After the Burst of the AI Agent Bubble: What is the True Value of Web3 AI?

In the fourth quarter of 2023, the AI agent track suddenly exploded, with the market value skyrocketing from nearly zero to over 20 billion USD. Various "agent" projects emerged, ranging from humorous to quirky. The market also saw the appearance of "financial agents" that automatically trade cryptocurrencies, DAOs investing in other agents, and even organizations for "co-governance between humans and agents." In no time, the variety of gameplay proliferated, and people's fantasies of getting rich overnight were reignited.

However, trends come and go quickly. After the bubble burst, many projects collapsed one after another. Nevertheless, some practical infrastructure AI projects have begun to emerge. True value is starting to surface, and the next wave of Web3 AI is brewing. This time, it may not just be hype, and it's worth our close attention.

We all know that whenever a new track or hot topic emerges, the market often doesn't care about the fundamentals. As long as the project looks lively, has a gimmick, and a nice demo, regardless of its actual utility, the market cap can easily reach hundreds of millions.

In this wave, some projects are good at telling stories, accurately grasping the market and occupying users' minds, with outstanding narrative ability. As a result, developers have launched projects on their platforms, and retail investors have followed suit to speculate.

Later, a project emerged that took a completely different approach. It open-sourced AI, allowing any developer to easily get started and create value on their own. This concept quickly resonated widely, and the community grew rapidly, with a dramatic increase in the number of stars and forks on GitHub.

The total valuation of certain ecosystems once broke through 5 billion USD, while some other interesting AI agent projects reached a market value of 1 billion USD.

However, the current market environment has undergone significant changes. Newly launched agent projects that perform well mostly have a market value between 3 million and 10 million dollars; the market value 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, with the total market value dropping from its peak of 20 billion dollars to the current range of 4 to 6 billion.

The Rise of Infrastructure, Accelerating Development of Web2 AI

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

AI models launched by major tech companies are updated almost every month, becoming stronger, faster, and smarter. For example, a recently launched image generation feature by a certain chat AI platform quickly sparked a craze for "Ghibli-style" images as soon as it went live, rapidly trending on social media.

The consumer product side of Web2 is also evolving rapidly. With the enhancement of underlying AI capabilities, many product experiences that were previously unattainable are now possible. Emerging AI tools have significantly improved developer efficiency, with frequent and numerous feature updates. AI agents and intelligent workflows have permeated every corner, and the entry barrier is getting lower. For users, switching tools involves almost no cost—if something is not user-friendly or too expensive, alternatives with better UI and smoother experiences can be found immediately. The entire market is becoming increasingly competitive, but this also accelerates the rollout of truly valuable products.

Awakening of Data Sovereignty Awareness: Who is the Real Owner of the Data?

Amid all this rapid development, more and more people are starting 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 an 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 major AI company—its chat AI's "memory feature" can now reference all past conversations with users to generate more personalized responses. This feature is indeed cool; imagine a future where everyone has their own AI personal assistant, chat companion, emotional support... But this also means that users' data will be "long-term held" by a platform, and users are no longer the true owners of their data.

Once others control your conversations, preferences, emotions, and even lifestyle habits, the consequences can be far more than just "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, predictions indicated that by the second quarter of 2025, decentralized AI would truly enter the public eye. Especially against the backdrop of increasing emphasis on privacy security and data ownership, underlying infrastructures that can provide confidentiality, verifiability, and transparency of user data ownership will gain more attention and usage.

Currently, we see three main trends emerging:

Venture Capital Trends in Web2 AI

• Some startups supported by a well-known startup incubator are launching vertical domain AI agents (specifically designed to address issues in a certain field);
• A well-known venture capital firm has begun to lay out its strategy for 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.

Web3 AI Venture Capital Trends

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

Retail Trend of Web3 AI

• AI Agent ecosystem remains 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 stage. The year 2025 will be a crucial moment to validate the value of decentralized AI.

Web2 vs Web3 AI: Completely Different Rhythms and Gameplay

The Web2 market size is much larger than that of Web3. Many traditional enterprises are seeking to leverage AI for transformation and optimize their business processes—such as acquiring more customers, improving conversion rates, and increasing sales. These companies typically 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 who are targeting these niche demands to develop vertical AI agents.

Compared to traditional SaaS, the benefits brought by AI agents are more direct—either saving a lot of costs or directly attracting more customers to generate revenue. Therefore, the subscription prices for such AI tools can also be sold at a higher price, and many startups can achieve annual revenues of millions or tens of millions of dollars within a few months of launching, which is not without reason.

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

Venture capital firms in Web3 have begun to plan for this future.

Why Retail Investors Prefer AI Agents

For Web3 retail investors, DeAI (Decentralized AI) is indeed difficult to understand: a bunch of new words and concepts that sound like a foreign language. Therefore, what initially attracts them most are those AI agents that are easy to understand and entertaining — like talking chatbots that can tell jokes and be funny. These "entertainment AI agents" are indeed very appealing, but over time, retail investors also begin to realize that these things seem to have no practical use. Additionally, with the recent market downturn, many useless projects are slowly being eliminated, while those with practical value that can provide functionality, although their valuations have also decreased, are still surviving.

This wave of "market cleansing" has made more and more people realize that only AI projects with practical use cases and core product capabilities have a future. As a result, project parties have begun to turn in two directions. Either they develop genuine AI products independently to solve real problems; or they collaborate with truly technical and valuable DeAI projects.

This transformation brings two positive effects: it prompts people to pay attention to the originally "hard-to-understand" underlying infrastructure; and it enables AI agents to be more than just performance tools, but products that can solve real problems. Some projects have already become typical cases—powerful not only in functionality but also introducing some cool DeAI technologies into the public eye. This indicates a trend: retail investors may not understand the technology, but they will gradually be educated by "truly useful" products.

One of the most interesting aspects of certain DeAI projects is that they are decentralized AI ecosystems in which ordinary people can invest. Currently, most DeAI projects are still in their early stages, and only venture capitalists or "strategic partners within the circle" can invest, with many not even having issued tokens. However, some projects are different. Users can directly vote with tokens to support promising subnets, effectively "reallocating" to these DeAI project's sub-tokens and positioning themselves in advance.

Although some people have complained before: cross-chain bridges and trading experience can be a bit troublesome... Their underlying technology, product logic, and overall atmosphere are indeed very strong. Especially with the presence of certain teams, the entire ecosystem's UX/UI design is evolving towards a "user-friendly" direction. Because in the mechanisms of certain projects, each subnet must rely on market recognition to obtain more rewards (mining incentives) — those who are useful and powerful can receive more distribution.

Therefore, for these subnets, "making users understand what you are doing" becomes crucial. Some teams are doing this. Their product direction is very clear: optimizing UI/UX for ordinary users. They not only have several practical subnets (such as a super convenient AutoML platform where users can train models directly and get them running with just a click of a button. Their latest flagship product is also very cool: the AI Agent platform, where users can drag and drop modules to create AI agents, truly achieving "zero-code AI agent building." This experience is somewhat like a "foolproof AI factory" for Web3, making it very suitable for users who do not understand technology to get started.)

Overall, certain DeAI ecosystems are not only among the most technologically advanced projects but also lead in terms of user-friendliness for general user participation. The clarity of product logic and the user-friendly nature of the team are key factors 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 businesses optimizing operations with AI in Web2 or retail investors experiencing the convenience of new agents in Web3, future data sovereignty and user participation will be crucial. Web3 AI has yet to reach its peak. The real performance has only just begun.