Blockchain Capital: Lessons from the Crypto Market, AI Investors Should Revisit

Author: @jonah_b, Researcher at Blockchain Capital

Compiled by: Jiahua, ChainCatcher

It is hard not to notice the striking similarities between the current AI frenzy and previous waves of enthusiasm in the cryptocurrency market.

The cryptocurrency market is a suitable place to observe how people act in the face of a significant technological wave: there are positive aspects, negative aspects, and the most unbearable aspects. This is because the cycles in the cryptocurrency market advance quickly, allowing early projects to gain liquidity that other markets struggle to provide, and market behavior is publicly accessible by default, as blockchain data is inherently open.

We have learned a lot from this market, and these experiences are equally applicable to AI. If you are investing in AI, this article is for you.

Lottery-style Betting

Huge successes trigger follow-the-leader behavior and create FOMO. Once a type of asset produces a game-changing winner, investors rush in, trying to replicate its success path. However, the second wave of projects rarely reaches the heights of the first wave, and many projects ultimately turn out to be castles in the air, resulting in significant losses of capital.

Bitcoin became a trillion-dollar asset. Subsequently, Ethereum became a hundred-billion-dollar asset, and Solana also became a multi-billion-dollar asset, proving that this market can produce more than one giant winner.

Thus, the craze for investing in new public chains arose. Venture capital firms treated these projects as lottery tickets; hitting on just one could yield multiples of the entire fund size.

Today, many emerging AI labs' valuations are also built on lottery-style expectations.

OpenAI and Anthropic are both moving towards trillion-dollar valuations. The formula behind this is simple: first, there is an extremely popular market, then a proven successful follower emerges. Thus, every new entrant is seen as the next lottery ticket to wealth.

In the past, many Layer 1 projects achieved valuations of billions of dollars based almost solely on a white paper and a founding team. The stories they told were also very enticing:

What if the global economy operated on our chain?

Similarly, emerging AI labs have raised billions based on a set of research ideas and founding teams poached from OpenAI, Anthropic, or Google DeepMind.

What if they really could create the "God of Machines"?

However, many times, this investment merely bets on the next higher-priced buyer appearing.

Many early investors are not necessarily evaluating the company's current fundamentals and future prospects. They know that the addition of a star talent or a partnership with a massive cloud provider could attract investors to push the company's valuation higher once again. Coupled with the increasing liquidity in the secondary market, they assume that the next buyer will always appear.

Market Chaos

As BCAP GP @CremeDeLaCrypto mentioned in the Bankless program, when large amounts of capital flow into the market, there will always be "a group of speculators and scammers chasing quick money, rushing wherever it is hot."

For over a decade, cryptocurrency investors have witnessed this phenomenon firsthand: batch after batch of projects claiming "tokens are the product" have been pushed to market under dubious market-making practices, investor-unfriendly high FDV/low circulation structures, unfavorable SAFT agreements, and a plethora of other market tricks.

Now, the AI industry is playing out a similar scene. For example, those structures composed of three layers of SPVs, charging exorbitant fees…

However, there is a key difference between the two: in the cryptocurrency market, prices are public, and tokens can be traded openly. The prices and valuations of AI companies are formed in opaque, illiquid secondary markets.

Even without discussing speculative behavior, AI investors can see from the cryptocurrency market that changes in market structure will determine where profits ultimately flow.

For instance, the heavy bets AI investors are placing now may ultimately turn into a standardized commodity.

Case from the Cryptocurrency Market: Block Space Became a Commodity

Block space was once scarce and expensive, leading capital to flood in, attempting to expand supply.

However, the industry later went a bit overboard: more and more Layer 1 chains launched, and Ethereum also added Layer 2. Ultimately, block space transitioned from scarcity to abundance, even excess.

This is good for technological development, but not necessarily for investors. Today, many alternative Layer 1s still generate very limited revenue.

The early market generally bet on the "fat protocol" theory, but ultimately the "fat application" theory prevailed.

As block space became cheaper, users paid less for the underlying infrastructure, while spending more on upper-layer applications. Applications like Tether, Hyperliquid, Aave, and Polymarket ended up capturing the bulk of the revenue.

Case from AI: Models May Become Commoditized

AI may be replaying the early experiences of the cryptocurrency market.

Chinese AI labs are continuously making increasingly powerful model weights public. I have provided an explanation regarding the incentives behind this.

If model weights become standardized commodities, and model prices continue to decline, value will shift upstream and downstream in the tech stack.

Applications will become the primary beneficiaries: if the marginal cost of using models gradually approaches the marginal cost of running models, applications will no longer have to pay high profits at the model layer.

OpenAI and Anthropic may not be directly impacted by this change, as they already have user bases, enterprise customer relationships, and distribution channels aimed at developers. But they are exceptions.

In this sense, they are more like the Hyperliquid of the AI field, rather than just providing underlying infrastructure like Layer 1.

In other words, OpenAI and Anthropic have achieved vertical integration. Other AI labs that cannot directly reach end users may find themselves in a more difficult position.

Another result of declining model profit margins is that the energy and hardware at the bottom of the tech stack have the opportunity to achieve higher profit margins. Especially when the supply of computing power is constrained by physical conditions, this trend may become more pronounced.

In short, the hardware layer and application layer may take away more profits, while the model layer will be squeezed. Yet, a significant amount of investment capital is flowing precisely into the model layer.

None of This Is New

From a broader cyclical perspective, this is actually not surprising.

Carlota Perez, in "Technological Revolutions and Financial Capital," argues that such technological revolutions go through several stages: infrastructure laying period, frenzy period, crash period, and deployment period.

Financial capital tends to over-invest in infrastructure during the frenzy period, but these overbuilt infrastructures, now priced low, will support the development of the next generation of applications.

This helps explain why there was overbuilding of block space in the cryptocurrency market. The expansion of models in the AI field may become the next case.

Investing in emerging AI labs requires believing that they can yield huge returns on R&D investments. The history of the cryptocurrency market and alternative Layer 1s should at least make us question this assumption.

Of course, if AGI emerges, the above judgments may not hold. Because we have no idea what the economy will look like after the arrival of AGI.