Author: Teng Yan
Compiled by: TechFlow
I often tell my friends that when we look back on the period from 2022 to 2024, we will find that this was a critical time of rapid technological acceleration for humanity.
AI may be the most transformative technological trend we encounter in our lifetime, unless we discover a miracle that extends our lifespan by hundreds of years.
This means that AI is now a hot topic, and everyone wants to be involved.
In just the first half of 2024, AI startups attracted $35 billion in investment. And this is just the investment in the private sector - the internal investments of the big tech companies are even greater, as they are purchasing massive amounts of GPUs from NVIDIA, driving NVIDIA's market capitalization to an astounding $3 trillion.
However, in this frenzy, there is a huge untapped potential area that many people are overlooking: Crypto AI (or Decentralized AI).
As early as 2019, a forward-looking cartoon hinted at this.
History has shown that every decade there are investment opportunities that seem highly unlikely or even foolish, but ultimately prove to be incredibly prescient.
Social media was once seen as just a trivial pastime for teenagers, with no real business model. Yet today, Meta (formerly Facebook) has become one of the most influential companies in the world, with early investors achieving over 1000x returns.
The story of Crypto AI is both urgent and compelling. When I explain it to others, most people can immediately grasp the concept.
The essence of AI is the centralization of power. If left unchecked, it could concentrate power in the hands of a few monopolistic organizations, which would inevitably use AI to profit and pursue dominance. Therefore, decentralized AI is crucial to our future and is key to achieving a brighter, more equitable society. I have previously delved into this philosophical issue.
However, skeptics argue that combining crypto with AI is just hype, pointing out that crypto has yet to have a lasting impact or widespread adoption in entertainment, gaming, and social media. I've even heard this concern from some savvy investors, which is a reasonable one.
But I believe this time will be different.
There are several reasons why Crypto AI will take a markedly different path. I'm writing this article to elucidate these reasons.
The content I want to cover is more than I initially envisioned, so I've decided to break it into several parts.
In this three-part series, I will delve into the technology and investment landscape of Crypto AI. I will highlight the most promising areas and show how I position myself to capitalize on this emerging megatrend.
Part 1: Why Crypto AI is a Focus Area
Part 2: My Thoughts on AI Agents, Decentralized Training, Verifiable Reasoning, Data Networks (and other Crypto AI Subdomains)
Part 3: Multiple Ways to Extract Value from this Opportunity
As the world becomes increasingly AI-driven, we need more crypto technology to constrain it and establish price and protective barriers for our most precious and irreplaceable resource - time. —— Preston Bryne
At the Intersection of Tech Trends
As savvy investors and entrepreneurs, we are always seeking to ride the waves of transformation.
To catch the biggest waves, we need to find the places where multiple major trends intersect, and position ourselves there.
This means identifying the deep behavioral shifts driven by technological progress that can redefine entire industries - the long-term tech trends.
Crypto is one such tech trend, transforming the way we think about and use money. Other examples include cloud computing, mobile technology, and clean energy.
However, simply following a single tech trend is not enough. There are countless others who have seen the same trend and made similar investments. To truly stand out, we need to go beyond the surface-level obvious.
This is why discovering the early convergence of two major tech trends is so compelling.
This is where the magic happens.
Convergence + Scale = Opportunity
( 1) The Power of Convergence
When two long-term trends intersect, it creates the best opportunities for innovation and value creation in often overlooked areas.
Multiple Growth Drivers: Companies at the intersection of multiple long-term trends can benefit from the growth of each trend.
Reduced Competition: Due to the need for deep expertise in multiple domains, these companies can occupy unique positions in markets with high barriers to entry.
Cross-Pollination of Ideas: Enabling innovative products and business models.
( 2) The Power of Scale
In investing, market size is crucial.
Think about Amazon in the early 2000s. Its success was not just due to e-commerce, but also because it seized the emerging cloud computing trend and launched Amazon Web Services (AWS). AWS now generates billions in revenue annually and is the leader in the rapidly expanding cloud infrastructure market.
The larger the long-term tech trends, whether in terms of total addressable market or growth potential, the more opportunities there are at their intersection, especially if you get in early.
Big trends not only provide a buffer against failure, but can also greatly amplify potential returns.
Crypto x Artificial Intelligence
CPUs and GPUs have long been the pillars of computing, and now they are powering AI, which is becoming a global supercomputer, harnessing the collective intelligence and creativity of humanity.
Crypto technology, on the other hand, has made it possible to create open, decentralized networks, laying the foundation for the next generation of the internet.
Together, these supercomputers and supernetworks complement each other perfectly.
AI enhances the user experience of crypto. Users can interact with blockchains through natural language, without needing to learn how to manage wallets, seed phrases, and sign transactions.
Crypto provides the trustless, permissionless, and secure infrastructure for AI to ensure its openness and censorship resistance. It also serves as a powerful coordination layer for building decentralized networks.
This convergence paves the way for entirely new business models.
The combination of crypto and AI will drive exponential growth, as the two fields continue to advance along different trajectories.
The key is to identify the specific ways in which crypto can help AI achieve functionalities that were previously impossible - this is where the magic lies. Hint: Distributed training, data networks, and privacy-preserving data. More on this in Part 2.
Learning from History: The Cases of NFTs and DeFi
The floor price of BAYC NFTs once skyrocketed. Yuga Labs raised $450 million in funding in 2022 (data from Coingecko).
The rise and fall of NFTs provide important lessons.
NFTs shone brightly in the crypto community's speculative frenzy, but without the support of another long-term tech trend, they struggled to develop further. Entertainment and gaming - the primary use cases for NFTs - are complex and mature markets, with strong traditional players, and their growth momentum is not solely dependent on technology.
Therefore, Non-Fungible Tokens (NFTs) have been unable to maintain their early growth momentum. Although their application scenarios are real, it will take longer to realize their potential.
In contrast, Decentralized Finance (DeFi) is a successful example of the convergence of long-term technological trends.
DeFi has revolutionized financial services by combining financial technology and cryptographic technology, providing alternative solutions to traditional banking, lending, and asset management, and meeting the real-world financial needs. Currently, the total market capitalization of stablecoins has reached a historic high ($170 billion) and continues to grow, while $82 billion has been locked in DeFi protocols.
Token: Open-Source AI Needs Cryptography
The closed nature of large tech companies' large language models hinders the realization of "AI democracy". Every developer or user should be able to contribute algorithms and data to large language models and share in the future returns of the models. AI should be accessible to all, relevant to everyone, and collectively owned. —— Catrina Wang (Portal Ventures)
When people ask me why AI needs cryptography, my answer is simple: Token.
Traditional software can be scaled at almost zero cost. Once the code is written, it can be deployed anywhere.
However, AI is completely different. It requires high capital and marginal costs.
Training and deploying large-scale AI models requires massive computing resources, so efficiency and accessibility of infrastructure are key factors for success.
Currently, we live in a world dominated by centralized giants - such as OpenAI, Anthropic, Google - who have abundant talent, hardware, and capital. But let's be honest: corporate-owned AI is always aimed at maximizing profits.
Although Meta's contributions to open-source AI are invaluable, who can guarantee that they won't stop releasing top-tier models like Llama 3? Developing these systems costs hundreds of millions of dollars, and if Zuck is in a bad mood one day, the project may be shut down.
Relying solely on ideals or goodwill to expect the open-source movement to compete with these giants is frankly unrealistic. We need a new strategy.
In fact, outside the top AI research labs, there is a wealth of underutilized computing resources, research, and talent in the open-source world. This includes universities, research centers, collaborative platforms like Hugging Face, and contributions from individual AI researchers. However, these resources are currently fragmented, lacking the necessary coordination to achieve large-scale breakthroughs.
This is where Tokens shine.
As @long_solitude stated in Zee Prime, Tokens embody the most powerful feature of cryptography - permissionless capital formation.
Tokens can do things that traditional financing models cannot:
They provide bottom-up financing opportunities for experimental AI projects that may never see the light of day under traditional venture capital models.
They can kickstart decentralized networks, transforming them into thriving and self-sustaining ecosystems by distributing ownership and value to contributors. Bittensor is an early proof-of-concept example.
They create entirely new business models through DePIN-style token economics, significantly reducing costs. For example, shifting the massive upfront infrastructure costs faced by AI startups to the network itself; or leveraging the idle GPU resources in global households through a GPU marketplace to reduce the cost per unit of computation.
Tokens provide a tremendous opportunity for ordinary investors to participate in the AI wave, which is often underestimated.
Cryptography excels at discovering new markets that people are willing to trade in. Look at Non-Fungible Tokens (NFTs) and cultural assets, creator economy social Tokens, and the virally spreading Memecoins.
I believe the demand for participation from ordinary investors in early AI projects is huge and largely untapped.
As AI becomes more deeply integrated into our daily lives and enables us to do things we couldn't before, people are starting to realize its scale and real-world significance. While there will be a lot of speculation, this is also about letting everyone participate in the biggest technological revolution of our lifetime - and giving everyone a chance to catch the next big opportunity.
More exciting developments are expected in the future.
So, Why Now?
Source: Syncracy Capital
New technologies often follow a clear innovation cycle.
The Gartner Hype Cycle is one of the most well-known frameworks. It depicts how innovations go through a hype stage, plunge into the trough of disillusionment, and eventually reach practical application.
For investors, the ideal investment timing is either at the discovery of a new innovation trigger, just before the hype peak; or in the trough of disillusionment, when one can identify the startups that are about to mature.
So, the million-dollar question is: Where are we in the innovation cycle of Crypto AI?
I like to use the chart from Syncracy Capital to illustrate the current consensus.
The chart shows that Decentralized AI is approaching or has already reached the peak of expectations. Crypto AI has performed strongly this year, with the valuations of multiple protocols reaching billions of dollars.
But I don't agree with this view. I believe the peak of Crypto AI is still far away.
The current consensus underestimates the tremendous potential of this field in the future. We haven't even entered the frenzy stage yet. Ask around, how many people truly understand Bittensor, which is a key indicator of Crypto AI? In fact, I believe it may take 1 to 2 years to see the real peak.
The reasons are as follows:
1. The total market capitalization of Crypto AI Tokens is $30 billion, only 2.9% of the total Altcoin market ($1.04 trillion).
The largest Crypto AI protocols - TAO, NEAR, FET, and ICP - have market capitalizations ranging from $5 billion to $10 billion, only four of them.
Excluding these four, and considering that ICP and NEAR are not dedicated Crypto AI Tokens, the total market capitalization of Crypto AI is only $11.7 billion. This is not even 25% of the Memecoin market value.
Additionally, only four Crypto AI projects (RENDER, GRT, AKT, AIOZ) have valuations between $500 million and $5 billion, and most projects have market capitalizations below $100-200 million.
Given the enormous potential of this trend, such market capitalizations seem negligible. Crypto AI covers both infrastructure and applications, including next-generation smart contract platforms designed for AI.
In comparison, the current total market capitalization of smart contract platforms is close to $600 billion. There are 8 Layer 1 protocols with market capitalizations over $10 billion, and another 12 between $1 billion and $10 billion.
How big could the Crypto AI market be? It's still too early to predict accurately.
According to Bloomberg Intelligence, the generative AI market is expected to expand at over 30% annual growth rate, reaching $1.3 trillion by 2032. If decentralized AI can capture 10% of the entire AI market, and considering the typical 3x speculative premium in the crypto market, this suggests a market size of $390 billion by 2032 - a 13x increase from the current $30 billion.
I intuitively feel this forecast is too conservative, and the time frame is too long to be a practical reference.
Another perspective is that, assuming that within the next 3 years (by 2027), Crypto AI occupies 10% of the total market capitalization of the Altcoin market, as AI applications and smart contract platforms are launched and gain momentum. If by then the Altcoin market capitalization reaches $2.7 trillion (a 50% increase from the 2021 peak of $1.8 trillion), then the market capitalization of Crypto AI will reach $270 billion - about a 9-fold increase from the current level. This means that there is $240 billion in potential value waiting to be unlocked.
But these numbers are more for illustrative purposes rather than precise forecasts. Because there are too many variables. However, they do show the vast opportunity and provide a rational reference point as we consider valuations.
2. Top Crypto AI teams are just getting started
Many high-caliber teams have been working on this for 1 to 2 years, and haven't even launched their products to the mainnet yet.
Some of these teams have already raised tens of millions in venture capital, including Sentient, Sahara, Vana, Story Protocol, Gensyn, Space and Time, Ritual, and Nillion. In the next 12 months, we will see the launch of major mainnets and Token issuances - such as the AO Computer ecosystem, which is a project worth keeping an eye on, and one I mentioned earlier this year.
3. AI is advancing rapidly
For example, OpenAI's recently released o1 model has made significant strides in reasoning capabilities. The scaling law still holds, and Crypto AI will closely follow the broader AI growth trend.
Nevertheless, there is a lot of noise in the current market, perhaps more than any other crypto domain. Many startups and protocols, even if they achieve some short-term success, are bound to fail.
Therefore, selectively identifying potential winners may be more effective than a broad net casting strategy.
Positive Catalysts for 2025
I expect multiple favorable factors to emerge in the next year, driving the development of the Crypto AI industry and related market expectations.
Traditional tech VCs entering Crypto AI: While a16z is already involved, major investors like Sequoia, Lightspeed, and Accel have only dipped their toes in this area so far. As they increase their investment in decentralized AI, it will bring more capital and legitimacy to the field.
OpenAI IPO: OpenAI's private valuation has reached $150 billion, and its revenue continues to grow. If it goes public (IPO) in 2025, it may ignite retail enthusiasm for AI investments. Apart from NVIDIA and the hardware space, retail investors have had few opportunities to directly benefit from the AI trend. This pent-up demand may flow into Crypto AI Tokens, which can be easily accessed on platforms like Coinbase.
More crypto-friendly US government: If the US government adopts a more favorable policy towards crypto Tokens after the November 2024 elections, it would provide a significant boost to the Crypto AI industry, offering a clearer regulatory environment and encouraging wider adoption.
AI breakthroughs: There are no signs of a slowdown in the rapid progress of AI technology. In a recent blog post, Sam Altman described his optimistic view of the bright future of AI. Advancements in AI agents and distributed AI training will unlock new use cases and drive further development.
Risks for Crypto AI
While I am optimistic about the immense potential of Crypto AI, I also recognize that nothing is certain. If the following risk factors materialize, my view may change:
Unfriendly regulatory environment: In major markets like the US, if the crypto regulatory environment becomes unfavorable, it could stifle innovation, limit capital inflows, and push crypto projects into a gray area. Tokens are the core of decentralized networks, and strict restrictions will undermine the key value drivers of crypto. If Trump is elected, this scenario is less likely, but if Kamala becomes president, the probability is higher.
AI failing to deliver on promises: Despite the current market enthusiasm and billions of dollars in investment, AI may fail to live up to its promises. If progress stalls or encounters roadblocks, the AI bubble could burst. I consider this scenario unlikely.
Inability to find a large market and product-market fit (PMF): No matter how innovative, Crypto AI projects must find viable business models and product-market fit. If decentralized AI fails to integrate meaningful business applications, the industry may stagnate. In this case, Crypto AI may become a niche domain.
Talent shortage: The number of top-tier machine learning scientists and engineers is limited. If Crypto AI is unable to attract enough of the best talent who believe in the vision of an open, democratized AI future, innovation will slow, and the industry will struggle to remain competitive.
Conclusion
Ironically, AI may be the best opportunity the crypto space has seen so far.
It paves the way for true mainstream applications and real-world use cases, which gaming, NFTs, and social apps have failed to achieve.
We are heading towards a decentralized AI future driven by open and public networks.
The most visionary founders and investors have already noticed this trend.
In the Crypto AI space, the biggest challenge is that you cannot just focus on the crypto domain, or you will have a very narrow and superficial understanding of the future. You need to keep up with the latest developments in machine learning, dive deep into the latest Arxiv papers, and engage with founders who believe they are building the next big breakthrough in AI.
To be honest, I have never been more excited.
