Self-paying loans and the magic of Alchemix
The world of DeFi is establishing itself as highly innovative when it comes to financial solutions, from the permissionlessness of its lending/borrowing platforms to the eye-poppingly high APRs on offer for providers of liquidity. We have always been convinced on the idea that building an alternative set of banking rails from first principles, free of the shackles of “legacy” banking as we know it (with all its challenges) opens up amazing possibilities.
Yet a couple of months ago, we noticed a new project launch that seemed to defy all the rules of financial gravity. Founded by the pseudonymous @scupytrooples, this protocol was aptly named Alchemix, and its value proposition was that it could make loans that PAY THEMSELVES OFF AUTOMATICALLY, with no risk of liquidation.
The uproar on “traditional” financial Twitter was a categorical declaration of “PONZI”. Here was evidence that crypto and DeFi was just another bubble waiting to implode. After all, how can loans repay themselves? The leverage that would accrue in the system would cause the entire DeFi ecosystem to crater and collapse and all things crypto would be relegated to the history books alongside Tulips.
Watching from the sidelines, we had mixed initial opinions. On one hand, the TradFi Twitter had a point: self-repaying-loans with no liquidation risk seemed too good to be true. On the other hand, if the numbers behind Alchemix did add up, they may have triggered the apocalypse for banking as we know it.
In the 2 months or so since inception, Alchemix has picked up more than $1bn of Total Value Locked (TVL) – small compared to the likes of AAVE (with US$12.6bn of total value locked), but still an impressive feat to accomplish in 2 months. More importantly, as a protocol it generates (from its 10% cut of yield cashflow) north of US$20k worth of revenues DAILY.
Lest we forget: all this in 2 months.
With that backdrop, we set ourselves the task of digging through their whitepaper and doing our own research in order to work out if DeFi had just given birth to a miracle or a ponzi-antichrist.
Our views are laid out below. As always, the risks are abundant and we’re not in the business of giving out financial advice to purchase anything on this blog.
Do your own research – and if needed, call up your friendly private banker. They’re the ones that are paid to give out advice.
Alchemix 101
The basic mechanics of Alchemix are simple: a borrower deposits an amount of DAI stablecoin (1 DAI is 1 USD), and is able to make a loan of up to 50% LTV of the deposited amount, disbursed in the form of alUSD, Alchemix’s USD stablecoin.
Over time, the 50% LTV loan is paid off using the net returns from staking the initial principal of DAI deposited, which is itself staked into one of the staking pools operated by Yearn (which we have previously written about here).
The idea is that the yield gained from staking the principal DAI then pays back the loan over time, and the borrower doesn’t need to actually pay it back.
Diagrammatically, it would look something like this:
Source: Three Body Capital
Three things stand out from this set up:
There is only one person involved – the borrower borrows from him/herself.
There is no explicit ongoing interest due on the borrowed alUSD.
The “cost of capital” for the loan is really 10% of the interest which would’ve been earned by depositing DAI amounting to 2x the borrowed amount.
Economically, the cost of borrowing is effectively a percentage of the overall opportunity cost of depositing the collateral amount – whatever that opportunity cost may be, expressed as the optimised staking yield offered by yEarn.
House of cards? Ponzi?
The question one must reasonably ask is where the point of failure in this very optimistic set up is located.
One possible candidate as a point of failure would be the amount of leverage. Fortunately, at 50% LTV, even the most aggressive of fractional reserve banking math wouldn’t get the protocol into any trouble: maximum leverage in the system would be 200% which is hardly any cause for concern, assuming users recycle their loan proceeds to re-stake continually ad infinitum (which at some point stops making financial sense).
Another possible candidate would be the viability of the collateral. Again, credit to the team behind Alchemix for picking what is probably the most stable of stable coins as the first form of loan collateral to accept. For those unfamiliar with DAI, it is the collateralised stable coin issued by the Maker protocol, one of the oldest, most established decentralised lending protocols. It isn’t the biggest lending platform, but it certainly ticks the box of “slow and steady”. DAI is pegged at a value of $1/DAI, but each DAI maintains a collateralisation ratio of at least 150%. Collateral quality isn’t the issue here.
Ironically, given Alchemix’s contention that liquidation will NEVER happen (we will discuss this point in a while), even liquidity isn’t really an issue.
(Future versions of Alchemix are expected to accept ETH, BTC and other USD stablecoins as collateral, which would further expand the scope of application of the protocol – and its fee-earning potential. This podcast featuring Scoopy with Jason Choi of Delphi Digital is mandatory listening, as well as this one with the team at Bankless.)
Rather, the critical assumption is that the yEarn vaults will continually pay a decent rate of return for staked DAI. As of time of writing, the yield rate of DAI deposits is around 20%, and assuming that the rate is stable (it isn’t), to repay a 50% LTV loan with a 20% return on principle value, less 10% of interest going to protocol fees, would take just over 2 years.
The lower the rate, the longer the time it takes for the loan to be paid off.
Now, regardless of one’s view as to the sustainability of yields in DeFi (and whether it is simply the function of an ongoing mania, driven by excessive demand for speculative assets via borrowing etc.), the worst-case scenario would be for yields to fall to 0%. What would happen then?
The obvious consequence would be that the yEarn vaults would return 0% returns, and the loans would never get repaid. Income to the ALCX protocol would fall to zero, rendering the income stream of the ALCX tokens worthless, and the tokenholders of ALCX also get zero-ed. Interestingly, because of the 50% LTV cap, the alUSD tokens may not necessarily fall in value to zero, and depending on the prevalence of alUSD in the crypto ecosystem, the fallout on the broader system as a direct consequence of Alchemix’s breakdown may well be relatively limited.
Additionally, there is a risk that the $1 USD peg of alUSD breaks for any number of reasons. While the protocol aims to have sufficient collateralisation (2x collateralisation with an asset that is itself 150% collateralised to $1 value – DAI), breaking the $1 USD = 1 alUSD peg simply turns alUSD into a bond whose value converges to $1 as the incoming interest payments repay the debt. The key variable here is time – the maturity of the bond is unknown and indeterminate.
That’s the worst case. But short of that apocalyptic scenario, every other positive yield value makes the concept behind Alchemix viable – and potentially revolutionary.
Banking without cost
Breaking it down to its roots, Alchemix represents an ideal that is the epitome of banking (just not for bankers) – a bank that has no bank: no branches, no rent, no staff, no advertising, no funding costs, no operating costs. Just code.
Because it has no costs, the “opportunity cost” of lending is effectively zero. It has no costs to cover once the smart contract code has been deployed, and its return hurdle for being a viable banking entity is zero. As a result, any yield it can collect is pure upside, even if it’s just a 10% slice.
Consider this: the capital is provided by the borrower, the collateral for the debt is the future non-zero yield from staking the deposit, and the maturity of the loan is a function of the rate of return that the deposit earns. The borrower and depositor are one and the same.
Because the bank of Alchemix doesn’t need to cover ongoing costs of operation or funding, it can afford to make loans that are paid off at an unspecified future date – the liability of the borrower is effectively to his/her own deposit.
Such a set up would be impossible under a traditional banking structure: wages and rent would need to be paid, funding costs for deposits and the need for demand deposits to be made available would mean that reserves need to be held in case of withdrawals, failing which a run on the bank may become a reality.
The other difference is that traditionally, the bank which makes the loans absorbs any yield from the collateral posted in addition to the interest costs charged. Of course, there are exceptions – in residential real estate for example, by occupying the home, the borrower is arguably enjoying the “yield” from the collateral asset, although what is to say that the interest rate charged by the bank doesn’t already compensate for that “loss” of carry for the bank?
In the majority of cases, especially where financial assets are posted as collateral, the bank gets the luxury of charging both interest and enjoying the convenience yield.
The Alchemix approach fundamentally changes that paradigm. Without a minimum hurdle rate of return to beat, all loans become “viable”, at least nominally. The bank of Alchemix can wait for as long as is needed for the yield on the deposited collateral to repay the loan – hypothetically, it could wait forever since it has no bills to pay on an ongoing basis.
Without an ongoing demand for liquidity, the need for liquidation of collateral, especially of overcollateralised DAI stablecoin collateral, disappears.
The cynics would be right in pointing out that there is a cost of time and opportunity that needs to be accounted for, an implicit hurdle which would need to be surpassed for such a loan to make sense. With neither a pre-defined maturity of the loan, nor a pre-determined rate of return on the loan, how would such a loan be priced?
And therein lies what we think is the absolute stroke of genius: because lender and borrower are one and the same, the loan actually doesn’t need to be priced. How does one default on oneself? If at any point the “rate of return” doesn’t make sense, the loan can be repaid by extinguishing the “deposited” amount, the balance unlocked to be reinvested. And if the borrower doesn’t want to repay the loan in this way, and is unable to stump up liquidity in any other form (e.g. the proceeds were used to buy something which went to zero), then the “penalty” to the borrower would be having 2x the borrowed amount locked up for an indeterminate amount of time – until the debt is paid up.
The beauty of it all is this: Alchemix is a risk-limited credit time machine. If the borrower is also the lender, then there’s very little reason (and in fact, very little point) in liquidating a “default”. When both time and default risk are simultaneously taken out of a credit structure, and a borrower borrows from his/her future self, things start getting very interesting.
Old trick, new level
Creating money out of thin air isn’t anything new. Fractional reserve banking has been in this business for centuries, once the bankers realised that not all depositors demanded their deposits at the same time. So, too, has the Eurodollar market – the market for US Dollars outside the US, where financial institutions issue USD credit to each other for the purposes of facilitating trade and lending denominated in USD, without themselves actually having the physical dollars to lend, requiring only the promise that – if needed – dollars can be made available. Leaps of faith are taken daily around the world, not least by people and parties who barely know each other and rely only on the trust and credit of their bankers.
Alchemix simply takes this to a singular and personal level: a borrower is given a facility to collateralise the future returns on his/her deposits and borrow against that in the present. Borrower and lender become one, and the protocol simply facilitates the discipline to save up to repay the consumption that has been brought forward to the present from the future – effectively mapping out what the economists among us would remember as intertemporal consumption smoothing, but without the banks.
The one caveat: one cannot borrow against what one doesn’t have. A starting deposit is required to get the ball rolling, and that is key – this is more about managing wealth and savings over time than creating aggressive leverage out of thin air.
Put in this way, Alchemix looks less like a ponzi than a way for individuals to make agreements with their future selves to pay themselves back by deferring consumption if things don’t go as planned. At a personal balance sheet level, assets and liabilities are created across TIME, rather than across entities, and that is a completely different paradigm to the status quo, albeit the best approximation to intertemporal utility function smoothing we’ve seen so far.
Disintermediating banks? That’s pretty interesting as an application of smart contracts.
Enforcing financial discipline over time, enforced by smart contracts – now that’s perhaps something that could turn out to be quite magical after all.
After all, better to be indebted to your future self than to a bank.