ETHE - Ethereum: Late Bloomer Of The Artificial Intelligence Boom

2023-08-10 12:34:55 ET

Summary

• AI has the potential to accelerate the development of Ethereum's decentralized network and increase the value of ETH.
• AI can enhance the efficiency and security of dapp development on Ethereum, leading to more activity and growth on the platform.
• AI can also improve decentralized finance (DeFi) and decentralized trading by enabling automated on-chain strategies and providing insights into smart contract behavior. This can attract more users and capital to Ethereum.
• AI systems will require interoperability in the future, and the Ethereum network's fast finality, transparency, and programmability serve as an optimal solution.
• I deconstruct ETH tokenomics and a framework for valuing ETH as an asset.

This article explores how artificial intelligence ((AI)) could be a strong catalyst for Ethereum’s growth as a decentralized network and for Ether ( ETH-USD ) as an investment.

Ethereum is a decentralized, proof-of-stake (PoS) open source blockchain that enables the creation and execution of smart contracts. It has quickly become one of the most prominent and influential blockchain networks in the world.

Unlike Bitcoin, which is primarily a digital currency, Ethereum is designed to be a versatile platform that hosts a wide variety of decentralized applications (or “dapps”). At the heart of Ethereum's functionality are smart contracts, which enable developers to create complex, automated interactions without the need for intermediaries.

ETH serves as the fee (called “gas”) for executing transactions and performing computations on Ethereum. After Ethereum’s switch to PoS in 2022, ETH has taken on a deflationary and value accrual economic design. The thesis is that AI should have a positive impact on ETH’s price through the following channels:

• AI greatly enhances the efficiency of developers and the security of written code. This will lead to ETH capturing the long tail of more AI-assisted dapp development.
• AI’s ability to generate unique insights from data can be applied to the vast amount of data provided by public blockchains. Today this application is probably best used for on-chain trading based on money flows, and such applications will lead to more activity on Ethereum. However, many more blockchain data use cases could appear in the future.
• The next frontier of AI will be interoperable AI systems. Revenue models of AI providers will likely remain as pay-as-you-go schemes for systemic reasons. Thus, interoperable AI systems will probably require a programmable fast-finality settlement layer, where auditability is a perk. Ethereum is the natural candidate for addressing this need.

The result of this is that ETH is the “late bloomer” for the current boom in AI – ETH has explosive upside as AI begins to catalyze the development of Web3 and blockchain, and even more so when Ethereum is used as a tool for AI interoperability.

First Things First: ETH is Not an AI Pure Play

The pure play of AI is going to be semiconductors and cloud computing providers. This is no secret to anyone who does some research.

Ethereum remains at the forefront of an evolving Web3 and DeFi landscape, serving as the base layer for a majority of dapps. The price catalysts that AI brings to ETH will be confounded with a lot of other factors. Even before generative AI’s explosion of popularity, Web3 and DeFi were growing industries. One should consider the secular growth trend of Ethereum to be assisted or accelerated by AI.

Also, present day AI and decentralized systems are incompatible from a technical perspective. The details are outside the scope of this article. Be very skeptical of Web3 projects that claim to use AI in blockchain applications. Ethereum cannot feasibly “run AI,” despite its Turing completeness. Readers will quickly see that what I present here conceptualizes ETH as a beneficiary of AI through more roundabout methods rather than a direct merging between blockchain and AI as technologies.

What is AI and its Current State?

Artificial intelligence is a big mathematical model that is “weighted,” or “trained,” or “parameterized,” using real data so that the model can be used to make predictions when given new data. Where it gets extremely complicated, to the tune of being a multi-billion-dollar industry, is in the exact way these big mathematical models become parameterized. One way to imagine what is going on is with one of the simplest predictive models: Ordinary Least Squares (OLS) regression. The weights in OLS, or coefficients for the “line of best fit” are found by doing matrix algebra on a matrix representation of the dataset. Details here .

OLS is exceedingly simple because the weights are linear, and the variables are assumed to lack collinearity (ie. unrelated to each other). But what if we wanted more complicated and deeper insights? Many things in the real world are nonlinear and have self-referencing or interactive effects. The answer is that we need a different model, a different way of finding weights, and almost definitely a bigger dataset with more variables. AI is the most complicated and cutting-edge iteration of this journey to find deeper predictive insights.

So, there are a few important things we can learn from this. The first is that data is the lifeblood of AI . Without data, there is no AI because the mathematical model cannot be parametrized. This is like trying to find a line of best fit without the actual dataset to fit the line through: it’d be a meaningless task.

The second conclusion is that AI systems are limited to the scope of the data which has been used to train them . This is also where the AI prejudice or bias narrative comes from: minority groups may be misrepresented in AI training sets, which may cause unintended consequences in which conclusions drawn by AI inherit the biases in the data. A more generalized way to state this problem is that AI today exists as siloed systems . Each model had its own specific dataset and training method, which will impact their respective conclusions, even if they are given the same inputs.

The third thing to note is that the machines used to create models are not your average laptop. AI is normally trained in data centers such as those offered by AWS, Azure, or Google Cloud . This is also why big tech, especially cloud computing providers , are more of a pure play on AI. The fact that AI is dependent on specialized hardware contributes to them being siloed systems.

The fourth conclusion is that because AI is so dependent on specialized computing solutions, they will inevitably keep cloud computing’s pay-as-you-go model (PAYG) of revenue generation . Cloud computing has great operating leverage because the costs are relatively fixed (land and electricity) but the revenues are practically unbounded by physical limitations. Since AI uses cloud computing, it must adopt PAYG as well, capturing a spread between what AI users pay and what the cloud computing costs. (In practice, because much of AI is owned or developed by big tech, it may be more reasonable to look at AI not as a spread capture business, but rather a specialized instance of cloud computing.)

AI Will Catalyze Web3 Development

The most direct way AI will impact ETH’s price is by accelerating the pace of Web3 development. AI tools like GitHub Copilot and GPT-4 have become so proficient at writing code that they have dramatically increased the efficiency of developers everywhere. Many of my programmer friends have said that they aren’t even sure if they can code anymore without the assistance of these tools. Yet, they are coding much faster than before. To be clear, humans are still required for writing production level code, but AI shortens the development process by:

1. Providing autocompletion suggestions: It’s a bit like how a smartphone might suggest words when you are texting. You end up being able to turn your thoughts into words much faster.
2. Assisting with debugging and test cases: GPT-4 and even ChatGPT (which is GPT-3.5) are pretty good at identifying issues in code which lead to bugs. They can also generate code to test what has been written. Debugging and testing easily consumes the most time in the development process.
3. Auditing code and finding vulnerabilities: Web3 has a history of exploits, and AI can be used to improve the security of Web3 by identifying patterns which may introduce security flaws.

This not only speeds up the development process but also ensures that the dapps are more secure and robust. As a result, we can anticipate a surge in the number of high-quality dapps being launched, and given Ethereum's dominant position in the dapp space, it stands to benefit the most.

Many of these protocols will be built on Ethereum, which will drive enormous value to ETH. Most of the actual projects need not succeed for ETH to benefit. However, the existence of a very competitive landscape, further bolstered by AI, ensures that something will emerge which captures substantial market share. No matter what emerges, ETH gains if Ethereum is used as some kind of base layer for Web3. ETH is like the pick and shovels to a Web3 gold rush.

AI Can Give DeFi and Decentralized Trading a Boost

One of the micro trends in DeFi right now is Telegram trading bots . These enable users to place trades through their Telegram account. Another trend is social copy trading, in which one trader’s actions can be copied by many others using smart contracts.

AI stands to bring a lot more trading activity through automated on-chain strategies. These strategies can leverage AI to make intelligent decisions within crypto ecosystems. While the current utility of this approach may be primarily restricted to token speculation, it would greatly enhance the speculating experience. AI's ability to analyze vast amounts of data and make informed decisions is leading to more efficient and potentially profitable trading strategies. This increased activity and sophistication within the platform can attract more users and capital, further boosting Ethereum's prominence and contribute to an increase in the price of ETH.

Furthermore, AI can be used to understand the behavior of specific smart contracts. By analyzing the frequency, type, and outcomes of interactions with a particular contract, AI can provide insights into its utility, potential risks, or even detect anomalies that might indicate a bug or a potential exploit.

For traders and strategists in the DeFi space, this is invaluable. By leveraging AI-driven chain analysis, they can develop more informed strategies that consider not just market sentiment or global economic indicators, but the very flow of money on the Ethereum blockchain itself. This on-chain intelligence can lead to better risk management, more accurate speculation, and the identification of emerging opportunities before they become mainstream.

Note that the use of AI in this manner isn’t technically new. Quantitative funds have long used big data models to analyze vast amounts of data and execute trading strategies. What is new is that there has never been the plethora of data offered by public blockchains. The granularity of blockchains leaves nothing unanswered. AI can be trained to follow “smart money” wallets, or to make deep insights based on the visible flow of funds on-chain.

The AI model would not be on-chain, but smart contracts can be made to receive instructions from an off-chain node which hosts the AI. These setups will probably start popping up over the next year, and they could generate a lot of interest in crypto speculation.

AI trading insights based on real-time on-chain data will add to the existing tailwinds of real-world asset (RWA) RWA Tokenization: What Does it Mean to Tokenize Real-World Assets? and institutional adoption. As more RWAs enter Ethereum, it will lead to more trading opportunities. Similarly, as more dapps are developed, many will release their own protocol tokens, which also leads to more trading opportunities. On-chain trading means higher demand for gas, which directly increases the value of ETH.

AI Incentivizes the Adoption of Public Blockchains

This is tangentially related to the last point. Again, data is the lifeblood of AI. Public blockchains have data like nothing else. Every bit of transaction data is permanently recorded. Every interest rate, funding fee, swap, liquidation, buy, or sell is recorded. Although Web3 is commonly pitched as allowing users to “own their data,” the reality is that the current infrastructure of Web3, public blockchains, ensures that no has exclusive access to any raw data. This could be good or bad, but it is definitely good for those developing AI because they will have unlimited access to extremely comprehensive datasets.

Therefore, I would not be surprised if big tech begins to encourage the usage of programmable public blockchains, like Ethereum. The need for more data will make this both a logical and desirable course of action for them. Because crypto is heavily swayed by narratives, a narrative of established firms actively encouraging the use of Ethereum will be great for ETH.

Ethereum May be Needed for AI Interoperability

Remember, AI models are siloed: each model is trained on different data using different parametrization methods. Having models communicate with each other will become very important. Interoperability will become a major forefront of the industry to unlock the insights of multiple AI systems working together.

As mentioned, today’s AI operates on a PAYG or pay-per-use model. Users might be allocated a certain number of "credits" that can be used before incurring additional costs. This model is efficient in a Business-to-Consumer (B2C) context, as it typically involves just one hop between the AI provider and the consumer. However, as AI continues to evolve, the need for interoperability between different AI systems will become increasingly apparent, and this simple model may no longer suffice.

Unlike B2C, AI-to-AI could involve many hops. In truth, no one knows how a resilient system of interoperable AI might look like or interact. We do know that the industry has a nearly guaranteed likelihood of advancing toward AI interoperability. Why? Because like telecommunications networks, Metcalfe’s law should apply to AI networks. The value of a network scales quadratically with the number of interacting individuals in the network. However, AI will not have the lag times of these human networks. An online argument in an Internet chat forum could take weeks to settle down because people aren’t always looking at their computers. Unlike adults with nothing better to do than argue about politics with a stranger they’ll never meet, the AI models would be considering real issues, constrained only by the speed of light. Billions of back-and-forth’s between dozens of AI models might be settled in minutes. That would be a lot of hops.

Since each hop is subject to a PAYG model (due to cloud computing’s revenue model), interoperable AI systems require a payment and settlement layer that can handle digital payments at these speeds. This layer must not only be swift but also possess a high degree of programmable logic to address the intricacies and potential frictions that arise when different AI systems communicate. For example, what if one query within this multi-hop route is unexpectedly big that the machine takes a bit longer to process it? This could lead to undesirable race conditions .

Ethereum's smart contracts offer a solution. These self-executing contracts, with the terms of the agreement directly written into code, provide the necessary programmability and flexibility to facilitate seamless interactions between multiple AI systems. To be clear, it won’t be the Ethereum base layer that settles these. Rather, it would be a highly optimized L2 that supports nano-payments and high throughput. But ETH is still required to publish the state to Ethereum.

Traditional financial systems, with their inherent delays and potential for reversals, are ill-suited for the rapid and final settlements required for interoperable AI systems. On-chain settlements, on the other hand, offer both the speed and finality essential for these transactions. Once a transaction is confirmed on the blockchain, it is irreversible, ensuring that settlements between AI systems are both quick and final.

Also, a crucial aspect of this interoperable future is the assurance of fairness and transparency. AI providers need the ability to audit transactions and ensure that they are not underpaid for what they provide. Ethereum's transparent and immutable ledger provides this capability. Every transaction, every smart contract execution, is recorded on the blockchain, allowing for easy audits and verifications. This transparency ensures that all parties involved can trust the system and be confident in the fairness of their interactions.

In conclusion, as AI systems evolve and seek greater interoperability, the need for a reliable, fast, and transparent settlement layer becomes paramount. Ethereum, with its smart contracts and on-chain settlement capabilities, is poised to play a pivotal role in this next frontier of AI development. The integration of these technologies can further solidify Ethereum's position and increase the price of ETH.

ETH Valuation

To value ETH, we should conceptualize it as a mix between a security (with yield) and a money (with liquidity premium). Gas on Ethereum are paid in ETH and separated into a protocol-determined base fee and a user-determined priority fee. The base fee is “burned,” which means to be removed from the supply. The priority fee is kept by the validator and is used to incentivize the validator to include the transaction in the block. Validators must stake ETH to build blocks and earn priority fees. Another reward stakers/ validators earn is the block issuance, where about ~1700 ETH/day is added to the supply. This inflation is more than cancelled out by the burned base fees, such that ETH is deflationary.

It's easy to see that the cash flows received by ETH owners resemble dividends and buybacks – the traditional means by which cash is returned to shareholders. The distribution of these cash flows, however, are dependent on the actions that the ETH owner takes; it’s not as clear cut as simply buying ETH and holding it. For example, if you just held ETH without staking it, then the only cash flow you would receive comes from the net deflation of ETH supply due to the gas burn. This is like if you were an AAPL shareholder, but you were unable to receive any dividends from Apple Inc. – the only cash you would get comes from AAPL buybacks, which increase AAPL’s price.

Staking ETH makes your ETH position resemble an equity position – you benefit from both the dividend (from the gas that is distributed to stakers) and the buybacks (from the gas that is burned). Technically, stakers also benefit from a Cantillon effect where they get newly issued ETH from the block reward. Think of this as issuing shares exclusively to shareholders who are promising not to sell their shares.

At any given time, Ethereum’s “dividends” are evenly distributed to ETH stakers. Therefore, it is better for you as an individual investor if a very small amount of staked ETH (ideally, only your ETH are staked, so you get all the dividends Ethereum pays out). At the same time, if every single ETH was staked, the network would halt completely because no ETH can be spent on gas and the source of all yields will be gone.

This is where the money aspect comes into play, and where things get convoluted. Staked ETH creates the economic effect of an equity security. ETH that isn’t staked has the economic role as money. Money, being the most liquid good in an economy, of course gains a liquidity premium – people will either pay more or require less yield for liquidity. However, the existence of this premium is contingent on what the owner decides to do with the ETH: there is an embedded option. And whenever ETH is acting as money, ETH’s value comes from the demand for the economic services on Ethereum, which is also where the cash flow to ETH owner’s (both stakers and non-stakers) come from.

To make it even more complicated, Ethereum itself is like a mix between a company and an economy. Income flows to ETH stakers, but the means of generating this income comes from an open network in which anyone can participate and set up their own enterprise. For example, Uniswap ( UNI-USD ) is a decentralized exchange which causes the most burned gas on Ethereum. Uniswap also charges a trading fee, earning revenue for itself.

Thus, any valuation should evaluate the equity, the money, and the embedded option aspects of ETH. The market doesn’t seem to be pricing in AI’s impact because the YTD return of ETH has been almost perfectly correlated with BTC, albeit significantly lagging. The dominating crypto narrative this year seems to be Bitcoin, and by extension the rest of crypto, emerging as a legitimate alternative asset class and garnering institutional support. Crypto has gained notable strength especially after the banking scare in March.

Staked ETH, the most equity-like, receives all Ethereum fees as either a direct ETH distribution or as the gas burn. There a few numbers are out there for annualized fees. Token Terminal gives $2.2 billion as the annualized fees given the last 30 days of fees. Over the last year, there was $2.0 billion in fees.

But remember, staked ETH also exclusively receive the issuance of new ETH, and this is a cash flow which is not captured by fees alone. Ultrasound.money is a website which tracks ETH statistics in real time. Currently, the issuance yields about $1.4 billion a year.

Therefore, staked ETH receives an annual cash flow of about $3.5 billion. With a market cap of $223.5 billion, ETH’s price to cash flow ratio is about 64x. This is much more expensive than MSFT and GOOG, a bit more than TSLA, and much less than NVDA.

But this multiple is treating ETH as having a singular equity use case. It can also be used as money on Ethereum. And, since every ETH can be switched from staked to unstaked with relative ease, there is a perpetually embedded option which the owner can freely exercise to convert the use case from equity to money or back. Thus, ETH’s true multiple if we tried to make all things equal to equities would be noticeably lower. For instance, if we assumed the liquidity premium warranted us to pay twice as much, then the true multiple might be very close to that of MSFT, especially when considering the embedded option too.

When we add the potential for AI to catalyze an increase in the use of Ethereum, we see that there is strong fundamental upside for ETH. Right now, the multiple is priced like a big tech company. Even without AI, Ethereum has far higher growth potential than most tech companies because its open network allows for unbounded, free market innovation while a tech company will inevitably be constrained by its own operational economies of scale. Ethereum also stands to take enormous market share through Web3 – it is more like an early-stage company operating in a niche market that has a good chance of going mainstream, if we desire to keep using the company analogy. For now, Web2 is the dominant form of the Internet. Capturing a small fraction of this market share would mean huge upside for Web3, and by extension ETH. Also note that the higher expected growth alone warrants a higher multiple.

Now if we consider the AI possibilities outlined above, the bull case becomes even stronger. AI leading to better Web3 development will accelerate Ethereum’s already promising growth trend. In the last two years, smart contract deployments on Ethereum and Ethereum L2s have grown rapidly. And that was mostly without the help of the AI tools we have today. In the next few years, Web3 should see a lot more innovation heavily assisted by AI.

In the next few months, I wouldn’t be surprised to see projects that use AI for on-chain trading, which will bring a lot more activity and interest to crypto. And in the long run, there is a good chance that AI interoperability will require the use of Ethereum.

Given these AI catalysts, I can see fees doubling over the next few years, and I believe ETH’s multiple should be 30-60% higher than it is today, making ETH a buy for investors who can stomach some volatility. At 100x cash flows, ETH would still be much cheaper than NVDA, which is a true AI pure play. Given Ethereum’s unbounded potential to scale, I would say this is a fairer valuation than what it is right now.

Risks

There are many risks to this thesis. The main one is that ETH is not the only blockchain. It is possible to make something far more performant that attracts much more developer attention. This blockchain could be the new launchpad of Web3, and ETH would get crushed as a result. I think this outcome is highly unlikely because although Ethereum isn’t perfect, it has the most people working to perfect it. There is already a very strong network effect. Ethereum will probably be a long-term winner in Web3. The biggest threat in this aspect is Bitcoin. It is possible to build Turing complete layers on top of Bitcoin. One method is through merged-mining, which reuses Bitcoin proof-of-work’s expended work to validate transactions. Rootstock is such a Bitcoin merge-mined blockchain which runs the Ethereum Virtual Machine on its network. However, these alternative chains have very low activity compared to Ethereum.

The other risks relate to the AI assertions. It is possible that AI is not used to accelerate Web3 development. However, one would need to answer the question of what those developers would be doing instead. The reality is that there are only a handful of spaces in tech this decade that are particularly lucrative, including: AI, blockchain, Web3, quantum computing (which is mostly hardware today). If developers randomly picked one, there is a good chance they’ll land in Ethereum.

Another risk is that AI interoperability will not involve Ethereum at all, or only on a minor scale. For this, I don’t have a strong defense. However, Ethereum is the system that exists today which seems best suited to address the issues AI interoperability would likely run into. It is possible another public P2P network is created specifically for AI interoperability, but this would require a lot of separate profit-seeking parties to coordinate on building a new base layer when a good solution, Ethereum, already exists.

Conclusion

Overall, ETH is a buy given secular trends. It is even more so a buy given the boost that AI could quite feasibly give to Ethereum over the next few years. Don’t expect these results to come immediately. Some, like AI trading bots which follow on-chain flows, will likely appear very soon. Others, like AI-assisted Web3 development, will take longer. Finally, AI as an industry requires time to grow into a need for interoperability.

Editor’s Note: This article was submitted as part of Seeking Alpha’s Best AI Ideas investment competition, which runs through August 15. With cash prizes, this competition -- open to all contributors -- is one you don’t want to miss. If you are interested in becoming a contributor and taking part in the competition, click here to find out more and submit your article today!

For further details see: