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

Introduction

In the fourth quarter of 2023, the AI agent sector experienced explosive growth, with its market value rapidly rising from nearly zero to over $20 billion. Various types of "intelligent agents" have become a global sensation, ranging from funny and entertaining ones to "financial agents" that automatically trade cryptocurrencies and help you make money, to DAOs that invest in other agents, and even organizations that feature "co-governance between humans and agents." The gameplay is rich and diverse, leaving people dazzled.

However, this wave of enthusiasm came quickly and went just as fast. After the bubble burst, many projects collapsed one after another. However, some AI infrastructure projects with practical value are gradually taking a dominant position in the market. Real value is beginning 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.

As we all know, whenever a new track or hotspot emerges, the market often ignores the fundamentals. As long as a project looks lively, has hype, and a nice demo, regardless of its actual utility, the market capitalization can easily surpass hundreds of millions of dollars.

In this wave, some projects have captured market demand with excellent narrative ability, occupying users' minds. Developers are launching projects on their platforms, and retail investors are also following the hype.

Subsequently, an open-source AI project emerged, taking a completely different path. It allows any developer to easily get started and create value independently. This concept quickly resonated widely, and the community grew rapidly, with the number of stars and forks on GitHub increasing exponentially.

The total valuation of a certain AI ecosystem once surpassed $5 billion, while another open-source project approached nearly half of that amount at its peak. Other interesting AI agent projects have also seen their market capitalization exceed $1 billion at one point.

However, the market landscape has changed dramatically today. The newly launched and performing well agent projects mostly have a market capitalization between 3 million and 10 million dollars; the market capitalization of older projects has also been compressed to the range of 10 million to 50 million dollars. The valuation ceiling of the entire sector has been compressed, with the total market capitalization decreasing from a peak of 20 billion dollars to the current range of 4 to 6 billion dollars.

The Rise of Infrastructure, Accelerating Development of Web2 AI

The current market no longer blindly chases after those "seemingly powerful" bubble projects, but instead focuses on 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.

The AI models of major tech giants are almost updated and iterated every month, becoming more powerful, faster, and smarter. For example, the recent launch of a well-known AI assistant's image generation feature immediately sparked a craze for "Ghibli-style" images, quickly taking over the screen.

The consumer products end of Web2 is also evolving rapidly. Due to the enhancement of underlying AI capabilities, many previously unattainable product experiences are now achievable. Emerging AI development tools have significantly improved developers' work efficiency, with frequent and abundant feature updates. AI agents and intelligent workflows have infiltrated various fields, and the entry barriers are getting lower. For users, switching tools is almost costless — if one is not user-friendly or is expensive, an alternative with better UI and experience can be found immediately. The overall market competition is becoming increasingly fierce, but it also accelerates the landing of truly valuable products.

Awakening of Data Sovereignty: Who is the True Master 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 actually owns my data? Where will my chat records go? If I discuss some private content with the 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 after a recent update from a well-known AI company—its chatbot's "memory function" 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 it also means that users' data will be "held long-term" by a certain platform, and users are no longer the masters of their data.

Once others control your conversations, preferences, emotions, and even lifestyle habits, the consequences may not be as simple as 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 were made as early as last year 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, foundational infrastructures that can provide confidentiality, verifiability, and transparency in user data ownership will receive more attention and usage.

Currently, we see trends emerging in three main directions:

1. VC Trends of Web2 AI

• Some startups supported by a well-known incubator are launching vertical AI agents (specifically addressing issues in a certain field);
• A well-known venture capital firm has started to lay out the next phase of consumer AI product trends and has presented its own investment logic;
• A certain AI search engine has launched a fund specifically for investing in AI.

2. The VC Trends of Web3 AI

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

3. Retail Trends of Web3 AI

• The AI Agent ecosystem remains one of the hot topics;
• Consumer-grade AI applications are gradually becoming more diverse, attempting to materialize 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 are intertwined, collectively driving DeAI from concept to practical stage. The year 2025 will be a crucial moment for validating the value of decentralized AI.

Web2 vs Web3 AI: Completely Different Rhythms and Playstyles

The Web2 market is significantly larger than Web3. Many traditional enterprises are seeking to transform through the use of AI to optimize their business processes—for example: acquiring more customers, improving conversion rates, increasing sales, etc. 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 specialized AI agents.

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 such AI tools can also be sold at a higher rate, and many startups can achieve annual revenues of millions or tens of millions of dollars within a few months of launch, which is not without reason.

But the way 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; and users can truly own their own data, models, and use cases. In short, the future of Web3 AI will allow users to understand how their data is used, comprehend the decision-making processes of AI, have autonomous control over models and use cases, and even profit from it.

Venture capital firms in Web3 have begun to bet on this future.

Why Retail Investors Favor AI Agents

For Web3 retail investors, DeAI (Decentralized AI) is indeed quite difficult to understand: it is filled with a lot of new terms and concepts that sound like an alien language. Therefore, initially, they are most easily attracted to AI agents that are understandable and fun—like talking chatbots 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 things seem to have little practical use. Additionally, with the recent poor market conditions, many useless projects are gradually being eliminated, while those with actual value and functionality, although their valuations have also decreased, still survive.

This round of "reshuffling" 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 independently develop genuine AI products to solve real problems, or collaborate with truly capable and valuable DeAI projects.

This transformation brings two positive impacts: it makes everyone start paying attention to the underlying infrastructure that was originally "incomprehensible"; it allows AI agents to no longer be just performance tools, but products that can solve real problems. Some projects have already become typical cases—not only are they powerful, but they also bring 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 interesting aspect of certain decentralized AI ecosystems is that they are platforms where ordinary people can also participate in investment. Currently, most DeAI projects are still in the early stages, and only venture capital or "insider strategic partners" can invest, with many projects not even having issued tokens yet. However, these ecosystems are different. Users can directly use tokens to vote in support of promising subnets, essentially "reallocating" into the sub-tokens of these DeAI projects, positioning themselves in advance.

Although there have been criticisms before: cross-chain bridges and trading experiences can be somewhat troublesome... the underlying technologies, product logic, and overall atmosphere of these ecosystems are indeed very strong. Especially with the presence of certain teams, the UX/UI design of the entire ecosystem is evolving towards "user-friendly". Because in the mechanisms of these ecosystems, each subnet must rely on market recognition to gain more rewards (mining incentives) — who is useful, who is strong, only those can receive more distribution.

Therefore, for these subnets, "making users understand what you are doing" becomes crucial. Some teams are working on this matter. Their product direction is very clear: optimizing UI/UX for the average user. They not only have multiple practical subnets (such as some automated machine learning platforms where users can directly train models with just a few clicks), but they have also launched some flagship products: for example, the AI Agent platform, where users can create AI agents by dragging and dropping blocks, 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 tech-savvy to get started.

Overall, these ecosystems are not only at the forefront of DeAI projects technologically but also lead in user-friendliness for general user participation. The clear product logic and user-friendly team are key factors that make this ecosystem attractive.

We are in a transformative era dominated by Web3 AI. The past bubble driven by hype to inflate market value has been replaced by real infrastructure, decentralized AI, and genuine application scenarios. Whether it's enterprises optimizing their businesses using AI in Web2 or retail investors experiencing the convenience of new agents in Web3, data sovereignty and user participation will be key in the future. Web3 AI is still far from its peak; the real play has just begun.