After the AI Agent Bubble Bursts: What is the Real Value of Web3 AI?

In the fourth quarter of 2023, the AI agent sector rapidly rose, with its market value soaring from almost zero to over $20 billion. Various types of "intelligent agents" emerged, including humorous, peculiar ones, as well as "financial agents" that claim to trade automatically and help generate profits. There were even DAO organizations claiming to invest in other agents, as well as DAOs promoting "co-governance between humans and agents." The diversified innovations have sparked people's hopes for the possibility of getting rich overnight.

However, this wave of enthusiasm came quickly and left just as fast. After the bubble burst, many projects failed in succession. However, some AI infrastructure projects with practical value are steadily developing. Real value is beginning to emerge, and the next wave of the Web3 AI trend is brewing, which may not just be hype this time, and is worth our close attention.

We all know that whenever a new track or hot topic emerges (such as the combination of Web2 AI or cryptocurrency and AI), the market often pays little attention to the fundamentals. As long as the project looks exciting, has a lot of hype, and the demo is appealing, regardless of its actual utility, the market capitalization can easily exceed hundreds of millions.

Later, a project took a completely different route. It open-sourced AI, allowing any developer to easily get started and create value on their own. This concept quickly gained widespread recognition, the community grew rapidly, and the number of stars and forks on GitHub continued to rise.

The total valuation of a certain ecosystem once exceeded 5 billion USD, while another project's peak was also close to half of that scale. Other interesting AI agent projects also saw their market capitalization surpass 1 billion USD.

However, the market environment has changed dramatically now. The newly launched agent projects that are performing well mostly have a market capitalization between 3 million and 10 million; the market capitalization of older projects has also been compressed to the range of 10 million to 50 million. The valuation ceiling of the entire sector has been lowered, and the total market capitalization has decreased from the peak period of 20 billion to the current range of 4 to 6 billion.

The Rise of Infrastructure, the Acceleration of Web2 AI Development

The current market is no longer blindly chasing those "seemingly powerful" bubble projects, but is beginning to focus on the real fundamentals. Especially against the backdrop of the rapid development of AI models in Web2, everyone is paying more attention to the long-term value of infrastructure and decentralized AI.

AI models of large tech companies are almost updated every month, becoming more powerful, faster, and smarter. For example, a certain chat AI platform's newly launched image generation feature has caused a sensation as soon as it was released, with related style images rapidly flooding the screen.

The consumer products side of Web2 is also evolving rapidly. Due to the enhancement of underlying AI capabilities, many previously unattainable product experiences are now possible. New development tools have significantly improved developer efficiency, with frequent and numerous feature updates. AI agents and smart workflows have now permeated various fields, and the entry barrier is getting lower and lower. For users, switching tools incurs almost no cost; if a tool is poorly designed or expensive, it is easy to find alternatives with better interfaces and experiences. The overall market competition is becoming increasingly fierce, but it also accelerates the deployment of truly valuable products.

Awakening of Data Sovereignty: Who is the True Owner of Data?

Amidst all this rapid development, more and more people are beginning to realize a problem: there are various AI agent applications everywhere, but most of them use centralized technology. So, who really owns my data? Where will my chat records go? If I discuss some private content with AI, will it really keep it confidential? Or will it be uploaded, analyzed, and used to train other models?

This issue has become more prominent following a recent update from a certain AI company. The company's AI assistant can now reference all past conversations with users to generate more personalized responses. This feature is indeed cool; just imagine, in the future, everyone will have their own AI personal assistant, chat partner, emotional support... But this also means that users' data will be "long-term held" by a certain platform, and users are no longer the true owners of their data.

Once others have control over your conversations, preferences, emotions, and even your habits, the consequences may not just be as simple as "a better experience."

This is why the topic of "data sovereignty" is becoming the next focus of AI + Web3. Data that truly belongs to users is the most valuable future.

The Rise of Decentralized AI (DeAI)

Last year, I predicted that by the second quarter of 2025, decentralized AI would truly enter the public eye. Especially in the context of increasing public concern over privacy and data ownership, those underlying infrastructures that can provide confidentiality, verifiability, and transparency of user data ownership will gain more attention and usage.

Currently, we see trends forming in three main directions:

1. Venture Capital Trends in Web2 AI

• Some startups supported by a well-known incubator are launching vertical AI agents (specifically designed to solve problems in specific fields).
• A well-known venture capital firm has begun to layout the next phase of consumer AI product trends, presenting its own investment logic.
• An AI company has launched a fund specifically for investing in AI.

2. Venture Capital Trends in Web3 AI

• Start investing in decentralized AI infrastructure
• Distributed training networks, computing power sharing and other fields are gradually heating up

3. The Retail Trend of Web3 AI

• The AI Agent ecosystem is still one of the hotspots.
• Consumer-grade AI applications are gradually becoming more diverse, attempting to land from productivity tools to emotional companionship.
• Users are increasingly concerned about whether "the AI products I use are truly serving me, rather than harvesting my data"

These trends intertwine to drive DeAI from concept to practical phase. The year 2025 will be a pivotal moment for validating the value of decentralized AI.

Web2 vs Web3 AI: Completely Different Rhythms and Playstyles

In terms of Web2, the market size is much larger than that of Web3. This is because many traditional enterprises are seeking to transform through AI, optimizing their business processes, such as acquiring more customers, improving conversion rates, and increasing sales. These enterprises usually have clear needs, and many focus on specific niches, so they hope to find AI tools that can accurately address their "specific pain points." This has also attracted many young entrepreneurs to develop vertical AI agents targeting these specific needs.

Compared to traditional SaaS, the benefits brought by AI agents are more direct - either significantly reducing costs or directly attracting more customers to generate revenue. Therefore, the subscription prices for these AI tools can also sell for higher amounts, and many startups achieve annual revenues of millions or tens of millions of dollars within a few months of launching, which is not without reason.

But the gameplay of Web3 is completely different. The blockchain itself is a foundational layer tailored for decentralized AI (DeAI). All actions can be verified on the chain and are immutable; it naturally provides a trustless environment; supports decentralized computing; users can truly own their data, models, and use cases. In simple terms, the future of Web3 AI is to enable users to understand how their data is used, comprehend the decision-making process of AI, autonomously control models and application scenarios, and even profit from it.

Web3 venture capitalists have begun to position themselves for this future.

Why Retail Investors Love AI Agents

For Web3 retail investors, DeAI (Decentralized AI) is indeed quite difficult to understand: a bunch of new terms and concepts that sound like a foreign language. So initially, they are most easily attracted to those AI agents that are comprehensible and entertaining—like chatbots that can talk and tell jokes. These "entertainment AI agents" are indeed very appealing, but over time, retail investors begin to realize that these things seem to have little practical use. Moreover, with the recent downturn in the market, many useless projects are slowly being eliminated, while those agents that have practical value and can provide functionality, although their valuations have also decreased, are still surviving.

This wave of "cleansing" has made more and more people realize that only AI projects with practical applications and core product capabilities have a future. As a result, project parties have begun to shift towards two directions: either independently developing real AI products to solve practical problems, or collaborating with truly technical and valuable DeAI projects.

This transformation has two positive meanings: it prompts people to start paying attention to the underlying infrastructure that was originally "unintelligible"; it allows AI agents to no longer be mere performance tools, but products capable of solving real problems. Some projects have already become typical cases—not only powerful in functionality, but also bringing some advanced DeAI technologies into the public eye. This indicates a trend: although retail investors may not understand the technology, they will gradually be educated by "truly useful" products.

One of the most interesting aspects of a certain DeAI project is that it is a decentralized AI ecosystem that ordinary people can also invest in. Currently, most DeAI projects are still in their early stages, and only venture capitalists or "strategic partners in the circle" can invest, with many not even having issued tokens yet. But this project is different. Users can directly vote with their tokens to support the subnets they are optimistic about, equivalent to "switching positions" into the sub-tokens of these DeAI projects, allowing for early positioning.

Although there have been complaints before about cross-chain bridges and the trading experience being somewhat cumbersome... their underlying technology, product logic, and overall atmosphere are indeed very strong. Especially with the existence of a certain laboratory, the entire ecosystem's UX/UI design is developing towards a "user-friendly" direction. Because in the mechanism of this project, every subnet must rely on market recognition to gain more rewards (mining incentives) — those who are useful and strong can receive more distribution.

Therefore, for these subnets, "making users understand what you are doing" becomes crucial. A certain lab is doing this. Their product direction is very clear: optimizing UI/UX for ordinary users. They not only have multiple practical subnets (for example, a super convenient AutoML platform where users can train models directly with just a few clicks), but their latest flagship product is also very cool: an AI Agent platform where you can drag and drop modules to create AI agents, truly achieving "zero-code building of AI agents." This experience is somewhat like a "foolproof AI factory" for Web3, making it very suitable for users who are not technically inclined to get started.

Overall, this ecosystem is now not only one of the most technologically advanced DeAI projects, but it is also at the forefront of user-friendly participation for regular users. The clear product logic and user-friendly team are the key roles that make this ecosystem popular.

We are in a transformative era dominated by Web3 AI. The past bubble of inflated market values driven by hype has been replaced by actual infrastructure, decentralized AI, and real application scenarios. Whether it's enterprises optimizing their business with AI in Web2 or retail investors experiencing the convenience of new agents in Web3, future data sovereignty and user participation will be key. Web3 AI is far from its peak. The real show has just begun.