The market capitalization has exceeded 85.4 billion US dollars. Although we have not yet fully realized a “ownership network,” we have already seen the vibrant innovation that cryptocurrencies have brought to the current internet. For example, stablecoins such as Tether (USDT) and Circle (UDSC) are quietly changing the landscape of the global payment network. According to Coinbase’s research report, stablecoins have become the fastest-growing payment method. Stripe recently completed the acquisition of the stablecoin infrastructure project Bridge, with a value of up to 1.1 billion US dollars, making it the largest acquisition in the crypto world.
Blackbird, founded by the co-founder of Resy, focuses on changing the dining experience by allowing customers to pay for meals with cryptocurrencies, especially using their own token $FLY. This platform aims to connect restaurants and consumers through a cryptocurrency-driven application, while also serving as a loyalty program.
Worldcoin, co-founded by Sam Altman, is a pioneering movement to promote universal basic income, relying on zero-knowledge proof technology. Users scan their irises through a device called Orb, which generates a unique identifier called “IrisHash” to ensure that each participant is a unique human, thereby combating the growth of false identities and robot accounts in the digital space. Worldcoin has more than 10 million participants worldwide.
If we go back to the summer of 2017, we probably wouldn’t have imagined what the next seven years would mean for the crypto industry – we wouldn’t have expected so many applications to grow on the blockchain, or for hundreds of billions of assets to be stored in smart contracts.
How AI reflects on cryptocurrencies
Now I want to discuss the similarities and differences between cryptocurrencies and AI. After all, many people often compare these two. If we compare cryptocurrencies to AI, it’s like comparing apples and oranges.
But if we look at today’s AI investments from the perspective of cryptocurrency investors, we might find some similarities: both are comprehensive technologies, each with its own infrastructure layer and application layer.
But the confusion is also similar: it is still unclear which layer will accumulate the most value, the infrastructure layer or the application layer?
“What if the headline does what you want to do?” – this may be a nightmare for all entrepreneurs. The history of internet development in the past has proven that this nightmare is not groundless, from Facebook and Zynga breaking up and making their own mobile games, to later Twitter’s live streaming and Meerkat, the resource advantage of large companies makes it difficult for startups to compete.
In the crypto industry, because the economic models of the protocol layer and the application layer are different, the focus of each project is not to do every layer in the ecosystem. Take public chains (ETH, Sol, etc.) as an example, the economic model determines that the more people use the network, the higher the gas income and the higher the token value. Therefore, the top projects in the crypto world spend most of their efforts on ecosystem construction and attracting developers. Only the emergence of a popular application will increase the usage of the underlying public chain, thereby increasing the project’s market value. Early infrastructure projects may even provide subsidies ranging from tens of thousands to millions of dollars to eligible application developers.
Our observation is that the value capture of the infrastructure and application layers is difficult to distinguish, but for capital, both the infrastructure and application layers can be equally successful. For example, a large amount of capital flowing into public chains has improved the performance of leading public chain projects, giving rise to new application models and eliminating less popular public chains. Capital flowing into new business models, coupled with user growth, has resulted in higher requirements for underlying infrastructure, driving infrastructure upgrades.
So, what can be used as a reference for investment? The simple truth is that investing in both infrastructure and application layers is not wrong, but the key is to find that top player.
Let’s fast forward to 2024 and see which public chains have survived. Here are three simple conclusions:
Disruptive technology does not play a major role in project success. “Ethereum killers” that were previously sought after by Chinese and American VCs and focused on professor and academic concepts (such as Thunder Core, Oasis Labs, Algorand, etc.) only Avalanche ultimately emerged, and it did so under the premise of professors leaving and fully compatible with the Ethereum ecosystem.
On the other hand, Polygon, which was not favored by investors because of its lack of innovation (forking ETH), has become one of the top five ecosystems in terms of on-chain assets and users.
Unfortunately, projects like Near Protocol, which focuses on sharding technology and can outperform Ethereum in TPS, with nearly $400 million in funding, now only have on-chain assets of about $60 million.
Of course, the numbers fluctuate daily with market conditions, but the trend is quite clear.
Developer and user stickiness comes from the ecosystem. For public chains, users include not only end users but also developers (miners are a completely different model, so we will ignore them here). For end users, the ecosystem with rich applications and more trading opportunities will have more stickiness. For developers, the ecosystem with more users and better infrastructure, such as complete wallets, block explorers, and decentralized exchanges, will be given priority for development. Overall, this creates a flywheel effect where developers and users drive each other.
The network effect of the top players is greater than imagined. The number of Ethereum users and the amount of funds in on-chain applications is greater than the sum of all “Ethereum killers.” Everyone (especially those outside the industry) thinks of Ethereum first when they think of smart contract chains, just like how people think of OpenAI when they think of AGI today – it has almost become an industry standard for developing blockchain applications. In addition, existing top public chains already have a large amount of cash that can be invested or donated to developers, which new startups cannot achieve. Lastly, because most blockchain projects are open-source, mature ecosystems provide more possibilities for decentralized application building blocks.
So, what are the significant differences between public chains and large models?
Requirements for infrastructure. According to a16z’s statistics, 80-90% of early-stage funding for most AI startups is spent on cloud services. AI application companies spend an average of 20-40% of their revenue on fine-tuning costs for each customer. In simple terms, the money is earned by Nvidia and AWS/Azure/Google Cloud. Although public chains also have mining rewards, the cost of hardware/cloud is borne by decentralized miners, and the scale of data currently processed by blockchains is still negligible compared to the billions of data labels required by AI models. Therefore, the cost of infrastructure is much smaller than that of large models.
Liquidity, liquidity, liquidity. Public chains that have not launched their mainnets can issue tokens, but AI model companies without users and revenue find it difficult to go public. Therefore, although various “professor chains” may not ultimately perform as expected (after all, Ethereum is still deservedly No. 1), from an investor’s perspective, they are not likely to lose money and are unlikely to go to zero. Large model companies, on the other hand, are different – if they can’t raise the next round of funding and there are no white knights, they are easily doomed. From this perspective, venture capitalists should be more cautious.
Actual productivity improvement. Through ChatGPT, LLM has found its product-market fit and has truly begun to be widely used by B2B and B2C, improving productivity.
Although public chains have experienced two bull and bear cycles, they still lack a killer app, and application scenarios are still in the exploration stage.
Perception by end users. Public chains and end users are strongly related – if users want to use decentralized applications, they must know which public chain it is on and then make an effort to move their assets to that public chain, thus forming a certain stickiness. On the other hand, AI is more silent, like cloud services and processors inside a computer. No one cares whether the ride-hailing software is backed by AWS or Aliyun. Because ChatGPT’s memory is very short, no one cares whether they are chatting with ChatGPT on the homepage or on an aggregator. Therefore, it is more difficult to retain C-end users.
As for the application scenarios of cryptocurrencies in AI, many teams have given their own insights, and it is generally believed that decentralized financial networks will become the default financial transaction networks for AI agents. I think the following figure accurately summarizes the current stage.
In order to find a needle in a haystack, my former boss, who made a lot of money in the crypto industry, once said, “You still have to work hard, otherwise, you will become a rich ordinary person.” I think there is a respected investor who described the work of VCs as “finding needles in haystacks.” For me, VC investment in the crypto world is also such a process. The only difference is that the haystack in the crypto world may be moving faster. So we must always stay agile.
The author of this article is JW (@bestmosquito), the founder of Impa Ventures. Impa Ventures is a fund focused on early-stage investments in the Web3 industry.
The article is a collaboration reprint from: DeepFlow.
