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

Introduction

In the fourth quarter of 2023, the AI agent sector rapidly emerged, with market value skyrocketing from nearly zero to over $20 billion. Various "intelligent agents" became a sensation, ranging from the humorous to the bizarre. People began to fantasize about "financial agents" that could automatically trade cryptocurrencies and help them make money, and various projects claiming to invest in other agents emerged. There were even organizations advocating for "co-governance between humans and agents," with a plethora of gameplay options, making it seem like getting rich overnight was just around the corner.

However, the windfall comes quickly and goes just as fast. After the bubble burst, many projects collapsed one after another. However, some AI infrastructure projects with practical value are gradually dominating the market. The real value is starting to emerge, and the next wave of Web3 AI is brewing, which may not just be hype this time—something we should pay serious attention to.

As we all know, whenever a new track or hotspot emerges (such as Web2 AI, the combination of Trump supporting cryptocurrencies and AI), the market often does not pay much attention to the fundamentals. As long as the project looks lively, has hype, and an impressive demo, regardless of its actual utility, the market cap can easily soar to over a hundred million dollars.

In this wave, certain projects excel at storytelling, accurately grasping the market and occupying users' mindsets, with outstanding narrative capabilities. As a result, developers are rushing to launch projects on their platforms, and retail investors are following suit to speculate.

Later, some projects took a completely different approach - open-source AI, allowing any developer to easily get started and create value on their own. This concept quickly resonated widely, the community rapidly grew, and the number of stars and forks on GitHub surged dramatically (and continues to rise to this day).

The total valuation of a certain ecosystem once surged to over 5 billion USD, while another project's peak was also close to half of this amount. Other interesting AI agent projects have also once broken the 1 billion USD market cap.

Of course, the current market environment is completely different now. Newly launched and well-performing agent projects mostly have a market capitalization between 3 million to 10 million USD; the market capitalization of older projects has also been compressed to a range of 10 million to 50 million USD. The upper limit of the entire sector's valuation has been compressed, and the total market capitalization has decreased from its peak of 20 billion USD to the current range of 4 to 6 billion USD.

Part One: The Rise of Infrastructure, Accelerated Development of Web2 AI

The market no longer blindly trusts those "impressive-looking" 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 being updated and iterated almost every month, becoming more powerful, faster, and smarter. For example, as soon as the image generation feature of a certain chatbot was launched recently, it immediately triggered a "Ghibli-style" image craze, quickly going viral on social media.

The consumer products side of Web2 is also evolving rapidly. With the enhancement of underlying AI capabilities, many previously unattainable product experiences are now possible. Some new tools have significantly increased developer efficiency, with rapid and numerous feature updates. AI agents and smart workflows have now permeated every corner, and the entry barriers are getting lower. For users, switching tools involves almost no cost—if a tool is not user-friendly or is expensive, an alternative with better UI and smoother experience can be found immediately. The overall market competition is intensifying, but it is also accelerating the realization of truly valuable products.

Part Two: 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 really 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 critical after the recent update from a certain AI company—its chatbot's "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 "held long-term" by a certain platform, and you are no longer the master of your data.

Once others control your conversations, preferences, emotions, and even lifestyle habits, the consequences may not be as simple as "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 the users is the most valuable future.

Part Three: The Rise of Decentralized AI (DeAI)

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

Currently, we see three main directional trends emerging:

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 a certain field);
• A well-known venture capital firm has started to lay out 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 sectors are gradually heating up.

Three, the Retail Trend of Web3 AI

• The AI Agent ecosystem is still one of the hot topics;
• Consumer-level 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 truly serve me, rather than harvesting my data."

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

Part Four: Web2 vs Web3 AI: Completely Different Rhythms and Playstyles

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

Compared to traditional SaaS, the benefits brought by AI agents are more direct—either you save a lot 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 reach 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 of Web3 AI will allow users to understand how their data is used, comprehend the AI decision-making process, autonomously control models and use cases, and also profit from it.

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

Part Five: 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 with actual value and 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 use cases and core product capabilities have a future. As a result, project parties are starting to shift in two directions: either develop genuine AI products themselves to solve real problems, or collaborate with DeAI projects that truly possess technology and value.

This transformation has two positive implications: it encourages everyone to start paying attention to the underlying infrastructure that was originally "incomprehensible"; it allows AI agents to no longer just be performance tools, but rather products that can deliver real results. Some projects have already become typical cases—not only powerful in functionality but also bringing 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 some DeAI projects is that they are decentralized AI ecosystems that ordinary people can also participate in investing. Currently, most DeAI projects are still in the early stages, and only venture capital or "strategic partners in the circle" can invest, with many projects not even having issued tokens yet. However, some projects are different. Users can directly vote with tokens to support promising subnets, which is equivalent to participating early in the sub-tokens of these DeAI projects.

Although the cross-chain bridges and trading experiences of these projects may still have some troubles, their underlying technology, product logic, and overall atmosphere are indeed strong. In particular, the presence of certain teams has led the user experience design of the entire ecosystem to develop in a more friendly direction. Because in the mechanisms of certain projects, each subnet has to rely on market recognition to obtain more rewards (mining incentives)—whoever is useful and powerful can receive more distribution.

Therefore, for these subnetworks, "helping users understand what you are doing" becomes crucial. Some teams are doing this. Their product direction is very clear: optimizing the interface and user experience for ordinary users. They not only have multiple practical subnetworks (such as a super convenient AutoML platform where users can directly train models with just a few clicks), but they have also launched some very cool flagship products: for example, an AI agent platform where you can create AI agents just by dragging and dropping, truly achieving "zero-code AI agent building." This experience is somewhat like a "foolproof AI factory" version of Web3, making it very suitable for non-technical users to get started.

Overall, certain DeAI ecosystems are not only leading in technology but also at the forefront of user-friendliness for ordinary user participation. The clear product logic and user-friendly teams are key roles that make this ecosystem attractive.

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