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

In the fourth quarter of last year, the AI agent sector rapidly emerged, with its market value soaring from nearly zero to over $20 billion. Various "intelligent agents" became all the rage, ranging from humorous to quirky. People began to envision "financial agents" that could automatically trade cryptocurrencies and help them make money. There even emerged DAOs claiming to invest in other agents, and some organizations proposed "co-governance of humans and agents". The ways to participate are endless, and it seems that overnight wealth is just around the corner.

However, the opportunities come quickly and go just as fast. After the bubble burst, many projects fell one after another. However, some AI projects and infrastructure with practical value are gradually taking over the market. Real value is starting to emerge, and the next wave of Web3 AI 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, the market often disregards the fundamentals. As long as it looks lively, has a gimmick, and is presented well, even if its actual utility is minimal, the market cap can easily exceed hundreds of millions.

In this wave, some projects are particularly good at storytelling. They accurately grasp the market, occupy users' minds, and have outstanding narrative abilities. As a result, developers compete to launch projects on their platforms, and retail investors follow suit and hype them up.

Later, another 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 resonated, and the community grew rapidly, with a surge in the number of stars and forks on GitHub.

The total valuation of a certain ecosystem once broke through 5 billion USD, while another project approached nearly half that scale at its peak. Other interesting AI agent projects also reached a market value of around 1 billion.

Nowadays, the situation has changed dramatically. The newly launched agent projects that perform well mostly have a market value between 3 million and 10 million; the market value of older projects has also been compressed to the range of 10 million to 50 million. The upper limit of the entire sector's valuation has been compressed, and the total market value has dropped 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 "superficially glamorous" 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 from major tech companies are being updated and iterated almost every month, becoming stronger, faster, and smarter. For example, the recent launch of an image generation feature by a certain chat AI platform has sparked a "Ghibli-style" image craze, flooding social media.

The consumer products side of Web2 is also evolving rapidly. With enhanced underlying AI capabilities, many previously unattainable product experiences have now become possible. Some new tools have significantly improved developer efficiency, with frequent and diverse feature updates. AI agents and smart workflows have now permeated every corner, and the entry barriers are getting lower. For users, switching tools is almost cost-free—if a tool is poorly designed or too expensive, they can immediately find alternatives with better UI and experience. The entire market is becoming increasingly competitive, but it is also accelerating the rollout of truly valuable products.

Awakening of Data Sovereignty: Who is the True Owner of the 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 private matters with AI, will it really keep them confidential? Or will they be uploaded, analyzed, and used to train other models?

This issue has become even more critical after the recent update of a certain AI platform - its "memory function" can now reference all your past conversations to generate more personalized responses. This feature is indeed cool; just imagine, in the future, everyone will have their own AI personal assistant, chat companion, emotional support... But this also means that your data will be "long-term held" by a certain platform, and you are no longer the owner of your data.

Once others control your conversations, preferences, emotions, and even habits, the consequences may be more than just a "better experience."

This is also 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)

Predictions made last year indicated that by the second quarter of 2025, decentralized AI will truly enter the public eye. Especially against the backdrop of increasing emphasis on privacy, security, and data ownership, foundational infrastructures that can provide confidentiality, verifiability, and transparency regarding user data ownership will receive more attention and usage.

Currently, we see three main trends emerging:

Web2 AI Venture Capital Trends

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

Venture Capital Trends in Web3 AI

• 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 hot topics;
• Consumer-grade AI applications are gradually becoming richer, attempting to land from productivity tools to emotional companionship;
• Users are increasingly concerned about whether "the AI products I use truly serve me, rather than harvesting my data."

These trends intertwine to drive DeAI from concept to practical stage. The year 2025 will be a critical moment to validate the value of decentralized AI.

Web2 vs Web3 AI: Completely Different Rhythms and Playstyles

The market size of Web2 is much larger than that of Web3. This is because many traditional enterprises are seeking to transform through AI to optimize their business processes—such as acquiring more customers, improving conversion rates, and increasing sales. These enterprises usually have very clear needs, and many are focused on specific subfields, so they hope to find AI tools that can accurately address their "specific pain points." This has also attracted many young entrepreneurs aiming to become vertical field AI agents targeting these subfield demands.

Compared to traditional SaaS, the benefits brought by AI agents are more direct — either save a large amount of money or directly attract more customers to make money. Therefore, the subscription prices for these AI tools can also be sold at a higher price, 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 way of Web3 is completely different. The blockchain itself is a foundation layer tailored for decentralized AI (DeAI). All actions can be verified on-chain and are immutable; it naturally provides a trustless environment; supports decentralized computing; users can truly own their data, models, and use cases. Simply put, the future of Web3 AI is to let users know how their data is being used, understand the decision-making process of AI, have autonomous control over models and use cases, and also profit from it.

Venture capitalists in Web3 have already started 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 words and concepts that sound like an alien language. So initially, what attracts them the most are those easily understandable and fun AI agents—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 real utility? Moreover, with the recent poor market conditions, a large number of useless projects are slowly being eliminated, while those agents with practical value that can provide functions, although their valuations have also declined, still survive.

This wave of "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 teams have begun to turn in two directions: either develop real AI products themselves to solve actual problems, or collaborate with DeAI projects that truly have technology and value.

This transformation has two positive implications: it encourages everyone to start paying attention to the underlying infrastructure that was originally "difficult to understand"; it turns AI agents from mere performance tools into products that can deliver real results. Some projects have already become typical examples — not only are they powerful in functionality, but they also bring some cool 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 certain DeAI projects is that they are decentralized AI ecosystems in which ordinary people can also participate in investment. Currently, most DeAI projects are still in their early stages, with only venture capital or "insider strategic partners" able to invest, and many have not even issued tokens yet. However, some projects are different. Users can directly vote with tokens to support the subnet they believe in, which is equivalent to "reallocating" into these DeAI project sub-tokens, allowing for early positioning.

Although there have been complaints in the past: cross-chain bridges and trading experiences can be somewhat troublesome... their underlying technology, product logic, and overall atmosphere are indeed 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 gain more rewards (mining incentives) — those who are useful, those who are strong, are the ones who can receive more distribution.

Therefore, for these subnetworks, "making it clear to users what you are doing" becomes crucial. Some teams are working on this issue. Their product direction is very clear: optimizing UI/UX for ordinary users. They not only have several practical subnetworks (such as a super convenient AutoML platform where users can directly train models and run them with just a click of a button), but 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. Teams that create products with clear logic and user-friendly designs are the key players that make this ecosystem vibrant.

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, data sovereignty and user participation will become key in the future. Web3 AI is far from its peak. The real show has just begun.