Decentralized bank architecture

More will benefit from a system that is global and composable

Andy Singleton
12 min readAug 17, 2022


Andy note on 1/12/2023- We posted documentation and code for an implementation at

MakerDAO, Frax, Aave and Circle are converging to build the future of banking.

This article will describe the emerging architecture. It is being driven by historically wide interest rate spreads. Now we can fit these pieces together to expand the supply of savings and credit.

DeFi banks will reach more savers and builders

Banks handle $100T in assets worldwide. They fill a systemically important role in every modern economy. Banking connects savers and builders.

Decentralization can make banking more scalable, so that it reaches more savers and builders.

Savers want a way to earn some interest that they don’t need to think about. They also want to get their money back quickly, when they need it. Builders want money for investments that take risks and require a lot of attention. Their most useful investments lock up money for some time. A bank converts term loans and investments with intermittent payoffs (on the builder side) to safe and liquid demand deposits (on the saver side).

A typical bank makes about a 2% return on assets for doing this job. If the bank puts in 10% capital to cover losses and liquidity demand, the bank will make about a 20% return on equity. A decentralized bank can run more efficiently, and it can expand by tapping into capital and risk management from a decentralized network of “bankers.”

DeFi must deliver some advantages

DeFi has more technical risk than CeFi. A DeFi service must deliver significant advantages in order to compete. It should:

  • Use transparency to reduce overall risks
  • Expand the network of experts and investors that provide services
  • Improve access for new categories of customers around the globe
  • Use online coordination to pool funds quickly and at a large scale
  • Use automation to react quickly to market conditions
  • Increase the rate of innovation with open source contribution

Tokens are better than bank accounts

DeFi gives us a better packaging for a savings account. You can’t do much with a bank account number. You can do more with a token.

  • You can buy it and sell it in multiple channels. Your software can acquire the token through an API. You can program your own software for front ends, smart sweeps, payments, and other services.
  • You can use it as collateral for other investments. You can stack yield by borrowing against it. You can use it as the deposit on an apartment or the guarantee of a contract.

DeFi should contribute to the real economy

Crypto speculation is a small market with limited social benefit. DeFi should support the real economy. The new decentralized banking architecture takes on the difficult task of funding “real world assets” — off chain loans and investments.

High spreads are driving action

Stablecoin yields in DeFi collapsed to around 1%. In the off-chain banking system, interest rates have increased. T-bills are at 2.5%, and high yield bonds with risks equivalent to DeFi are offering 8.5%. The spread between 1% and 8.5% is motivating us to put money work in the real economy.

This drives growth in “uncollateralized lending” protocols such as Maple, Goldfinch, and TrueFi. They accept stablecoins into on-chain pools, and lend it off chain. They have grow rapidly during the past year to over $1.5B in assets.

From Orthogonal Trading, “2022 lending update

Demand for investment options is coming from the big stablecoin providers — MakerDAO, Frax, and now Aave. MakerDAO ended up with $4B in USDC that they acquired as collateral in order to meet a demand for DAI. Frax has about $1B. Aave recently announced a new stablecoin, GHO, which will engage some of their $1.5B in USDC. This is just the tip of the iceberg, with an estimated $90B of USDC and Tether sitting idle in wallets.

Holders of USDC do not get interest. Circle gets the interest. Circle makes a spread from selling USDC and investing the USD proceeds in investment grade money markets. That spread has been increasing, and is currently about 2.5%. Other stablecoin providers want to capture this spread. They want to redeem out their USDC, and invest the resulting USD into treasuries and investment grade bonds. They would make an extra $150M/year by redeeming and investing $6B of USDC.

They would also like to invest in high yield bonds and in higher rate loans. That takes them directly into the banking business.

Architecture is converging

Stablecoin providers are using different words to describe similar architectures.

  • MakerDAO has been working for years to put DAI into the safe tranche of what they call “real world assets” or RWA. Now they have $4B of USDC earning 0% interest that is burning a hole in their pockets. This is accelerating their normally slow and painful governance process. They have approved plans to put $500M into treasuries and investment grade bonds through a structure that has been approved by Swiss regulators. They have approved a plan to fund $100M of the loan book of an American bank.
  • Frax deploys “AMOs,” which are contracts that can sell Frax and invest the resulting USDC. They have programmed AMOs to put money into the senior tranche of Goldfinch and TrueFi lending pools.
  • Aave has announced a stablecoin called GHO that will put money into a range of assets provided by “facilitators”.

These plans include three components that fit together into a DeFi banking system.

  1. Liabilities. Savers purchase and hold a dollar coin that is the liability of the system. In these examples, they get a “stablecoin” that pays zero interest. However, most stablecoin protocols actually do pay interest by paying rewards into AMM pools and staking pools. They need to do this in order to generate demand for their brand of coin. A new generation of “savings coins” will simplify this system by paying interest directly the coin holders.
  2. Assets. The system earns interest by investing dollars into overcollateralized loans, other on-chain strategies, and off-chain loans and investments. The off-chain investments are labeled as “uncollateralized.”
  3. Agents. Someone needs to figure out where to put money, and manage it actively for a good result. A normal bank has a centralized management team that oversees investment into a portfolio of assets. DeFi can decentralize this structure, so that each asset has an agent. Maple calls them “pool delegates.” Aave calls them “facilitators.” I will call these agents “bankers”.

Aave shows how it will work

Aave is planning a new stablecoin, GHO, that will compete with MakerDAO’s DAI. Their diagram shows the full scope of a decentralized bank.

Liability: GHO is a stablecoin. It will officially pay zero interest. GHO minters or “borrowers” pay interest that goes to Aave, and some of that money can be passed through.

Andy’s note: Aave will need a way to convert their GHO coin into dollars that borrowers and investment sellers will accept. Frax has a pure implementation of this. Their system sells Frax to get USDC, which can be used directly in on-chain investment strategies, or redeemed into bank accounts for off-chain investments and loans. Aave will need to spend some of their net interest margin to pay interest, and to persuade people to trade their redeemable coins for GHO.

Assets: They plan to accept a wide range of assets. They are not taking a position about whether these assets are held on-chain, or in some off chain custody arrangement. This opens them up to funding securities and loans.

Agents: “Facilitators” are the “bankers”. Facilitators pay an interest rate to borrow GHO that is set for the whole system by Aave governance. They earn yield from their collateral. So, they make a spread whenever their collateral or investment returns are greater than the GHO borrow rate.

Top-level insurance: Aave pays a discount to GHO borrowers who also stake into Aave’s safety module — a type of insurance pool.

Asset allocation: According to Aave, “A facilitator (e.g., a protocol, an entity, etc.) has the ability to trustlessly generate (and burn) GHO tokens.” Aave will approve a deal with a facilitator. After that, they don’t do any asset allocation. They “trustlessly” match the facilitator. This simplifies DAO governance enough to be a game changer. It uses a “waterfall” structure to push risk onto the facilitator.

Waterfalls protect savers


Loans and investments can be risky. Financial products handle this type of risk with a “waterfall,” where “senior” investors get paid first and get a safe investment with a low and predictable return, and “junior” investors get paid last and take excess losses or profits. Here are three examples.

In the simplest banking structure, bank deposits make up the senior tranche. Bank equity investors put money into the junior tranche. The bank agrees to pay a low fixed rate to the depositors. If they make a return on loans and investments that is greater than this fixed rate, they can distribute the extra profit to equity holders. If they take a loss, they have to pay it out of equity.

Loans often have additional junior tranches, such as a “first loss” from the originator, and a down payment from the borrower. Each layer contributes increased safety for the savers in the senior tranche.

A decentralized bank will tap into a community of experts to set up investments and loans, and provide the junior tranche money. We will call these experts “bankers”. They are essentially borrowing money from savers. Their goal is to make a spread from loans and investments.


We propose “stabilizers” as a standardized asset that fits into multiple stablecoin and bank architectures. A stabilizer is a smart contract that implements a waterfall. It has a set of parameters that define the obligations of the junior tranche holders — the bankers. We can use these parameters to design a senior tranche that fits the risk, return, and liquidity requirements of a specific protocol.

Protocols like MakerDAO, Centrifuge, Maple, and Goldfinch use different contract structures, parameters, and legal structures to define their waterfalls. Some of these structures suffer from complexity and a lack of legal standing.

This diagram shows a simple structure that does the job.

  1. It sets some rules that protect senior tranche holders and specify how money comes in and goes out
  2. It adds a pluggable asset — the investment or loan — that can earn returns and hold collateral on behalf of the pool
  3. It bridges from the unregulated DeFi money in the senior tranche, to real world asset investment. It provides a legal entity, the banker, that can make loan agreements and invest in securities.

A stabilizer defines the rules for working with junior tranche holders. Bankers have to put in some money to pay for possible losses, take money out last, etc. The bankers — human professionals with expertise and money at risk — can make active allocation decisions.

We can simplify DAO governance

A stablecoin or savings coin is defined by the list of assets that it will fund. We can define a stablecoin or savings coin by changing the list.

Asset pools in this list (guided by bankers) can act to invest and redeem.

Frax pioneered this architecture. It seems weird because the all-important flow of money is not controlled by a central authority. The Frax coin has a “minter” list that authorizes specific AMO’s (investment strategies) to mint it and sell it. After that, it trusts the AMO to mint and redeem at favorable times. Aave is planning to use this system with a “facilitator” list. MakerDAO is using automatic matching in their deal with Huntingdon Bank, in which they match 50% of the funding for Huntingdon’s loan book.

Automated minting has some advantages for a blockchain protocol.

  1. Simplified governance. After a DAO authorizes a strategy it usually does not need to more decisions. This minimizes reliance on slow and tedious DAO voting.
  2. Utilization. If we have a big enough pool of strategies that are minting and redeeming, each strategy will not need to keep excess cash. This increases returns for investors, and it lowers costs for borrowers.
  3. It is cross-chain. Investments that are administered on a particular blockchain can buy dollars on that chain. This delivers dollars without going through (insecure) bridges.

This is a game changer for DAO governance

The “minter list” simplifies the governance of a DAO that manages a stablecoin or savings coin. It is up to the bankers to set up structures, investments, loans, and stabilizers that meet the requirements of the DAO. Then, they can request funding for their senior tranche. The DAO only needs to approve or disapprove the funding.

It also simplifies the legal structure of the DAO. The legal theory of a decentralized DAO is that it is not an organization. It is a group of people that use the same software. DAOs can run active governance as long as they focus on proposals that modify the shared software, or set parameters for the shared software. Proposals that add and remove assets on a minter list fit into the role of a decentralized DAO.

Composable and forkable

In this architecture, assets are composable. Protocols can shop around for assets. Bankers can target the design of their safe tranches to particular aggregators.

Coins can fork. They can take the basic structure of the coin, and implement a different minter list, with different requirements, and different governance. An option to fork would solve a lot of problems at MakerDAO, where two factions are locked in a dispute about how to improve their process for evaluating real world assets.

We may see a range of instruments with different returns, risks, and liquidity constraints. I think that the successful coins will all be perpetual instruments that bundle together a lot of different assets. People don’t want to shop around and roll over between individual assets. However, they are willing to shop for one perpetual savings or money market account.

Unbundling Circle

In this world, some of Circle’s job will get unbundled. I estimate that they make a spread of about 2.5% (250 bips). In return, they do a great job at creating a single asset (USDC) and a single bridge (a Circle account) for moving between banks and various blockchains. That’s why USDC is the core unit of exchange in the emerging decentralized banking system.

However, the market rate for managing a safe portfolio of treasuries, commercial paper, and investment grade bonds is about 35 bips. So, there is economic pressure to unbundle in favor of “bankers” that provide less expensive service. And, we will see competition in capital backing (currently under 1% at Circle), bridging (new bridges from vendors like Aave allow you to “borrow” on other chains with a specific cost) and transparency (can be improved to daily or real time).

We need to solve some problems

A lot of people have argued that decentralized banking can never work. It is difficult to provide credit without losing money. Maybe it is fundamentally too risky and too difficult for a decentralized system.

The decentralized system does provide some solutions. Typically, bankers will put some of their personal money at risk. This can reduce “agency” problems in banking bureaucracy. And, the transparency of the system reduces the probability of cascading defaults, which can be caused when creditors pull money away from lenders with unknown risks.

We need to address some problems.

Providing liquidity

Financial systems often experience “panics” where everyone wants their money back at the same time. Yikes! This is an important case that I will cover in a complete article.

Managing risk and defaults

DeFi bankers should put in more capital. JP Morgan is the biggest, most diversified, most government supported bank on the planet. They keep $285B in capital, covering 12% of assets. Much riskier pools at Maple have 3% “cover”.

We can improve lender representation, custody of assets, and transparency. DeFi reduces risk by exposing where the money is, in real time. DeFi banking will pull all of its related investments and loans toward real-time transparency.


Finance is heavily regulated. It’s systemically important, historically vulnerable to fraud and instability, and targeted by governments as a point of control. Global DeFi does not fit cleanly into nationally regulated systems. I’ll cover several scenarios for regulation in a new article.

Maxos will make three contributions


Stabilizers are standardized assets for decentralized banking. They put DeFi money to work with expert investors and lenders. They bridge from on-chain pools to real world investment with a streamlined software and legal structure.


We’re taking action to originate loans with better lender protections and transparency

SWEEP savings coin

SWEEP gives wallets, apps, and treasuries a reliable way to earn interest from unused dollars.

Follow this channel

Future articles will cover:

  • Stabilizers — a standard structure for composable assets
  • Liquidity — How do we pay money back? What happens when everyone wants money back?
  • Compliance — How do we fit this global system into national regulations?
  • Money supply — How does DeFi contribute to macro investment?

If you have read this far, you are in an elite group. Please consider following Maxos on Medium or on Twitter, or chatting with me on the Maxos discord.



Andy Singleton

Software entrepreneur/engineer. Building DeFi banking at Maxos — . Previously started Assembla, PowerSteering Software, SNL Financial.