Finance, but Decentralized
A DeFi investment thesis, part one
In this series of blogs, I set out my long-term investment thesis for decentralized finance (DeFi) — finance on the blockchain. I end with a hopeful call to arms: To truly recreate (all of) “finance, but decentralized”.
It starts out as a bit of a sad tale, however. Which might surprise many, as on the face of it DeFi has had a blockbuster year and a half, since “DeFi summer” 2020.
A DeFi Cambrian explosion emerged from the primordial soup of Ethereum smart contracts in the ecosystem drove a 4x increase in capital (measured by “total value locked”, or TVL) in the last year alone:
More than 70 chains have some amount of DeFi occurring on them. A few of the top 10, including Terra, Avalanche and Fantom, are now hosting tens of billions of DeFi TVL, up from next to nothing a year ago.
There are many success stories. Protocols founded in the early days thrive even without spending on user acquisition and retention, such as Uniswap v3, MakerDAO, Aave, Compound, and wrappers and bridges such as renBTC.
Yet the gains in TVL and users as well as the increasing number of developers building on an increasing number of layer 1 blockchains mask a potentially fatal problem in DeFi.
Devs and users alike have been spoiled by (1) the ease and speed in which crypto traders have adopted the on-chain infrastructure needed to replace rickety and dangerous centralized exchanges of the past, as well as (2) the rise of governance tokens to incentivize protocol usage and “pool 2” liquidity, in the face of extremely poor capital efficiency. Token incentives did well in the DeFi bull from DeFi summer until spring 2021. Not so well since.
The rise of DeFi on other L1s and indeed same chain forks, as well as the launch of new yet uneconomic use cases, has generally resulted in a never-ending battle to secure what almost always turns out to be temporary users. But the current DeFi ecosystem, built to support trading and speculation, cannot easily scale beyond its current uses. Growth is needed, and that won’t come from the same old use cases. And inflationary tokens do nothing to remedy this problem.
What token incentives do accomplish is to create an elite class of rentiers who use insider knowledge and crypto-native abilities to earn “passive” income on their capital by being early to the latest incentive bazaar. We do not appear to be shedding oursleves from a reliance on these rewards. The new new thing as of March 2022:
We have ended up with a hyper-capitalist ecosystem, where nobody is incentivized for the long run. Not VCs, who can often begin cashing out day 1. Not teams, who may not have the fortitude and incentives to build lasting companies. And certainly not users, looking for the latest and highest yields (APY) or whitelists or airdrops to get tokens before the masses arrive. Then dumping as soon as they can.
Rather than focus on highly inflationary token incentives and/or disappointing iterations on the same original themes of borrowing/lending, DEXs, interoperability and farming, I call on developers and VCs to pivot away from speculative financial plumbing and inflationary token rewards, and towards a DeFi that reinvents (all of) finance, for all.
In this blog, I trace DeFi’s development, culminating in the bull market in so-called DeFi 1.0 tokens that ended in the spring of 2021. I show that DeFi replicated much-needed basic tools for on-chain speculation, and that several excellent use cases were the result. But token rewards used to compensate for capital inefficiency resulted in extremely poor token performance once the token rewards faded, and users, VCs and teams sold their cheap tokens. Everybody moved on, quickly.
In part two I explain how the current state of DeFi was a direct result of earlier disappointments, and a failure to own up to the mistakes of the past. DeFi remains obsessed with chasing extraordinary yet temporary APYs, or the newest pre-launch. These early farmers and pump-and-dumpers have gotten very rich. Is this what we want DeFi to be?
But I do not come to bury DeFi, I come to praise it. I believe that DeFi can be the source of new hope: for the unbanked, for the retail crypto or traditional investor, and for those who don’t want to wait three-five days for their bank transfer to go through. In the final blog I explain why I remain so bullish, and how we can build and invest for the future.
Part 1: The Dawn of DeFi
The perils of early crypto trading
Though now in some circles thought of as largely irrelevant boomer projects, the earliest DeFi pioneers were really battling a dark empire. Not that of traditional finance, however, but of crypto exchanges.
That is, the dawn of DeFi that launched a thousand protocols had one goal in mind at first: Make speculation in cryptoassets safer than on the unregulated centralized exchanges of the day. But why?
Bitcoin was the original crypto asset. And its rise is without precedent
Yet in the mid 2010s, crypto exchanges operated in the financial market version of the Wild West, or at least the 1920s. At one point, Mt Gox, previously a platform for trading Magic The Gathering collectibles, accounted for 70% of all Bitcoin trading. I think we all know what happened next. But Mt Gox was not an anomaly. The list of offenders is endless: Quadriga CX, Coincheck, Bitfinex are a few of the bigger headlines.
Traditional safeguards were largely absent in these pioneering exchanges. There was often little governance, proper segregation of funds, or checks and balances. One person could control all of the assets. “Rug pulls”, such as that at Canadian exchange Quadriga CX ($200 million) and Turkish counterpart Thodex ($2 billion), were often the result. Poor opsec resulted in many high-profile hacks, often measured in the hundreds of millions of dollars. The Bitfinex funds are now worth almost $4.5 billion!
Of course, without legal separation between customer and exchange orders, front running and manipulation were rampant. Never mind that there were no disclosure requirements and no suitability checks.
DeFi “1.0”, an abridged history
Soon after Ethereum launched smart contracts, devs were in motion. And by the seventh day, DeFi had recreated all the tools needed for speculation. The first need was for borrowing/lending platforms that could recreate margin lending and facilitate short selling
Margin lending — Beginning in March 2019, MakerDAO allowed ETH holders to lever up by depositing their tokens into an overcollateralized smart contract called a CDP, backing the minting of a new USD-pegged stablecoin, now called DAI. DAI could then be sold to buy more ETH, or any other token. Instant leverage. Of course, if ETH fell enough, the borrower either had to pay down some DAI debt, or deposit more ETH. Otherwise they would be liquidated, as in any margin loan.
It is worth noting that using DAI for debt explicitly made the USD the unit of account for DeFi. Debt was denominated in USD, just like US stock market margin debt.
Shorting — MakerDAO’s stablecoin leverage was joined by platforms that matched borrowers and lenders, such as Aave and Compound. Overcollateralized borrowers of stablecoins could sell them to mimic margin lending as above, of course, but could also borrow riskier tokens to go short. Short selling increases the efficiency of markets by allowing traders to profit from price falls in assets that are considered overvalued. Another key component of markets was (re)created.
Exchanges — Of course it’s impossible to speculate without trading venues, but recreating them on-chain required overcoming one major hurdle: The traditional exchange’s central limit order book (CLOB) system for broadcasting and filling orders could not scale on Ethereum. There were some false starts before Bancor and Uniswap developed a solution. Markets would be made passively in what is known as an Automated Market Maker (AMM) by liquidity providers (LPs) using a deceptively simple formula.
Using the x * y = k model, prices would be set equal to the ratio of the two tokens on the exchange. On its own the system was far from ideal. But arbitrageurs who could trade on centralized exchanges and other decentralized exchanges would help keep prices in line.
Together, DEXs and borrowing/lending protocols make up 71% of DeFi TVL (see chart below), testament to their use cases. They fill a need. The biggest dominate the DeFi space: Five out of the current DeFi top 10– Curve, MakerDAO Aave, Uni and Compound– account for 35% of all capital deployed in DeFi. These are hardy, useful and battle-tested protocols that have survived crashes such as March 2020, when ETH and altcoins fell 50% or more in 24 hours.
There was but one problem, and it was a doozy.
Capital inefficiency and impermanent loss: The rise of the token
The first problem is that passive liquidity providers on AMMs can lose money as markets move: As one token rises in price, the LPs end up holding less of it. If trading fees do not compensate for these losses, the LPs end up with an “impermanent loss” that can often turn permanent.
The second issue is that DeFi 1.0 is extremely capital inefficient. AMMs need to hold enough capital to cover all possible price changes: The LPs make prices all the way down to near zero on either token (and of course near infinity on the other). Borrowing/lending protocols, with or without stablecoins, also require overcollateralization. MakerDAO’s maximum loan to value (LTV) is 66%.
First Synthetix and then Compound found a solution, rewarding users with their own protocol “governance” token.
Yield aggregators — As almost every protocol adopted a token, “harvesting” the token rewards became complicated and time-consuming. Iearn/Yearn was an early innovator, but now the space is full of so-called “yield aggregators” and bribing protocols (more on these in part 2) that maximize a user’s rewards. They have become very popular: The top 2 in this space account for about 10% of TVL.
Interoperability — The last original use case is the result of a bug rather than a feature of original DeFi. Native BTC cannot be used in DeFi, so we end up with wBTC, renBTC, and a host of others to take advantage of the trading-based financial infrastructure.
Bridging complicates this: there is a wormhole and native version of many tokens on exchanges such as Avalanche, for example. Additionally there are a wide variety of stablecoins, each with their own supporters and risks. As a result, a good deal of exchange and borrowing/lending volume revolves around the trading of USD-pegged tokens, such as USDT, USDC, DAI, UST, USDP, SUSD, BUSD, FRAX, FEI, MIM and alUSD. Increased bridging looks to make these problems even worse.
So we can now total all of the capital that is being used to facilitate trading, as follows:
There are some observations from this chart and the data behind it:
- DeFi is very concentrated: The top 7 protocols account for just under half of all TVL. The top 50 protocols account for 90%.
- DeFi is mostly explained by the four trading use cases above: Over 90% of the top 50 are related to facilitating trading.
- True TVL on DeFi is overstated: Overlap abounds, with, for example, Convex (#4) and Yearn (#14) sitting on Curve (#1).
- Ethereum is barely holding a 50% market share, down from near 100% a year ago. The rise of new chains means that the same infrastructure has to be replicated many times: each chain needs a DEX or three (including one for stables), a borrowing/lending protocol, and a yield aggregator. Avalanche, for example, has Aave and Benqi as top borrower/lenders, Curve, Platypus and Trader Joe as top AMMs, and Yield Yak to farm yields.
Now I won’t deny that within the broad categories above there is some real innovation occurring. And some of these earn real fees. Lido is one such protocol, providing liquid staking to those otherwise locked into ETH (until the Merge) or LUNA (21 days). I also would agree that there is a “long tail” of new DeFi that is innovating, but not yet in a position to dominate the top 50. Again, however, more than 90% of DeFi is in a DEX, lender/borrower, yield aggregator on top of the first two, or a bridging solution.
DeFi 1.0 after the hype
So much for the landscape. What about performance? Well it’s been very disappointing.
The DeFi Pulse index that consists mostly of the original use case protocols have flatlined, in dollars or in ETH.
It’s worth examining why prices have fallen, and there are many. Most have to do with the capital inefficiency of DeFi 1.0 — the base return for Curve’s largest pool is just 0.22% — and the solution to incentive activity using inflationary mostly-worthless governance tokens.
Very few tokens accrue value from protocol activities, even as they are minted heavily to incentivize otherwise capital inefficient behaviour. COMP, AAVE and UNI offer no cash flow to their tokenholders. Some, like Sushi and Curve, have some value accrual through fee distribution. A select few others have buyback-and-burn tokenomics. Worse still, some of the biggest growth areas of the future may not result in value accrual to the token. For example, Compound Treasury’s (permissioned markets with KYC/AML) success will not benefit COMP.
In bull markets, those that receive and then sell the tokens (farm and dump) don’t impact the market much. VCs and the team are also less inclined to sell their allocations as prices are rising. But in a more uncertain market, farmers, VCs and teams will be incentivized to dump, causing a price spiral that has many serious effects.
Vesper Finance is a poster child illustrating how farm and dump plus unlocks puts pressure on a token. And, worse, how users don’t stay once rewards reduce. TVL was highly incentivized for three months, and otherwise unsustainable. But even before the May correction and TVL started to drop, the price fell significantly.
Price collapses also affect the future of the protocol. Not only do token incentives lose their luster (and eventually run out), but teams become uninterested in working for tokens worth peanuts. It could even be the case that the team, partner and community tokens run out, and selling more tokens to new investors after a huge dip may not be feasible. Devs leave. Discords empty. Projects stop innovating, or even maintaining. Sushiswap and Barnbridge are two great examples of this phenomenon, but there are others.
I recently saw a secondary offering for a protocol’s token where the incentives had nearly ended and the team and VCs were unlocked. The product was almost there. PMF was a near certainty. But the devs had left and the remaining parties were mercenaries just looking for exit liquidity. Where does the protocol go from here?
One last issue is that, as a result of open source code and the rise of alternative layer 1s, we end up with many multiple versions of the same product, few of which gain much traction without inflationary incentives. How many fixed rate protocols does DeFi need, even cross-chain? How about DEXes?
There are 195 forks or close copies of Uniswap v2 that have more than $100,000 in TVL. Because the newest must reward their users in inflationary tokens, it becomes a player-versus-player market. Sure, there is (temporary) market share growth, but little of the innovation that is needed to really grow the pie. Worse still, aggressive customer acquisition obscures who the true long-term winners might be. As Nic Carter pointed out on 27 February,
It seems like we’re in a state of suspended reality where a lot of the liquidity has been enhanced due to the existence of the liquidity mining rewards. …We almost don’t know for now which are the self-sustaining protocols that have reached exit velocity.
We can’t yet identify the winners. But for the losers, will users stay after rewards run out? In general they do not.
Centralization is winning crypto
Meanwhile centralized solutions have smartened up since the earlier Wild West, and are raking it in, from Coinbase to Binance DEXs and OpenSea for NFTs. Many are regulated, insured and offer unparalleled UX. As a result they dominate:
The Gartner Hype Cycle and DeFi 1.0 and 2.0
So DeFi 1.0 clearly suffered from inflated expectations and then a steep decline, as predicted by the Gartner Hype Cycle. But when will we bounce onto the plateau of enlightenment?
My view, for many of these projects, is never. Why would you stay to try to work out a failing protocol when all you have to do is dump your tokens and start a new run towards inflated expectations. Enter “DeFi 2.0”. As DeFi 1.0 names like Balancer (BAL, below) and SUSHI flatlined in the spring of 2021, a new crop of DeFi protocols arose, such as OlympusDAO (OHM, below), Wonderland, Rari, Abracadabra and Convex. A narrative shift was all it took to breathe new life into the space, for users, devs and investors.
Some major improvements have resulted from this early grand DeFi experiment, with the successful replication of the financial infrastructure needed for trading crypto: (1) exchanges, (2) shorting and margin lending, and (3) interoperability allowing the use of the top token risks on any DeFi platform (e.g. wrapped BTC).
Yet capital efficiency remains low, and growth in DeFi 1.0, except where it has been resurrected with “2.0” economics, has been stagnating since the late spring of 2021.
DeFi 2.0 was supposed to solve all of this. No more boring capital-inefficient protocols. No more farming and dumping for liquidity providers. Enter “permissionless pools” with higher TVLs (more capital efficiency, but more risk), “protocol owned liquidity” and “monetary premium”.
But in fact, there is not much new in 2.0. It was more of a much-needed narrative shift, diverting attention away from token and business model failures in early DeFi.
How are these faring? I’ll explain next week.