The Problems With DeFi

Summary An overview of some of the potential opportunities and shortfalls of "Decentralized Finance" or "DeFi" for short. Analysis on various centralization risks and other risk associated with DeFi. Examples of misaligned advertising and problematic details around the DeFi industry. In the wake of the FTX/Alameda (FTT-USD) collapse, which came after prior collapses of Luna (LUNC-USD), Celsius (CEL-USD), Voyager (VYGVQ), and Three Arrows Capital, many crypto industry analysts have pointed towards DeFi (“Decentralized Finance”) as a potential solution. This article dissects some of the nuances of that view, grants that certain opportunities exist, but offers a rather critical assessment of the current state of the industry overall. CeFi: The Problem Of Opaque Leverage The purported benefit of DeFi is that the technology can supposedly decentralize and add transparency to various trading and leveraging services. So, it makes sense to start this article by analyzing crypto CeFi (“centralized finance”) before getting into crypto DeFi (“decentralized finance”). In the crypto asset industry, the two big types of centralized companies are exchanges and deposit/lenders. The term “CeFi” mainly refers to the latter, but more broadly can include both, especially since the two business models can be intertwined. Exchanges allow traders to swap around various crypto assets, and generally offer leverage to traders as well. Deposit/lenders (aka “liquidity providers”) allow people to deposit crypto assets and collect a yield, allow people to deposit crypto assets as collateral to borrow stablecoins or fiat currency, and allow some institutions to borrow money in an uncollateralized way. There are other companies as well, such as pure custodians, technology developers, and so forth, but those are the two types of companies that are relevant for a comparison to DeFi. Large portions of the legacy financial system are opaque. It’s difficult to determine how much liabilities any given entity has, unless it’s publicly traded. Even then, it’s still possible to commit fraud or to obfuscate the details with certain accounting tricks to some degree. This tendency carried over into the CeFi crypto industry, but on steroids. Various trading firms and funds frequently take on leverage to speculate on crypto assets, and especially altcoins. Since very few of these companies are publicly-traded, almost all of it is rather opaque, even though it involves many billions of dollars. Throughout 2021 and 2022, this industry began to run into problems for the cycle. The early problems began with the premium-over-NAV of the Grayscale Bitcoin Trust ( BTC )" target="_blank">OTC:GBTC ) flipping to a discount-under-NAV. In other words, there was a time when the market price of the fund was a dollar forty for every dollar worth of Bitcoin in the fund, and now the market price is below sixty cents for every dollar worth of bitcoin in the fund. YCharts Many trading firms had leveraged exposure to that vehicle, so this was a bigger issue for them than one might realize. For years, GBTC was one of the few ways to get bitcoin exposure in the form of a security, which a number of regulated entities wanted. Therefore, it traded at a premium over net asset value “NAV”, since it offered something that was otherwise hard to get in that type of securitized package, including within a brokerage account. As more funds and securities came to market, offering bitcoin price exposure for these regulated types of entities, GBTC became less unique, and its premium went away. Most close-end funds trade at a mild discount to NAV, and GBTC became no different. The vast majority of my bitcoin exposure is directly in bitcoin itself. However, in my brokerage portfolios, I have used a small GBTC position to replicate bitcoin price exposure for lack of better alternatives. I have rebalanced its exposure from time to time, especially when its premium became excessive, like in this case with my No Limits portfolio: YCharts In addition to just being a way for a portfolio to get some bitcoin price exposure, GBTC also gave accredited investors an arbitrage trade for many years. They 1) could short bitcoin, and 2) send funds to GBTC to create new units of GBTC at NAV. They had to hold this position for a six-month lockup, and then could sell their GBTC shares for a premium over NAV and close their bitcoin shorts, thus pocketing that premium over NAV as a profit without exposing themselves to the volatile bitcoin price exposure that underlies both bitcoin itself and GBTC. Then, they could repeat this trade again and again. This worked for a long time, until it became easier to access bitcoin exposure in various other regulated vehicles. At that point, there was no longer reason for the trust to have a premium over NAV. This caught a number of trading firms off guard, since they were stuck with a massive bitcoin short position and a massive GBTC long position that, rather than trading at a premium as it used to, was now at a discount. In other words, they experienced a major loss from what they viewed as a low-risk trade, and that they therefore had on in size and with leverage. So, that was strike one. I covered this premium-to-discount flip back in one of my premium reports at the time, although I mainly focused on it as a cessation of a portion of the buying pressure on bitcoin: Up against the structural bull market of the post-halving adoption cycle, there are two fundamental headwinds against bitcoin at the moment. The first is potentially bond yields as described previously. If they keep rising, and the dollar index remains firm, and liquidity gets more tight, that should continue to pressure highly-valued growth stocks, and bitcoin may temporarily be lumped into that category by investors as well, especially in a broad risk-off environment. The second is that the Grayscale Bitcoin Trust no longer has a positive premium over its net asset value or “NAV”. In fact, it’s trading at a slight discount to NAV. Canada now has a solid bitcoin ETF, firms such as NYDIG and SkyBridge have come to market over time and provided competition for institutional and accredited investor allocations into bitcoin. So, there is less of a reason for GBTC to have a premium. Most existing closed end funds for other asset classes trade at a discount to NAV, and GBTC has at least temporarily joined the discount club. This discount to NAV is both good and bad. It’s good because it makes GBTC a better vehicle for holding bitcoin than it used to be, since you’re getting it at around net asset value rather than at a premium. It’s bad because it removes one of the sources of bitcoin demand. GBTC was the biggest buyer of bitcoin in 2020. Some of this was for long exposure, and another big chunk was for market-neutral arbitrage. For that second part, accredited investors and institutions could buy into GBTC at NAV with a six-month lock-up period, and then sell six months later at the market price, and that market price generally had a premium over NAV. Simultaneously with their purchase at NAV, they could short bitcoin elsewhere, and thus make a non-directional investment that simply lets them extract the premium over NAV every six months with no exposure to bitcoin’s price. This process made a lot of money for folks, and permanently sucked up a lot of bitcoin. GBTC basically converted liquid bitcoin to illiquid bitcoin, locked away in cold storage. If the GBTC premium remains minimal or negative, that trade is done, and thus that specific source of demand and liquid-to-illiquid conversion process is done. -Lyn Alden, February 28th 2021 Stock Waves report Strike two was when Terra/Luna blew up, and that was basically a ponzi scheme from the start. The people involved lured retail investors in with artificial unsustainable 20% yields. I warned about Terra/Luna during April and May 2022 near its highs, and then wrote a post-mortem on it called “ Digital Alchemy ” in May 2022. One of the themes of that post-mortem is that when liquidity and general business conditions are improving, crypto ponzis tend to be built, and when liquidity and general business conditions are deteriorating, crypto ponzis tend to be revealed and break apart: Trading Economics, annotated by Lyn Alden A number of leveraged trading firms had exposure to Luna, which made the subsequent blow-up in the crypto industry even more severe than I would have guessed at the time. The biggest surprise to many was the fall of Three Arrows Capital or “3AC” for short, which had been impaired by the GBTC discount and then was hit again by the collapse of Luna, which the 3AC team had been particularly bullish on. Three Arrows Capital was a large crypto trading firm that had been around for ten years, and they were using a lot more leverage than most people realized, since it was opaque and they were quietly borrowing from many different sources that were not necessarily communicating with each other. Multiple lenders including Voyager and Genesis had huge unsecured loans out to 3AC. BlockFi had a huge loan to 3AC as well, but it was collateralized, which reduced the total impact to themselves. Celsius and a couple dozen other firms also had loans out to 3AC. So, the resulting implosion of 3AC brought down large swaths of the crypto lending industry, with just a handful still standing afterward. The overall risk management in the space was bad, although some were better than others. Many of the lenders were lending to just a handful of the same massive borrowers, like 3AC. Overall, the biggest losses in the CeFi industry came from either 1) making uncollateralized loans to leveraged entities with opaque balance sheets or 2) custodying assets with leveraged entities with opaque balance sheets. And the problem with opaque balance sheets, of course, is that it’s hard to know for sure how much they are leveraged, especially when there is allegedly outright fraud involved in some cases. What DeFi Aims To Accomplish In contrast to CeFi, trading and leveraging occurs on various blockchains as well, in what is generally termed “DeFi”. As previously mentioned, this stands for “decentralized finance”. The foundation for this activity consists of Layer 1 smart contract blockchains, such as Ethereum (ETH-USD) , BNB Chain (BNB-USD), Positron (TRON-USD), Avalanche (AVAX-USD), Solana (SOL-USD), and others. These systems allow a computing environment to be embedded into a blockchain. This effectively becomes a federated computing environment, meaning a computing environment that is spread across multiple machines, and thus able to be observed and confirmed by multiple different entities. This computing environment can also be built on layers/sidechains on top of other blockchains like the Bitcoin network (e.g. Liquid or RSK), but most of the activity thus far has been on these Layer 1 smart contract blockchains. After the computing layer, the next layer up consists of two main parts: decentralized exchanges and decentralized deposit/lenders. The decentralized exchanges allow traders to swap in and out of various crypto assets, similarly to how they would on a centralized exchange. The decentralized deposit/lenders (the most common term being “liquidity providers”) allow users to either 1) deposit crypto assets into the protocol to receive a yield or 2) deposit crypto assets as collateral to borrow another type of crypto asset, with the most common asset on the deposit and the borrowing side being stablecoins. Whether or not these are actually as decentralized as they claim to be will be discussed in a later section of this article. After that, there are other types of DeFi assets. Most of these are equity/governance tokens associated with controlling and potentially earning profits from those exchange protocols and deposit/lender protocols. However, there are a lot of questions around what any of this means in the long run. Here are some examples of things that can happen with DeFi: Person A has a bunch of bitcoin, transfers it into Wrapped Bitcoin (“WBTC”), and so now it’s a custodial bearer asset that is tradeable on the Ethereum network. They then deposit that WBTC as collateral on Aave (AAVE-USD) (a borrower/lender protocol) to borrow stablecoins. They send some of those stablecoins back to a centralized exchange and sell them for normal fiat currency, which they withdraw to their bank to use for daily spending. They now have leveraged bitcoin, with multiple layers of counterparty risk to hacks and custodian problems, and as well as a chance of liquidation if the price of bitcoin goes too low. Person B has a bunch of ether, and deposits it as collateral into Aave to borrow a bunch of stablecoins. They then trade those stablecoins on a decentralized exchange called Uniswap (UNI-USD) for some DeFi equity/governance tokens (let’s say AAVE and COMP) as well as UNI, which is the coin associated with that same decentralized exchange. So they have leveraged ether as collateral to buy a bunch of DeFi protocol equity/governance tokens (AAVE, COMP, UNI). Person C doesn’t own much crypto assets, but decides to convert some fiat currency into stablecoins, and deposit those stablecoins into Aave and Compound to earn a yield. In this sense, they are funding the collateralized loans to Person A and Person B. Until the Federal Reserve increased interest rates this year, these stablecoin deposits earned much higher yields than they would get with a safer FDIC-insured bank deposit or a Treasury bill, albeit with much higher risk, and so this usage of capital kind of competed with junk bonds or similar ways to earn a high yield in traditional markets. If this all sounds familiar, here is what I wrote back in January 2021: One of my concerns, when reviewing the biggest use cases for decentralized apps, is that a lot of the use-case is circular and speculative. Ethereum is heavily used for decentralized exchanges of crypto tokens, crypto stablecoins that serve as liquid units of account for trading crypto tokens, and lending and earning interest on crypto tokens which is a practice that serves as a liquidity/borrowing source for traders of crypto tokens. To a lesser extent, it is also used for gamified ways to earn or trade various crypto tokens. So, it’s a big operating system powered by crypto tokens, for the purpose of moving around… crypto tokens. A healthy banking system in the real world would consist of people depositing money, and the banks making various loans for mortgages and for business financing, to generate real-world utility. A speculation-based banking system, on the other hand, would consist of a bunch of banks taking deposit money, and then lending to speculators in the nearby stock market, along with technology providers that make this easier, and then what those speculators are trading mostly consists of shares of those banks, shares of those tech companies, and shares of the stock exchange, resulting in a big circular speculative party. The biggest use case so far for Ethereum is a decentralized version of that circular speculation-based system. –Lyn Alden, An Economic Analysis of Ethereum , January 2021 When we look back almost two years later from my January 2021 piece that initially covered this industry, I can’t really say that it’s any different. It has been a circular and speculative party with little or no real-world utility the entire time, and the industry has made little if any improvement on that front. In fact, a March 2022 study found that 97.7% of the coins traded on Uniswap were rug-pulls. A rug-pull is when developers and investors hype up a project, extract value from it as investor money pours in, and have no intention to actually execute on their plans. The Ever-Present Risk Of DeFi Exploits DeFi offers some benefits of transparency, but it also offers a huge attack surface for hackers/exploiters. These are commonly referred to as “DeFi hacks”, where someone figures out how to steal funds from a smart contract. In reality, it’s better to think of them as exploits or arbitrage opportunities than hacks. “Code is law” was a common adage in the early days of DeFi (and of course, stretching back prior to DeFi). If there’s a bug in a smart contract, then someone can exploit it and potentially give themselves an advantage, up to and including taking assets in a way that the contract did not intend. This is similar to a poorly-written real-world contract, where a clever lawyer can find loopholes and help a client to abuse a contract in a way that was not intended when it was written. After all, if code is not the final arbiter of a smart contract, then what is the point of a smart contract? The Bitcoin network is the most streamlined blockchain, with nearly 14 years of operating history. It’s purposely simple, purposely changes very slowly, and is purposely resistant to change except in the case of overwhelming consensus. Despite all of that, the Bitcoin network has still encountered a number of serious bugs over the years. So, what hope do ever-changing and highly complex smart contracts have? Anyone who holds funds in DeFi or other smart contracts should assume that there is an ever-present risk of code exploits and loss of funds. You’re getting a 3% yield? Great. Weigh that against the probability of a 100% loss occurring at any point over the year, and then repeat that year after year. The risk is amplified when smart contracts from different blockchains or different layers interact with each other. More than two and a half billion dollars worth of crypto assets have been exploited from smart contracts over the past two and a half years: The Block Collateralized Vs Uncollateralized: That’s The Key Many DeFi proponents point to the fact that most CeFi lenders failed in 2022 while DeFi contracts continue to operate. While there is a grain of truth to that, we should examine the underlying reason: collateralization. CeFi lenders that made fully-collateralized loans generally held up well also. If they had to liquidate a client, it was the client that was ruined rather than the lender. The trouble occurred as CeFi lenders made under-collateralized or non-collateralized loans to entities that they thought were trustworthy (who then went and gambled it away in DeFi, such as on LUNA tokens, and were unable to pay back the loans). Due to DeFi’s automated nature, it’s very hard to do it in a way that makes sense without over-collateralization. So, by its very nature, DeFi consists of mostly over-collateralized leverage. The lesson here is not that DeFi is better than CeFi; it’s that when it comes to volatile assets in particular, over-collateralized loans are safer than under-collateralized or non-collateralized loans. In a DeFi environment, a depositor mainly has exposure to collateralized lending (a good thing, as far as lending is concerned), but also faces constant smart contract exploit risk (a bad thing). In a CeFi environment, a depositor may have a exposure to some blend of collateralized and non-collateralized lending (a bad thing), but has greater protection from exploits (a good thing). My takeaway from the crypto industry events of 2022 is not that DeFi is better than CeFi. Instead, my takeaway is that with these highly volatile assets, chasing yield is inherently unwise . And to the extent that cautious lending is done, it should mainly be over-collateralized lending. Bitcoin offers the user the ability to self-custody their own units of a finite liquid asset, and to send or receive permissionless payments with that asset, without relying on a centralized third party. In my view, that’s the signal through the noise, and that’s more powerful than many people yet realize. Intermediate-term usage of stablecoins for savings and payments in developing countries, where accessing dollars is difficult, is another good utility use-case. They should be aware of counterparty risks, and overall I’d like to see greater and greater transparency of stablecoin collateral. Almost everything else involves speculating, or involves picking up nickels (yield) in front of a steamroller (opaque counterparty risks and/or code exploits). The fact that I dabbled in yield-seeking for a year, even with a small amount of capital, and even though I stopped well before the blow-ups happened, is something that I consider to have been an error when navigating through this industry. The Centralized Shortcomings Of DeFi Aside from the circular, speculative, and sometimes outright fraudulent nature of the crypto industry including the DeFi sub-industry, there are centralization problems at the heart of the technology, even though it is marketed as being decentralized. Centralized Smart Contract Blockchains The foundation of the DeFi tech stack, the underlying smart contract blockchains, already start with some centralized aspects compared to the more decentralized Bitcoin network. For example, the Ethereum network had difficulty bombs in the code for over seven years from its inception in 2015 until its proof-of-stake conversion in 2022. This reduced the power of miners and individual node operators, and enhanced the power of the head developers, which is a form of centralization. It allows them to push through a roadmap and change the protocol according to their vision, which basically makes it an investment contract. Even after the proof-of-stake conversion here in 2022, users of the network still have to wait for developers to implement staking withdrawals. Binance, meanwhile, can pretty much ask the Binance Smart Chain to pause if there is a problem, like it did between October 6th and 7th of this year . The Binance chain in theory is a separate and decentralized system from Binance the centralized company, but in reality: CZ, Binance CEO, via Twitter Similarly, when Solana unexpectedly went offline on five separate occasions in 2022, the validator operators had to meet via Discord to manually restart the chain. Many proof-of-stake chains operate like this. The technical or capital requirements to be a validator are quite high, which ends up making the operation of the system rather oligopolistic. And unlike proof-of-work protocols, proof-of-stake protocols have no unforgeable history of the ledger, and so if the system intentionally or unintentionally goes offline, it’s a rather manual process to determine where the appropriate checkpoint is and where to restart the network from. Since it is costless to create a nearly unlimited number of alternative histories, each seemingly as valid as the next, there is no unforgeable, self-verifiable way to determine what the “true” history of the ledger is in a proof-of-stake system (that’s what proof-of-work does specifically), and so with proof-of-stake checkpointing, some authority or set of authorities needs to be trusted. That’s why there are projects like Babylon Chain that allow proof-of-stake chains to insert unforgeable timestamps into the Bitcoin blockchain. It’s an attempt to mitigate some of the inherent circular problems with proof-of-stake systems by making use of the dominant proof-of-work system. They use the Bitcoin network as the checkpointing authority. In addition to centralized difficulty bombs, centralized developer decisions, centralized validation, and/or the centralized checkpointing authority problem, there’s also the simple problem that in most cases, the smart contract blockchain nodes are too big. Optimal blockchain privacy and decentralization occurs when a user runs their own node, or at least has the practical option to if they should need to. This allows them to verify aspects of the network for themselves, and allows them to initiate transactions themselves rather than asking a third party to initiate a transaction for them. However, by adding higher throughput or more code expressivity right on the base layer of the protocol, it increases the processing, storage, and bandwidth requirements of running a node. In order to run a Solana or Binance node, you pretty much need enterprise-grade equipment or use a cloud provider, for example. Ethereum nodes are lighter than nodes for Solana or Binance, but are still too bulky to run over Tor. According to ethernodes.org , there are only around 6,700 Ethereum nodes. Of those, over 4,400 are hosted by a provider (generally a cloud service), including 2,700 that are hosted via Amazon specifically. There are only around 2,300 non-hosted nodes. Most users and apps instead rely on third-party node operators like Infura and Alchemy (which themselves make heavy use of cloud providers). When the US Treasury Department sanctioned the privacy-focused smart contract Tornado Cash in August 2022, Infura and Alchemy complied, and ceased processing transactions related to Tornado Cash. This means that many foreign people, including those who aren’t even subject to US sanctions, can’t use Tornado Cash unless they are willing to run their own Ethereum node, which is nontrivial to do and can’t be run over Tor. In contrast to all of this, back when Satoshi Nakamoto invented bitcoin, he purposely made sacrifices in terms of bandwidth and complexity to make it as small and simple and decentralized as possible. This made it possible for individual users to run bitcoin nodes comfortably on a laptop with a normal internet connection. The requirements to run a node expand more slowly than technological improvements in processing, storage, and internet bandwidth, so it directionally becomes easier to run a bitcoin node over time rather than harder. The goal of the Bitcoin network from inception was to cut out all of the unnecessary fat to keep it as lightweight as possible. Governments are good at cutting off the heads of a centrally controlled networks like Napster, but pure P2P networks like Gnutella and Tor seem to be holding their own. -Satoshi Nakamoto, November 2008 Centralized Custodial Assets Aside from some of the centralization aspects on the underlying smart contract blockchains that the DeFi industry relies on, the actual DeFi use-cases tend to have even more reliance on centralized entities. The vast majority of DeFi total locked value relies on centralized custodial stablecoins and other centralized custodial assets. Stablecoins make up a large portion of the lending and borrowing in the pseudo-decentralized deposit/lending protocols, and stablecoins are common trading pairs in the pseudo-decentralized exchanges. A fiat-collateralized stablecoin, such as USDT or USDC, is one where a centralized issuer holds assets in the form of bank deposits, Treasury bills, repurchase agreements, commercial paper, or similar types of assets, and issues redeemable token liabilities that are tradeable on a blockchain. In other words, there is a centralized issuer that manages the collateral and processes redemptions, but the liabilities are digital bearer assets for the holders, and thus can be traded between people efficiently and automatically with no further action needed by the centralized issuer. The centralized issuer can, however, still choose to actively freeze specific addresses as requested by law enforcement or due to various code exploits. In addition to the heavy reliance on centralized custodial stablecoins, DeFi on Ethereum makes heavy use of Wrapped Bitcoin “WBTC”. This is a custodial product whereby bitcoin is held in custody, and its token liabilities can be leveraged or traded on the Ethereum blockchain. The amount of custodial bitcoin wrapped on Ethereum rivals the amount of custodial bitcoin held by the largest crypto exchanges. Much like stablecoins, this is a centralized product whose liabilities are bearer assets. The worst-case scenario for how this could play out on Ethereum, has already played out on Solana. There are tokens on Solana representing wrapped bitcoin and wrapped ether, to allow those assets to be leveraged or traded within the Solana ecosystem. The problem, however, is that FTX was the issuer of those assets, and FTX is now bankrupt. As a result, those custodial assets de-pegged and lost almost all of their value: Coin Gecko Attempts At Synthetic Stablecoins Some stablecoin developers have tried to get around this reliance on centralized issuers and custodians. After all, if most of the value of the assets traded on “decentralized finance” protocols are themselves completely centralized, then is that term even appropriate? Just because the liabilities of a centralized entity are bearer asset tokens, doesn’t make the system decentralized. MakerDAO, for example introduced DAI several years ago, which began as an ether-collateralized synthetic stablecoin. What this means is that rather than being redeemable for actual dollars like USDT or USDC are, DAI is backed by an over-collateralized amount of ether and balanced with a stabilization algorithm, to synthetically represent one dollar. Fiat currencies are centrally-managed ledgers. Trying to peg a pseudo-decentralized ledger (e.g. a crypto-collateralized stablecoin protocol) to an actively-managed and centralized ledger (e.g. the Federal Reserve system) is always going to come with various limitations and risks. In this case, using a volatile asset as collateral means that there is a reasonably high chance of sudden liquidation of that collateral. Sudden liquidations are likely to occur during tumultuous market periods, which leads to chain congestion and extremely high transaction fees as too many people scramble for the exit at once, and too many liquidations are triggered at once, for the network to handle properly. In March 2020, during the global market crash associated with COVID-19 becoming a pandemic and various economies locking down, the price of ether crashed, and in an emergency situation, MakerDAO ironically voted to add centralized USDC fiat-collateralized stablecoins as a form of DAI collateral. This is because a crypto-collateralized stablecoin (which is basically the attempt to back a low-volatility asset with a high-volatility asset) is inherently either fragile or capital inefficient. Ever since then, USDC has been a very large percentage of DAI’s collateral. USDC could freeze DAI’s collateral and basically end that project at any time, if asked to by the government. In order to have an arbitrarily low probability of liquidation, a crypto-collateralized stablecoin requires an arbitrarily high ratio of over-collateralization. In simplistic terms, if you want the stablecoin to avoid having to liquidate even with a 75% drawdown of the collateral, then you need 4-to-1 overcollateralization. If you want the stablecoin to avoid having to liquidate even with a 90% drawdown of the collateral, you need 10-to-1 overcollateralization. This would be extremely capital inefficient if handled that way, and therefore most protocols will try to get around these levels by relying on incentive mechanisms to bring in more collateral from the community when it is needed, rather than keeping it there all the time. For example, newer crypto-collateralized stablecoins like Liquity and Zero aim to be 100% crypto-collateralized via incentive mechanisms. Liquity (LUSD-USD) is collateralized by ether, and Zero (ZUSD-USD) is a fork of it on the RSK sidechain of the Bitcoin network that is collateralized by RBTC, which is RSK’s merge-mined federation-wrapped version of bitcoin. The way this works for Zero is that ZUSD synthetic stablecoins can be created when someone deposits RBTC into the smart contract as collateral and receives ZUSD as a loan on their collateral. If they get liquidated (which occurs if the value of the RBTC falls below the maintenance threshold on the contract), they lose their RBTC but keep the ZUSD. The system overall optimizes towards a 1.5x or higher level of system-wide overcollateralization, although individual loans can be as low as 1.1x overcollateralized. If the system as a whole reaches sub-1.5x levels of collateralization, it liquidates loans under 1.5x levels of collateralization and changes the incentive mechanisms for borrowing to encourage higher levels of collateralization. Importantly, there is a stability pool that incentivizes holders of ZUSD to deposit and support the lending. If loans get liquidated, these ZUSD depositors get their funds transformed into RBTC at a discount to the market price of RBTC (which is conceptually kind of like selling put options to wait and buy RBTC at a discount, since most of these participants would generally be bullish on RBTC). So, the system encourages excess ZUSD that is floating around in the market, to come back and get burned in exchange for the ZUSD holders receiving RBTC, to reduce the chances that the dollar amount of ZUSD is ever not fully backed by a greater dollar amount of RBTC. These incentive mechanisms are interesting. LUSD, with its longer history than ZUSD (and that ZUSD is based on) has managed to maintain its USD peg throughout the 2021 and 2022 bear markets, where ether (the collateral for LUSD) was at one point down by around 80% from its high. Here is a post from May 2021 analyzing how it held up well in an ether flash-crash scenario, thanks to a properly-working oracle and a properly-working stability pool. However, it has been less than two years of operation overall, and LUSD is tiny compared to fiat-collateralized stablecoins like USDT and USDC, meaning that DeFi as a whole is still incredibly reliant on custodial assets. I view bitcoin as being better collateral than ether, all else being equal, and so I naturally would prefer DeFi on the Bitcoin network, at least to the extent that I would be interested in DeFi to any meaningful degree. I don’t plan to leverage any of my bitcoin, and I’m happy to hold bank dollars, money markets, or T-bills to the extent that I want liquid dollar exposure, so I’m not exactly the target audience here. Ongoing holders of LUSD and ZUSD have to trust 1) that the underlying smart contracts won’t be exploited for the foreseeable future, 2) that the incentive mechanisms will continue to work for the foreseeable future to properly maintain over-collateralization through all market conditions, 3) that the price oracles won’t be gamed in any destabilizing way, and 4) that the governance of the smart contract won’t become misaligned with users or otherwise captured (referring either to the specific contract governance or the underlying computational layer governance). Centralized Oracles Tying digital assets to real-world information involves a centralized oracle, or a quorum of several centralized oracles to try to spread out the oracle readings. An oracle is a source of information that a smart contract requires in order to execute its function. For example, tying a crypto asset to the dollar means that the smart contract needs to know what the dollar-denominated price of that crypto asset is, which means it needs information from an exchange. Similarly, a sports betting smart contract needs an external source of truth that collects the information from the real-world sports game, so that the smart contract can award the gains to the winning speculators. This reliance on one or more oracles represents yet another point of centralization. Who controls the oracle or oracles? How easy is it to manipulate a given oracle or set of oracles and exploit a contract? Centralized Governance Actions Many deposit/lender protocols and exchange protocols have centralized web-based user interfaces, and centralized companies supporting them. The underlying technology may be open source and accessible without them (requiring the user to run a node, or otherwise be rather hands-on technically), but for most users, that web-based user interface is how they access the services. There have already been precedents for these centralized companies and interfaces removing certain tokens from their trading or leveraging environments, either when asked to by a government or pre-emptively to try to avoid breaking securities laws. And many of these systems have far more centralized control than they let on. For example, as Jameson Lopp pointed out earlier this month, the DeFi protocol called Serum on the Solana blockchain markets itself as fully decentralized in its Twitter bio and elsewhere, and yet right under its bio there is a pinned tweet about how it went defunct due to centralized control. This is what you see when you look at their Twitter bio as of this writing: Project Serium, via Twitter How can something be considered “fully decentralized” if one centralized entity holds the upgrade authority? Many of these protocols try to to answer that question by breaking up control of the service by using governance tokens. Either from an initial coin offering or from airdrops to users, these tokens can be distributed, and allow for voting on various governance actions that relate to the operation of the service. These same tokens may also earn a share of profits generated by the protocol. While this is interesting in theory, it begins with several problems. First, as with most things in crypto, the vast majority of governance tokens are held by whales, meaning that a small number of wealthy entities retain the controlling share of voting power. Second, voter turnout tends to be low, which gives highly active and incentivized parties even greater control over the protocol’s governance in practice. Third, because it is pseudonymous, these governance models tend to be more easily gameable than real-world democracies. Pay-for-votes, entities secretly controlling a larger total share of the governance than the market realizes, and other problems make it hard to operate in a truly decentralized way. The co-founder of Ethereum, Vitalik Buterin, recently brought up the economic problems associated with governance tokens, and I agree with him: Vitalik Buterin, via Twitter So, recurring profits may be able to support token value (e.g. an equity security in digital form), but governance alone is an insufficient reason for a token to hold sustained value. And regardless of whether they earn profits or not, they’re likely going to be rather centrally-held and controlled in practice. Overall, there is a high probability that DeFi protocols will face greater scrutiny and regulation over time. And because they have so many centralized attack surfaces, it’s not that hard for regulators to clamp down on them, reduce their usability, and increase their trackability. When it comes to the DeFi ecosystem, their blockchain nodes mainly run on centralized cloud providers, most of their locked value relies heavily on centralized custodians, and users primarily interact with the ecosystem through centralized web interfaces maintained by centralized companies. DeFi Doesn’t Solve The Fiat Onramp Bottleneck One of the criticisms that people in the DeFi industry tend to aim at bitcoin, is that bitcoin is heavily reliant on centralized exchanges and brokers. The vast majority of its buying and selling occurs in centralized exchanges or through centralized brokers. “That’s why we need DeFi,” many of them say. However, we need to conceptually separate the phases of 1) post-onramp speculation/trading vs 2) actual onramping and utility. If you want to trade in DeFi, how do you start? Do you magically teleport your capital into a DeFi ecosystem? No. First, you go through a fiat onramp exchange, like Coinbase or Kraken, by transferring money from your bank to the exchange. Or you go through some other centralized payment provider. Then, you can buy various crypto assets, and move those crypto assets into a DeFi environment. From there, you can trade and leverage those crypto assets across various smart contracts. So, DeFi isn’t cutting out the bottleneck of relying on centralized exchanges or centralized bank connections for the fiat onramp part of the process. DeFi merely offers post-onramp smart contract environments to trade or leverage crypto assets, as a competitor to the alternative of remaining on those exchanges to trade and leverage crypto assets there. But realistically, how many people should be trading or leveraging crypto assets anyway? These aren’t “banking” services; these services are mainly for speculators. The Bitcoin network, on the other hand, has a similar onramp bottleneck. You send money to an exchange or broker, buy bitcoin, and then from there you can withdraw your bitcoin from the exchange or broker. From that point, you are financially “self-sovereign”, meaning you can self-custody your own bitcoin, and use the decentralized network to send or receive permissionless bitcoin payments globally. Additionally, there are some multi-signature escrow services that you can use, if for some reason you want to get a loan (in fiat or in stablecoins) while using your bitcoin as collateral. The only way to go around the onramp bottleneck, should centralized exchanges ever all be shut out from the banking system, is peer-to-peer purchases. In addition to mining, that’s how people got into bitcoin in the beginning, before exchanges existed. There are now methods like Bisq or RoboSats or Hodl Hodl or Paxful or Azteco for people that want to buy bitcoin without going through a centralized exchange, and sometimes with anonymity. The downside is that there is limited liquidity; these types of services only work for buying or selling modest amounts of bitcoin. However, the number and size of these types of services would likely substantially increase if centralized exchanges are shut out from the banking system. As a tangible example, Nigeria shut out crypto assets from its banking system nearly two years ago. Nigerian banks are disallowed from letting customers send money to crypto exchanges. And yet, Nigeria has among the highest bitcoin/crypto adoption in the world on a per-capita basis, and bitcoin/crypto in the country has much higher adoption than the country’s central bank digital currency, the eNaira. How is this possible? Because they use various peer-to-peer methods to acquire bitcoin. They can even do remote work such as programming or graphic design, and get paid by foreign clients in bitcoin straight to their own self-custody, which they can then use for global payments. Where there’s a will, there’s a way. And when official consumer prices in aggregate have gone up about 5x since 2010 due to ever-expanding money supply, and when authorities have arbitrary power to freeze bank accounts of protestors, there’s certainly a will. Trading Economics The “killer app” of bitcoin is simply using it for what it was originally designed to do: self-custodying and sending/receiving permissionless payments in what has thus far been the most decentralized, secure, and immutable crypto asset. Trading/leveraging is something that a subset of bitcoin holders may choose to do, but it’s not as though the world is in desperate need for more ways to speculate on ten thousand crypto assets, especially when DeFi is only relevant post-onramp and has nothing to do with the on-ramping process from the existing fiat banking system. Summary Points In summary, DeFi environments are trying to address a genuine trading/leveraging problem, but generally consist of multiple layers of partial centralization. -The underlying smart contract computational layer is often not as decentralized as it claims to be, thanks to validator oligopolies or developer control, the reliance on a checkpointing authority due to there not being an unforgeable history of the ledger, and the general difficulty in the ability of users to run a node due to high bandwidth and storage requirements. -Any significant trade-able asset that refers to some external data, such as price or some real-world connection, requires some centralized or partially-centralized custodian or oracle. -Governance in practice, even for purely digital assets, often is centralized and can be traced back to a centralized business entity or small group of individuals. -Any potential areas of actual decentralization in leveraging/trading protocols that may occur in this industry only apply once the speculator gets past the fiat onboarding bottleneck anyway; DeFi doesn’t solve the fiat onboarding bottleneck.