate, using your existing altcoin portfolio (even if unsecured) as collateral. Then, you convert the $1 billion into $1 billion UST and earn 20%. Every year you would make $100 million, and managing this position requires no effort. Similarly, I do not know how large 3AC's UST position was, but if we believe they once managed nearly $18 billion in assets earlier this year, it is very reasonable to assume they borrowed billions to deploy this arbitrage trade.

Your ability to borrow money depends on your collateral and creditworthiness. Our goal is to borrow money with as little collateral as possible. The less collateral you are required to provide, the more trust the lender has in you.

3AC cleverly crafted an invincible aura by correctly predicting the market in the past and actively sharing their trading situations on social media. At its peak, 3AC was one of the largest cryptocurrency pools in the world. This stellar trading reputation, combined with the scale of assets they controlled, allowed 3AC to borrow under very generous terms. For example, Voyager, a publicly traded company, lent 3AC hundreds of millions of dollars without any collateral. Specifically, it was 15,250 BTC and $350 million.

Game Over

This leads me to my thoughts on how 3AC began to slide toward bankruptcy (pure speculation).

As mentioned above, I suspect 3AC not only used its own capital but also borrowed USD from individuals and funds to conduct UST arbitrage trades. For loans that required collateral, 3AC used Bitcoin, Ethereum, and, most importantly, various illiquid and volatile altcoins as collateral. On paper, 3AC's cryptocurrency asset portfolio looked very impressive. However, when UST decoupled and the entire $40 billion Terra ecosystem collapsed within a week, what was once a profitable arbitrage trade quickly turned into a rotten nightmare, and the collapse of 3AC became a matter of when, not if.

When the market declines, it does so in a correlated manner. This simple fact spelled doom for 3AC, whose inflated valuation largely stemmed from the market value of many illiquid altcoins they held. As the market began to decline, the liquidity of these altcoins evaporated, leaving only bids on the order book. As 3AC continued to sell into rapidly depleting demand, prices fell in a non-linear fashion, ultimately preventing them from recovering enough USD to repay their loans.

3AC suddenly owed a massive amount of USD that it could not repay, as the remaining assets it could liquidate to pay off those loans had dropped by 50% to 75%. If the broader cryptocurrency market had not declined, 3AC might have been able to recover from the capital losses tied to UST and LUNA. Instead, the market sensed a wounded lion and set about driving it out.

Then the questions began.

Who lent money to 3AC?

How much money was lent to 3AC, and what collateral was there?

What was the collateral?

Did 3AC pledge the same collateral multiple times?

Over time, various large centralized cryptocurrency lending institutions came under pressure. Some major companies declared bankruptcy and sought help from a few cryptocurrency OGs with enough capital to restore their solvency.

But you might wonder how the default of one fund could evolve into the near-collapse of the entire industry?

Good question. Let’s take a closer look at how 3AC brought down many of the largest and most prominent centralized cryptocurrency lending participants.

Centralized Cryptocurrency Lenders

First, let’s familiarize ourselves with the major players in the centralized lending space. Some of the large companies currently feeling the pressure from 3AC's demise include Voyager, Celsius, BlockFi, and Babel Finance.

Just a reminder: I have no non-public information about these companies. I only have reasoning abilities and public statements as a basis.

The lending business is very simple: lenders accept deposits and pay interest to depositors in exchange for the ability to lend out their funds. Then, lenders lend out the funds at a rate higher than what they pay to depositors to make a profit.

The simplicity of this model comes with some pitfalls, one of which is that lending businesses will inevitably come under pressure during any form of financial crisis. Unless you work at the company, depositors have no way of knowing whether the company is handling their capital prudently. Therefore, as soon as there are any signs of stress, you will quickly flee, and if lenders are taking on any term risk, they will soon go bankrupt.

With that in mind, let me describe the safest way to operate a lending business.

Suppose you have three clients: Mark Karpeles (former CEO of Mt. Gox), Do Kwon, and Su Zhu. Each has 1 ETH and wants a return on their assets. Each has a time preference (Mark = one month, Do = three months, Su = one year).

We will call the lending institution the Cryptocurrency Long-Term Capital Management (LTCM).

LTCM can be a safe lending institution or a risky one.

If LTCM is conservative, it will lend out matching terms with each person's time preference. Thus, LTCM lends 1 ETH for one month, 1 ETH for three months, and 1 ETH for one year. When Mark, Do, or Su asks for their money back, that money is lent out and cannot be returned to them, but there is no risk.

If LTCM is aggressive, it will lend out for terms longer than each person's time preference. To illustrate this absurd extreme, LTCM could lend out 3 ETH for ten years. Clearly, if any depositor requests their money back within their preferred time frame, there will be no funds available. This is why lenders go bankrupt. But LTCM can gamble that it will be able to convince each depositor to roll over their deposits multiple times, thus preventing LTCM from ever going bankrupt.

This analysis only involves the term of LTCM's loan book. The second aspect of the lending business is the quality of the borrowers. As a company, the value of lenders primarily comes from their ability to accurately assess the credit risk of borrowers and require appropriate collateral or guarantees based on those risk assessments before lending.

In the cryptocurrency industry, there are three sources of demand for credit:

1. Spread and basis trading. Borrowers do not provide collateral because they want as much cash as possible to use these strategies most effectively.
2. Margin loans for speculation. These are prepared for directional long and short traders. Typically, they will provide some form of collateral. However, failing to understand the volatility and liquidity changes of the underlying collateral can lead to losses for lenders.
3. Mining asset-backed loans. Miners either pledge hardware (like ASIC Bitcoin miners) or pledge cryptocurrencies and receive fiat and/or stablecoins. The risk for these borrowers is minimal because they have strong cryptocurrency cash flows supporting their loans. However, if you recover ASIC machines from a defaulting miner, you must have a site condition to plug in the machines and start mining to make full use of them.

The major missing use case in cryptocurrency credit is commercial loans. This space is so new and the risks so high that lending to cryptocurrency companies makes little sense. Cryptocurrency companies, similar to the underlying tokens themselves, should be viewed as call options. In the event of default, no matter where you are in the capital structure, you will be wiped out. Given this, it is better to just hold equity because at least you can participate in the upside.

Due to the inherent volatility of the cryptocurrency market and the so-called profitability of various yield farming, arbitrage, and basis trading, borrowers have been willing to pay extremely high interest rates. This means that lenders can offer extremely high rates to retail depositors while still maintaining a positive net interest margin (NIM).

I believe that the initial loans made by mainstream cryptocurrency lending institutions were very reasonable, but they grew too quickly afterward. Deposit rates were very attractive compared to the offers from TradFi banks and sovereign governments for fiat deposits and bonds. This was a result of central banks printing money indiscriminately and zero interest rate policies. Retail investors, eager for yield, could not resist these enormous returns on cryptocurrencies and fiat stablecoins.

The influx of billions of dollars into these few companies exceeded the supply of responsible borrowers. They had to deploy this capital because they had to pay to acquire these funds (remember, they had to pay interest to everyone). The pressure to lend out funds forced companies to lower borrowing standards, leading them to engage in arbitrage trades very similar to those of 3AC.

For example, BlockFi was once one of the largest holders of the publicly traded Bitcoin tracking fund GBTC. BlockFi accepted BTC deposits, created GBTC shares, and then sold GBTC shares in the market for a premium. That was the plan, anyway, but it took six months to create GBTC shares. During those six months (from late 2021 to early 2022), the premium on GBTC turned into a discount, and by the time they exited the arbitrage, they had incurred losses.

The above is a chart of GBTC's premium or discount to its net asset value (NAV). Given the six-month preparation time and humanity's infinite speculation on short-term trends, I bet many companies borrowed BTC to create GBTC, hoping to make a 40% profit six months later. As you can see, this clearly did not happen, as GBTC has been trading at a discount since the second quarter of 2021.

3AC's ability to engage in such large-scale arbitrage and directional trading was largely due to its pool of capital that had to be invested. In my view, there is no other explanation for why publicly traded companies like Voyager would lend them hundreds of millions of unsecured dollars and Bitcoin. 3AC claimed they would pay high interest, and Voyager believed them because, in the eyes of Voyager's leadership, no other company could absorb the capital that needed to be deployed like 3AC. As Chuck Prince said in response to Citibank's involvement in issuing subprime mortgages, "When the music stops, in terms of liquidity, things will get complicated. But as long as the music is playing, you have to get up and dance. We are still dancing."

Unfortunately, the lenders were all on the same side of the trade. They had the same borrowers. They held the same collateral. And the borrowers were all losing money on the same arbitrage trades. The only difference between 3AC and these lenders was their ability to market themselves. As we have seen, they all failed to conduct proper risk management to withstand this particular storm.

Worse still, these private companies do not have to regularly disclose the health of their loan books, nor do they have to reserve for expected bad loans like publicly traded banks. We can only guess the potential size of the losses these lenders face without any basis. Therefore, everyone tried to withdraw their funds at the first sign of stress. And that is exactly what happened, as the market wanted to know how big of a hole 3AC had blown in these lenders.

This was not just a retail exit. These lenders were also trading the same risks, meaning they had each other's risks and did not trust one another. This complete loss of confidence in centralized opaque lending institutions is why they almost simultaneously went extinct.

Halftime

To quickly recap: 3AC blew up because they took a boring, stable, and predictable arbitrage strategy and added astonishing leverage, ultimately succumbing to the market after the UST arbitrage trade collapsed.

Due to their sharp investment acumen and large asset pool, 3AC was allowed to borrow from eager lending companies looking to reallocate retail deposits into high-yield cryptocurrency credit instruments with little to no collateral. These lending companies, such as BlockFi, Babel Finance, Voyager, and Celsius, abandoned prudent risk management policies to secure as many loans as quickly as possible. As a result, when 3AC failed to meet margin requirements, these lenders were left with massive holes on their balance sheets. Unfortunately, these once-lauded "fintech" startups valued at over $1 billion are now heading toward bankruptcy and restructuring.

How these lending institutions became unable to repay their debts is not new or innovative. As long as there is centralized lending, there will be epic failures, like the ones these cryptocurrency lending institutions are experiencing in this case. The bankruptcies of 3AC and these lending institutions have nothing to do with the technology of cryptocurrency; it is merely a typical failure of a financial institution with poor risk management, allowing us to delve into how actual cryptocurrencies and DeFi applications handle these market pressures.

UST Worked Fine

The Terra ecosystem is centered around DeFi and operates in conjunction with the algorithmic stablecoin UST. The model of algorithmic stablecoins is indeed relatively Ponzi-like, but UST still executed according to its design model. Although UST ultimately collapsed inevitably, this was a potential outcome we foresaw, and it is not UST's fault.

DeFi Lending Protocols

Other entities that lent to 3AC are DeFi lending protocols, primarily Compound (COMP), Aave (AAVE), and MakerDAO (MKR).

For these protocols, 3AC is just a few addresses; there is no so-called extra credit. All borrowers must strictly adhere to the rules of the protocol, so you see, these protocols do not pause withdrawals and can continue to issue loans; everything proceeds normally.

CeFi vs DeFi

When you remove trust from the equation and rely purely on transparent lending standards executed by fair computer code, you get a better outcome. This is the lesson to be learned. Do not let the media blame the failures of these centralized companies on the shortcomings of blockchain.

Now the question is, due to the contraction of centralized cryptocurrency credit, with hundreds of thousands or even millions being taken down, who can still borrow funds? This is the problem that these DeFi lending protocols will strive to solve as we progress through this bear market.

At some point, the price-to-book value ratio of these DeFi lending protocols will become attractive again. Currently, given the recent outlook for loan demand, their trading prices remain too high. With the exit of speculators, large trading institutions, and centralized lenders, and without significant organic loan demand, I hope the market gives me an opportunity (maybe I am just too greedy) to own COMP, AAVE, and MKR at lower prices before we exit this bear market.

To Rescue or Not

Centralized cryptocurrency lenders are in urgent need of liquidity, but should capital be provided to rescue them?

For cryptocurrency hedge funds like 3AC, their intrinsic value entirely depends on their portfolio managers (PMs) wisely taking risks and consistently making money over the long term. If a group of PMs blows up due to improper use of leverage, what value is left in the funds they control? I suspect 3AC made frantic calls to CZ and SBF, but would they extend a helping hand? It is certain that the situation is chaotic, and I doubt 3AC and funds in similar situations will find a savior.

As for cryptocurrency lending companies, their only value lies in their client lists and a small portion of good loans. If these clients can be retained and more crypto financial products sold, it may make sense to cheaply acquire paralyzed centralized crypto lending companies and take on their debts. Their loan books may contain some properly underwritten loans that can be purchased at attractive prices. The problem is that the longer these lending institutions go without assistance, the less likely they are to reopen and allow their retail depositors to withdraw. The time pressure for due diligence on these opaque entities makes it increasingly difficult to accurately assess whether cryptocurrency tycoons will abandon their capital and save these sinners.

Of course, the Federal Reserve or other central banks could rescue these hedge funds and companies, but these entities trade in cryptocurrency. They are not part of the "too big to fail" financial institution club and will die a very undignified death. However, let us not shed too many tears, as we learn through these trials and tribulations that the promise of a new decentralized financial system has withstood yet another test.