Author: Spencer Applebaum, Multicoin Capital Investment Partner

Translation: Big Pliers

Five years ago, developing a decentralized exchange (DEX) or lending protocol meant raising $3 million in funding, hiring five engineers, and spending six months to build a minimum viable product (MVP). Today, founders can utilize Claude Code to accomplish most of the work in just one weekend.

By letting a cutting-edge AI model develop a perpetual contract exchange, one can directly obtain a matching engine, risk engine, trading interface, API, and most of the necessary foundational functions.

AI is lowering the competitive barriers in almost all areas of the software industry, but financial services may be one of the hardest industries to commoditize. This is because the most valuable parts of financial services have little to do with the software itself.

Building a loyal customer base, accumulating a large balance sheet, and developing proprietary data and risk assessment capabilities all require years to establish relationships, accumulate track records, and earn trust. AI can significantly reduce software development costs but cannot make these capabilities easier to replicate.

In other words, a model can write a set of lending agreements in an afternoon, but it cannot conjure up $10 billion in stablecoin deposits for a smart contract or attract 28 million customers to a newly established digital bank.

We believe this trend is very positive for DeFi. In the next decade, trillions of dollars in financial activities will move on-chain, and AI will drastically reduce the development costs associated with products to service these activities. However, at the same time, the core factors that determine the value and competitive barriers of a financial enterprise have not become easier to replicate.

A New Pillar of DeFi Value Capture Framework

For years, we have been thinking about the issue of value capture in DeFi and have continually adjusted our views as the market evolves.

In the early days, we focused on how application tokens capture value through models like Work Tokens and Burn-and-Mint Equilibrium (BME). Then we shifted our focus to DeFi's defensive capabilities and Unforkable State. This perspective holds that while anyone can replicate open-source code, they cannot replicate liquidity and network effects.

Recently, we further proposed that DeFi protocols can only capture more value when they take on and manage certain risks. These views have been validated to some extent, and this article will further expand our thinking.

It is important to note that we believe DeFi will continue to erode the market share of centralized finance (CeFi) and eventually cover the entire traditional financial space.

The market share of DEX relative to centralized exchanges (CEX) has already reached an all-time high and continues to grow.

At the same time, tokenized assets are starting to gain real market adoption, whether they are U.S. Treasury bonds, private credit, or stocks. Importantly, the driving force behind this growth is no longer just native crypto users. Traditional financial institutions, fintech companies, institutional asset management firms, and traditional brokerages are all building on-chain products and bringing their customers into the on-chain world.

But the problem is that today's leading DeFi protocols may not necessarily be the ultimate beneficiaries of this growth trend. As software development and data analytics costs continue to decline, value will flow to those participants that possess scarce resources that software cannot create.

We believe there are three major moats in DeFi:

1. Distribution Capability
2. Capital
3. Proprietary Data

Additionally, there is a fourth factor—Risk Tolerance. Its mechanism of action differs and is more like a breakthrough entry point, which we will discuss at the end of the article.

Moat One: Distribution Capability

Exchanges and Other Stablecoin Distributors

One of the most influential theories in the early crypto industry is the "Fat Protocols" theory proposed by Joel Monegro in 2016.

This theory posits that most value on the internet is generated at the application layer, for companies like Google, Facebook, and Amazon; however, blockchain will disrupt this relationship and most value will be captured by underlying protocols.

However, the reality is moving in the opposite direction, and this trend is accelerating.

As the development costs of protocols decrease, value begins to shift towards the parties that control customer relationships. Companies that control customer relationships can direct users and funds to the highest-yielding platforms. Sometimes this principal is the application; sometimes, as is the case with Hyperliquid, it is a blockchain that also possesses application attributes.

Our definition of distribution capability is simple: controlling the flow of order or the movement of funds.

However, it is important to note that not all distribution capabilities have equal user stickiness. Accounts that are already funded and difficult to move can form true competitive barriers, while a simple user interface is more susceptible to competitive threats. This will be discussed further below.

The stablecoin value chain is the most obvious manifestation of distribution capability.

Historically, as long as a stablecoin reaches a certain scale and interest rates are not close to zero, issuing stablecoins has been a very profitable business.

Tether and Circle hold dollar reserves and earn interest on them, while in most cases, they do not pay interest to holders. Tether alone has earned over $20 billion in interest income since 2021 using this model, with a profit margin approaching 100%.

For example, Hyperliquid's platform holds over $6 billion in USDC as collateral for derivatives trading. This funding brings substantial reserve income to Circle and its distribution partners.

Hyperliquid gradually realizes that it is its own platform that facilitates the formation of these USDC balances, allowing it to possess bargaining power: if Circle is unwilling to share profits, Hyperliquid can replace USDC with another stablecoin willing to share profits and theoretically could even issue its own stablecoin.

Earlier this year, Hyperliquid partnered with Coinbase through its Aligned Quote Asset (AQA) framework.

According to the agreement, approximately 90% of the income from USDC reserves on the Hyperliquid platform will flow into the protocol's Assistance Fund for purchasing HYPE, rather than being owned by Circle or Coinbase.

Based on the scale of $5.5 billion in USDC in the treasury, the 90% corresponding income can currently bring Hyperliquid about $185 million in revenue.

Hyperliquid is not responsible for any of the underlying infrastructure here, including token minting and redemption, treasury management, or regulatory licenses. It controls the distribution channel, and that is enough to compel it to seek more economic benefits.

Wallets and Front Ends

Wallets and front ends are another embodiment of distribution capability.

In the early crypto industry, non-custodial wallets struggled to monetize effectively. Then, they gradually realized that the most lucrative business in the crypto industry is trading, so they began to make trading their core product a few years ago.

In terms of transaction fees, MetaMask charges a 0.875% fee, while Uniswap's average transaction fee rate over the past 12 months has only been 0.13%. Moreover, most of Uniswap's transaction fees flow to liquidity providers (LPs) rather than the protocol itself. In short, the revenue wallets earn from transactions far exceeds that of the platforms executing the transactions.

On the other hand, exchanges must attract liquidity providers, maintain complex infrastructure, and compete on trade execution quality; whereas MetaMask only needs to win and maintain customer relationships.

If a DEX cannot provide competitive prices, MetaMask can route transactions to other platforms. As its order flow scale continues to expand, it can gradually internalize more transactions, thereby retaining more revenue.

The perpetual contract trading space is also beginning to see similar trends.

Phantom allows users to trade perpetual contracts through Hyperliquid and charges an additional 5 basis points on top of the 4.5 basis points fee charged by Hyperliquid.

Hyperliquid is responsible for running the risk engine, matching system, and providing liquidity, while Phantom can earn more revenue simply by leveraging its user interface and customer relationships.

However, this model also has limitations.

Currently, the development teams in the Hyperliquid ecosystem have entered a fiercely competitive red ocean market, where price wars are underway, and teams are lowering fees to improve user retention.

In other words, even at the front end, the fate of commoditization cannot be avoided.

More broadly, crypto wallets are also quite fragile, as the cost of user migration is low. Users just need to import their private key or mnemonic phrase into another wallet to access the same funds through a new interface within minutes.

In contrast, migrating bank or fintech services is much more complicated. Users need to open a new account, complete KYC, transfer funds through traditional payment channels, re-link payroll accounts, and may need to update other related services.

Fintech Companies and Digital Banks

Take Robinhood as an example; it currently has 28 million funded accounts.

Currently, Uniswap is the main DEX on the Robinhood Chain, and Robinhood does not dispute this.

However, as AI lowers the costs of developing financial software, no factor can stop Robinhood from developing and launching its own DEX protocol, directing users to its platform.

With its massive user coverage and advanced programming tools, Robinhood could launch one of the largest Layer 2 DEXs in DeFi almost overnight.

Of course, we are not saying that Robinhood will definitely do this, as historically, mastering liquidity has not been Robinhood's primary goal. However, the company has indeed begun to shift a significant amount of trading volume from Kalshi to its own prediction market, Rothera.

What we really want to emphasize is that developing protocols is no longer the most difficult part. The real challenge is attracting users, capital, and proprietary data.

As a result, a value distribution chain is forming in DeFi:

• Front ends (like Phantom) gaining more economic benefits from trading platforms;
• Trading platforms (like Hyperliquid) gaining more revenue from stablecoin issuers (like Circle).

In all these cases, companies that control distribution channels have stronger pricing power over the underlying infrastructure.

However, how much value each layer can eventually retain depends on the ease with which users can leave.

Hyperliquid successfully captures about 90% of reserve income, while front-end applications within the Hyperliquid ecosystem are starting to compete on price, driving transaction fees down to single-digit basis points.

Moat Two: Capital

A natural conclusion from the previous section is that protocols and backend infrastructures lacking direct user entry points are at a disadvantage.

In some cases, this is indeed true. But besides distribution capability, there are other moats, one of the most important being capital.

Capital is the lifeblood of financial markets. Traders need capital to use leverage, market makers need capital to continuously provide quotes, and lending protocols need deposits to issue loans.

AI can help write code and automate workflows but cannot conjure up funds, which financial systems depend on for their operation.

When problems arise, such as the Aave KelpDAO incident or MakerDAO's "Black Thursday" incident, someone always has to bear the losses.

Before further discussion, let's talk about liquidity.

Liquidity has a logarithmic (log(n)) network effect: traders will flock to places with ample liquidity, and liquidity will follow traders' migration.

Hyperliquid's open interest is a true competitive advantage, but this capital actually belongs to users, who can transfer it to other platforms at any time.

We have seen liquidity in DeFi migrate from one protocol to another, often driven by token incentives and "vampire attacks." When these incentives disappear, substantial liquidity often flows away as well.

The following two forms of capital can create value with long-term defensive capabilities.

DeFi Prime Brokers

Prime brokers (PBs) are key participants in traditional financial markets.

They provide leverage, finance trading, and allow hedge funds to use the same capital pool across different markets and strategies without having to provide collateral on each trading platform.

This is an extremely large business.

Driven by the growth of the hedge fund industry and the reliance on financing in multi-manager pod shop models, the entire industry's prime broker revenue has risen from about $15 billion in 2020 to approximately $27 billion in 2024.

We believe the opportunities in DeFi could be even larger.

Traditional prime brokers mainly serve large institutions because the operating costs of servicing small clients are prohibitively high. Smart contracts can automate much of this work, allowing such services to be accessible to everyone, whether individuals with only $500 in trading amounts or investment teams managing hundreds of millions at Millennium.

Currently, traders often have to provide collateral separately on platforms like Hyperliquid, Aave, Lighter, Morpho, Variational, and Kamino.

This model severely lacks capital efficiency. As tokenized stocks, U.S. Treasury bonds, private credit, and other real-world assets (RWAs) continue to enter on-chain, this issue will only worsen.

For example, if a fund holds a long position in tokenized stocks on one platform and a short position in index perpetual contracts on another platform, they should not be required to provide full collateral for both positions if the risks can offset each other.

Now, building software to manage such businesses is becoming easier and easier. The real difficulty lies in having enough of a balance sheet to finance these positions, acquire low-cost capital, and possess the necessary risk management capabilities to avoid significant losses.

Over time, prime brokers with lower-cost capital can offer better financing conditions, attract more clients, and leverage economies of scale to further reduce financing costs.

In summary, this business should exhibit significant economies of scale.

Providing Risk Backstops for DeFi Protocols

The second way capital forms a moat is by providing backstops for risks.

This directly relates to our previous article on DeFi DAO risk management. At that time, we argued that protocols capable of managing risks have their associated tokens be less easily forked and replaced compared to those that do not manage risks. For instance, Hyperliquid's HLP treasury provides liquidity by making markets and taking on liquidation positions.

Depositors can share profits but must also bear losses. In March 2025, HLP incurred losses of around $4 million when closing a large ETH position, and the subsequent JELLYJELLY incident put it at further risk.

These two incidents highlight the importance of the derivatives DEX insurance fund.

Aave's Umbrella also serves a similar function: Users provide backstops for bad debts in exchange for corresponding returns.

Anyone can replicate the HLP or Umbrella smart contracts, but not the capital within them. However, there is a problem: This capital comes from third parties. If depositors find a better yield elsewhere, they may withdraw their funds.

A rebuttal to this is that some capital has a degree of stickiness because there may be overlapping interests between token holders and depositors. For instance, HYPE holders may be willing to leave some dollar funds in HLP to enhance the exchange’s safety and protect their token investments.

The most durable form of this strategy is when protocols build their balance sheets with retained earnings. Such capital is less likely to flow away due to new, more attractive yield opportunities or points programs.

Binance’s recent performance illustrates the importance of this model.

After suffering a $6.3 billion loss from a hack on October 29, Binance utilized its SAFU reserve fund and company funds (primarily the latter) to cover the losses and reopened withdrawals within days.

Ultimately, the net outflow of assets accounted for only 3% of total assets.

It is still too early to draw conclusions, but initially, market trust in its balance sheet strength may have helped it avoid a bank run.

As DeFi gradually encroaches into private credit, tokenized stocks, and other real-world asset domains, the demand for insurance and risk capital will continue to grow.

The end result is that protocols that can truly provide and manage capital will capture more value than those that merely provide software.

Moat Three: Proprietary Data

The third moat is information, but there is a prerequisite to note here.

DeFi is built on the principle that financial data should be public, which is very effective for collateralized lending and spot trading involving crypto assets. However, as more and more real-world assets enter on-chain, we will see more businesses built around non-public information.

The real moat lies in the ability to combine proprietary data with the ability to utilize that data for risk assessment.

Credit Curators

Lending businesses are the most obvious embodiment of this trend. Curators on platforms like Morpho, Kamino, and Euler are currently responsible for assessing risks and managing depositors' funds, with a total managed scale reaching $9.5 billion.

These curators obtain limited control access within treasury smart contracts, enabling them to adjust fund allocations to which lending markets, set risk parameters, and monitor collateral positions.

Although they manage a total of $9.5 billion, the income generated is quite limited, especially compared to traditional asset management companies that have greater autonomy and responsibility.

In fact, many managers do not even charge any fees. Measured by assets under management (AUM), the largest risk curator in DeFi, Steakhouse Financial, has a management fee rate of only 22 basis points, while the average rate across all managers is 19 basis points.

Some of the reasons for this phenomenon are that current risk-adjusted returns in DeFi are not attractive, possibly due in part to large amounts of idle capital in Aave. Another reason is that Morpho and Kamino have commoditized the treasury layer and are positioned between risk managers and depositors, which again highlights the distribution capability issue we discussed earlier.

But the bigger issue is that most of the data used by risk managers is the same.

If using ETH or BTC as collateral for lending, then asset prices, liquidation rules, and collateral information are all public.

AI makes it very easy to replicate risk models, so risk managers can only compete based on judgment (e.g., risk tolerance, a significant part of which is predefined by algorithms) and historical performance records.

Over a sufficiently long time span, this will inevitably exert pressure on fee levels.

In contrast, OnRe is a different case.

OnRe is an on-chain reinsurance company that raises capital from investors to underwrite insurance risks in the real world. To do this, it needs historical claims data, loss probabilities, and the insurance contracts themselves, not to mention professionals who truly understand this data. This information cannot be obtained from blockchain explorers. A cutting-edge AI model that cannot access reinsurance business data, no matter how intelligent, will not be able to complete relevant risk underwriting tasks.

This is also why OnRe can charge 75 basis points in management fees and additionally receive a 20% profit share. Based on current return rates, its annual charges are about 3%, which is 16 times the average charged by DeFi risk managers.

Of course, these are two businesses with different risk characteristics, and some portion of the fee premium reflects the illiquidity of the assets and the risk of potential losses. But this precisely illustrates the problem: Companies that possess proprietary data, risk underwriting expertise, and customer relationships can charge significantly higher fees.

If your data and tools are exactly the same as those of all other participants, then you will likely end up as a commoditized product.

What Does This Mean for Value Capture?

This framework naturally raises a question: If DeFi experiences substantial long-term growth, which participants are most likely to capture value? Some groups will benefit from this, while others will face greater difficulties in monetization.

L1 and L2

A somewhat pessimistic interpretation is that infrastructure tokens such as L1 and L2 are facing challenges. The reason is that in the past, blockchains could provide two of the three moats—distribution capability and capital—for new DeFi applications.

Historically, blockchains with distribution capability arose because DeFi users usually kept assets on their most active chains.

If you develop applications on Ethereum, you can access Ethereum users; if you develop applications on Solana, you can reach Solana users. Starting a project today from scratch with Uniswap, Compound, Aave, Kamino, or MakerDAO, this ecological advantage is very important.

Because you don’t have ready-made customers, assets, or financial relationships, launching projects in already populated ecosystems can significantly reduce the difficulty of entrepreneurship. However, today, advancements in interoperability technology are weakening this advantage.

Now, users can enter DeFi directly through applications and trade any asset on-chain with almost no consideration for the underlying blockchain.

More and more new users will enter DeFi through fintech companies like Revolut, Nubank, and Robinhood. These platforms already have tens of millions of customers, and users can access relevant services without needing to fund their wallets first.

L1 also provides capital through ecological wealth effects. As the prices of SOL and ETH rise, users have more funds to invest in applications on these blockchains.

Additionally, due to past difficulties in cross-chain operations, and users often having a strong sense of belonging to a particular chain, capital has also been more likely to stay within the existing ecosystem.

However, today, cross-chain operations have become much easier, and capital can flow into DeFi via stablecoins, tokenized assets, and institutional funds. Thus, the reliance of new protocols on the ecological wealth effects of any single token is decreasing.

Issuers of real-world assets also want to deploy assets across all blockchains, with projects like Securitize, Superstate, and Ondo already showcasing this trend.

As for data, blockchains have never really formed a strong moat. Almost all on-chain financial information is public, so blockchains and applications themselves have never had much proprietary data to establish competitive barriers.

AI is also undermining another advantage that blockchains previously had: developer ecosystems.

The significant advantage Ethereum has over emerging blockchains like Solana or Aptos is its mature development tools, audited codebases, and rich integration solutions. However, as AI makes it easier to migrate applications between different virtual machines (VMs), the importance of this advantage is declining.

From a broader perspective, generic blockchains are losing their advantages in distribution capability and capital, and they have never truly established a data moat from the outset.

This means they will ultimately only be able to compete at the execution and settlement layers, where block space is abundant and transaction fees are continuously declining. But this does not mean that blockchains themselves lack value.

As Tushar articulated in a keynote speech at a summit years ago, L1 tokens possess the attributes of capital assets, commodity assets, and value storage assets simultaneously. If Solana ultimately processes one billion payment transactions per day and charges a fee of $0.01 per transaction, its annual revenue would reach $3.65 billion.

As shown in the figure below, transaction volumes continue to grow.

The core conclusion is that the source of value for L1 and L2 may differ entirely from what investors originally anticipated.

Applications and Protocols

For applications, the companies that can capture the most value must possess at least one of the following three moats:

• Distribution Capability: having hard-to-migrate direct customer relationships, such as Robinhood or Hyperliquid.
• Capital: having a balance sheet capable of bearing losses.
• Information: having the ability to underwrite risks utilizing data that other participants cannot access.

No matter how powerful AI models become, these three capabilities are far more challenging to replicate than the software itself.

We do not believe it matters much whether value ultimately flows to equity or tokens. Our view is that both can capture value.

Recent SEC guidance and rule proposals have provided a corresponding path for tokens to capture value in a compliant manner, and we fully intend to invest in opportunities in this area.

What is truly important is whether a business, whether adopting a token or equity form, possesses a certain scarce resource, whether it thus has pricing power, and if it can convert that advantage into returns for investors.

As more and more financial activities shift on-chain, the gap between protocols that possess moats and those that merely have software will continue to widen.

As one of the largest investment institutions in this field, we invest a lot of time researching which projects in the entire tech stack truly possess these moats and can maintain them over the long term.

Will All This Ultimately Only Benefit Industry Giants?

From the analysis above, it is easy to come to a reasonable conclusion: the ultimate winners will be the existing giants in the industry.

Robinhood has customers, Goldman Sachs likely has more capital than any DeFi prime broker, and Moody's holds data and relevant licenses.

If software development is becoming free, and the real moats are precisely those costly resources, then companies that already possess these resources naturally have a competitive advantage.

However, startups still have one advantage: they are willing to take on risks that industry giants are not.

Part of the reason is that AI has a two-way effect. The same set of tools that allows Robinhood to launch a DEX in a weekend can also enable a startup with just three employees to do the same.

As the costs of building a competitive product continue to decline, the real opportunity for startups lies in entering markets and servicing customers that industry giants are unwilling to touch.

In fact, several compelling cases have already emerged.

Ethena

Ethena has built a synthetic dollar product around a delta-neutral hedging strategy. Traditional financial institutions are undoubtedly familiar with this strategy, known as basis trading, but they previously could not or did not want to introduce it to crypto exchanges. Today, Ethena has reached an assets under management (AUM) of $5.4 billion.

Pump.fun and fomo

Pump.fun and fomo burst onto the scene, creating a brand new trading experience around meme coins, and have established a massive distribution channel as a result. They have become one of the most successful projects in this crypto market cycle.

Hyperliquid

Hyperliquid entered a derivatives market dominated by Binance, Bybit, and OKX, and quickly captured market share. It has also started offering perpetual contracts for pre-IPO firms like OpenAI and Anthropic, a space traditional exchanges would not venture into. These markets attract new traders and liquidity. The number of wallets created on the day of the SpaceX IPO reflects this.

In terms of capital, Project 0 provides financing for DeFi collateral that traditional prime brokers are unwilling to accept.

In terms of proprietary data, USD.ai can secure loans using GPUs as collateral, whereas banks have practically no experience in handling such assets.

The common thread among these cases is that they have all entered markets that traditional industry giants have yet to service.

The reasons for this could be regulatory restrictions, operational complexities, reputational risks, or simply because the corresponding opportunities seem too small in the eyes of giants. As long as these situations persist, startups will always have the chance to enter the market in a non-symmetric competitive manner. The history of the crypto industry has proven this multiple times, and we expect it to continue in the future.

This is precisely why the combination of AI and DeFi is so appealing. AI is significantly lowering the development costs of financial products, while DeFi allows startups to directly reach customers, acquire capital without having to rebuild the entire financial system from scratch.

Of course, risk tolerance is not a lasting moat in itself.

Regulators will eventually catch up, and industry giants will enter the competition after the market has been validated (the Bitcoin spot market is a prime example). However, startups do not need to establish a moat immediately upon their founding. They simply need to find a market entry point where an industry giant is unwilling to tread and secure enough time to develop that entry point into one or more of the three moats.

After that, they can begin to accumulate and expand their advantages.

Conclusion

We believe that in the next decade, tens of trillions of dollars in financial activities will move on-chain, and AI will make it unprecedentedly easy to develop financial products to service these activities. However, despite software becoming increasingly commoditized, the core factors that determine the value of financial enterprises have not changed.

Customers, capital, and proprietary data remain difficult to acquire, and companies that control these resources will possess pricing power. This is also why we are so optimistic about the next generation of DeFi companies.

As AI development tools become more widely available, startups can now compete with industry giants faster and at lower costs than ever before. Once a new company successfully establishes one of these moats, it has the opportunity to build a company with a remarkably enduring competitive advantage.

Ultimately, you can copy the code and use the latest, most advanced AI models to regenerate the code. But you cannot directly generate these moats, and this is precisely the opportunity we see as crypto investors.

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到[email protected]，本平台相关工作人员将会进行核查。