After the AI Agent bubble bursts, where is the real value of Web3 AI?

In the fourth quarter of last year, the AI agent track rapidly emerged, with its market value skyrocketing from nearly zero to over $20 billion. Various "agents" have emerged, including humorous and even quirky ones, as well as "financial agents" that can automatically trade cryptocurrencies and help you make money. However, this craze came quickly and left just as fast. After the bubble burst, many projects collapsed one after another. Nevertheless, some AI projects and infrastructure with practical value are still developing steadily, and real value is beginning to emerge. The next wave of Web3 AI is brewing, and this time it may not just be hype, which is worth our close attention.

When a new track or hotspot emerges, the market often pays little attention to the fundamentals. As long as the project looks lively, has gimmicks, and presents well, regardless of its actual utility, its market capitalization can easily reach hundreds of millions of dollars.

In this wave, some projects have precisely occupied the market and user mindset with outstanding narrative abilities. Subsequently, some open-source AI projects emerged, allowing any developer to easily get started and create value on their own. This concept quickly resonated widely, and the community grew rapidly.

However, the market atmosphere is completely different now. The newly launched agent projects that perform well mostly have a market capitalization between 3 million and 10 million dollars; the market capitalization of older projects has also compressed to the range of 10 million to 50 million dollars. The valuation ceiling for the entire sector has been lowered, with the total market capitalization dropping from a peak of 20 billion dollars to the current range of 4 to 6 billion dollars.

Rise of Infrastructure, Acceleration of Web2 AI Development

The market no longer blindly believes in those "seemingly powerful" bubble projects, but instead focuses 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 updated almost every month, becoming stronger, faster, and smarter. For example, the image generation feature recently launched by ChatGPT caused a sensation as soon as it went live, taking the internet by storm.

The consumer products side of Web2 is also evolving rapidly. Due to enhanced underlying AI capabilities, many previously unattainable product experiences are now possible. New tools have greatly improved developer efficiency, with frequent and abundant feature updates. AI agents and smart workflows have penetrated various fields, and the barrier to entry is getting lower. For users, switching tools incurs almost no cost, and they can find better alternatives at any time. The overall market competition is becoming increasingly fierce, but it also accelerates the implementation of truly valuable products.

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

As AI develops rapidly, more and more people are beginning to realize an issue: there are AI agent applications everywhere, but most of them use centralized technology. So, who really owns our data? Where do our chat records go? If we discuss 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 critical after recent updates from certain AI companies. For example, ChatGPT's "memory feature" can now reference all past conversations of users, generating more personalized responses. This feature is indeed cool, and in the future, everyone may have their own AI personal assistant, chat companion, and emotional support. But this also means that users' data will be "held long-term" 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 lifestyle, 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)

There are predictions that by the second quarter of 2025, decentralized AI will truly come into the public eye. Particularly against the backdrop of increasing public concern over privacy security 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 emerging in three main directions:

Web2 AI VC trends

• Some startup companies supported by well-known incubators are launching vertical AI agents (specifically designed to solve problems in a certain field);
• Prominent venture capital firms are beginning to position themselves for the next phase of consumer AI product trends, proposing their own investment logic;
• Some AI companies have launched funds specifically for investing in AI.

Web3 AI VC trends

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

The retail trend of Web3 AI

• The AI Agent ecosystem remains one of the hotspots.
• Consumer-grade AI applications are gradually becoming more diverse, trying to implement everything from productivity tools to emotional companionship.
• Users are increasingly concerned about "whether the AI products I use truly serve me, rather than just harvesting my data".

These trends are intertwined, driving DeAI from concept to practical stage. The year 2025 will be a key 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. 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. Such companies usually have clear demands, many of which 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 needs to develop vertical AI agents.

Compared to traditional SaaS, the benefits that AI agents can bring are more direct—either significantly reducing costs or directly attracting more customers to increase revenue. Therefore, the subscription prices for these AI tools can also be set higher, and many startups can achieve annual revenues of millions or tens of millions of dollars within just 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 tailor-made 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. In simple terms, the future goal of Web3 AI is to enable users to understand how their data is used, understand the decision-making process of AI, autonomously control models and use cases, and even profit from it.

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

Why Retail Investors Favor AI Agents

For Web3 retail investors, DeAI (decentralized AI) is indeed quite difficult to understand: filled with new words and concepts, it sounds like a foreign language. So initially, they are most easily attracted to those simple, easy-to-understand, and fun AI agents—like chatty robots that can tell jokes and entertain. These "entertainment-type AI agents" are indeed very appealing, but over time, retail investors also start to realize that these products seem to have little practical use. Furthermore, with the recent poor market conditions, a large number of useless projects have gradually been eliminated, while those agents with practical value that can provide actual 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 parties have begun to turn in two directions: either develop real AI products themselves to solve practical problems, or collaborate with DeAI projects that truly have technology and value.

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

One of the most interesting aspects of some DeAI projects is that they are decentralized AI ecosystems in which ordinary people can also participate in investing. Currently, most DeAI projects are still in the early stages, where only venture capital or "insider strategic partners" can invest, and many projects have not even issued tokens yet. However, some projects are different; users can directly vote with tokens to support promising subnets, effectively laying out for these DeAI projects' sub-tokens in advance.

Although some projects' cross-chain bridges and trading experiences still need improvement, their underlying technology, product logic, and overall atmosphere are indeed strong. In particular, the addition of teams focused on improving user experience has made the UX/UI design of the entire ecosystem more user-friendly. Because in the mechanisms of certain DeAI projects, each subnet needs to gain market recognition to obtain more rewards (mining incentives) — who is useful, who is capable, will be able to receive more distribution.

Therefore, for these subnetworks, "making users understand what you are doing" becomes crucial. Some teams are dedicated to this matter. Their product direction is very clear: optimizing UI/UX for the average user. They not only have multiple practical subnetworks (such as some convenient AutoML platforms where users can train models directly with just a few clicks), but they also launched some flagship products: for example, the AI Agent platform, where users can drag and drop modules to create AI Agents, truly achieving "no-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 tech-savvy to get started.

Overall, some DeAI ecosystems are not only technologically advanced but also lead in user engagement friendliness. This clear product logic and user-friendly team are the key factors that make this ecosystem attractive.

We are in a transformative era dominated by Web3 AI. The past bubble of inflated market values through 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 individual users experiencing the convenience of new agents in Web3, future data sovereignty and user participation will be key. Web3 AI is far from reaching its peak. The real show has just begun.