Coinspire is a research community that has long focused on the Web3 and AI sectors, providing project research reports, primary investment research, secondary market analysis, and other Alpha opportunities.
Introduction
In the fourth quarter of last year, the AI agent sector skyrocketed like a fire arrow, rising from almost zero to a market value of over $20 billion in no time. Various "agents" became popular—some funny, even bizarre—followed by a wave of excitement for "financial agents" that could automatically trade cryptocurrencies and help you make money, along with various InvestmentDAOs claiming to invest in other agents, and even DAO organizations advocating for "co-governance between humans and agents." The variety of playstyles made it seem like getting rich overnight was not a dream.
However, the wind comes quickly and goes just as fast. Once the bubble burst, many projects cooled down one after another. Nevertheless, some AI projects with foundational infrastructure and practical value are taking over the market. The true value is beginning to emerge, and the next wave of Web3 AI is brewing—this time it may not just be hype, and it deserves serious attention.
We all know that whenever there is a new track or a new hotspot (like Web2 AI or the combination of Trump supporting cryptocurrencies and AI), the market doesn't care about the fundamentals. As long as it looks lively, has a gimmick, and the demo is impressive, the market cap can easily soar to over a hundred million dollars, regardless of the actual utility.
@virtuals_io is the project that tells the best stories in this wave. They precisely hit the market and occupy users' minds, with top-notch narratives. As a result, developers want to launch projects on it, and retail investors follow the hype.
Later, @elizaOS emerged, taking a completely different route—it open-sourced AI, allowing any developer to easily get started and mine tokens themselves. This concept resonated widely, and the community grew rapidly, with the number of stars and forks on GitHub skyrocketing (and still rising).
The total valuation of the Virtuals ecosystem once surged to over $5 billion, with Eliza peaking at nearly half that size. Other interesting AI agents, such as AIXBT, also once reached a market cap of $1 billion.
Of course, the current landscape is completely different. Newly launched, well-performing agent projects mostly have market caps between $3 million and $10 million; older projects have seen their market caps compressed to the $1 million to $5 million range. The valuation ceiling of the entire sector has been compressed, with the total market cap dropping from a peak of $20 billion to the current range of $4 billion to $6 billion.
Part 1: The Rise of Infrastructure, Web2 AI Accelerates
The current market no longer blindly believes in those "impressive-looking" bubble projects but is starting to focus on real fundamentals. Especially against the backdrop of rapidly developing AI models in Web2, people are more concerned about the long-term value of infrastructure and decentralized AI.
Meta's Llama, OpenAI's GPT, Grok on X, DeepSeek, and Alibaba's Qwen—almost every month there are updates, with models becoming stronger, faster, and smarter. For example, recently when ChatGPT's image generation feature went live, it immediately sparked a wave of "Giraffe Power" images, flooding the screens.
Consumer products in Web2 are also evolving at breakneck speed. As the underlying AI capabilities have improved, many previously unattainable product experiences are now possible. New tools like Lovable, Bolt, Cursor, and Windsurf have significantly enhanced developers' efficiency, with rapid and abundant feature updates. AI agents and intelligent workflows have now penetrated every corner, and the entry barriers are getting lower. For users, switching tools incurs almost no cost—if one is hard to use or expensive, they can quickly find a better UI and smoother experience alternative. The entire market is becoming increasingly competitive, but it also accelerates the realization of truly valuable things.
Part 2: Awakening of Data Sovereignty: Who is the True Owner of Data?
Amidst 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 do my chat records go? If I discuss private matters with an AI, will it really keep it confidential? Or will it be uploaded, analyzed, and used to train someone else's model?
This question has become even more critical after @OpenAI's recent updates—ChatGPT's "memory feature" can now reference all your past conversations 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 your data will be "long-term held" by a platform, and you are no longer the owner of your data.
Once someone else controls your conversations, preferences, emotions, and even lifestyle habits, the consequences may not just be about "better experiences."
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.
Part 3: The Rise of Decentralized AI (DeAI)
I made several predictions last year, one of which is that by the second quarter of 2025, decentralized AI will truly enter the public eye. Especially against the backdrop of increasing concern for privacy, security, and data ownership, 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:
1. VC Trends in Web2 AI
▶ Some startups supported by Y Combinator are launching niche AI agents (specifically solving problems in certain fields);
▶ a16z is starting to lay out the next phase of consumer AI product trends, proposing its own investment logic;
▶ Perplexity has launched a fund dedicated to investing in AI.
2. VC Trends in Web3 AI
▶ Betting on decentralized AI infrastructure is beginning;
▶ Distributed training networks, computing sharing, and other sectors are gradually heating up.
3. Retail Trends in Web3 AI
▶ The AI Agent ecosystem remains one of the hotspots;
▶ Consumer-level AI applications are gradually diversifying, 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, collectively pushing DeAI from concept to practical stage. 2025 will be a key moment to validate the value of decentralized AI.
Part 4: Web2 vs Web3 AI: Completely Different Rhythms and Playstyles
On the Web2 side, the market size is far larger than that of Web3. Many traditional companies are trying to use AI to transform and optimize their business processes—such as acquiring more customers, improving conversion rates, and increasing sales. These companies usually have clear needs, often concentrated in specific niches, so they hope to find AI tools that can precisely address their "specific pain points." This has attracted many young entrepreneurs targeting these niche demands to create specialized AI agents.
Compared to traditional SaaS, the benefits brought by AI agents are more direct—either saving a significant amount of money or directly attracting more customers to make profits. Therefore, the subscription prices for these AI tools can be higher, and many startups can reach annual revenues of hundreds of thousands or millions of dollars within a few months of launching, which is not without reason.
However, the playstyle in 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 data, models, and use cases. In simple terms—Web3 AI aims to let users know how their data is used, understand the AI decision-making process, control their models and use cases, and even profit from it.
"Web3 VCs are already betting on this future."
Part 5: Why Retail Investors Love AI Agents
For retail investors in Web3, DeAI (decentralized AI) is indeed quite difficult to understand: a bunch of new terms and concepts that sound almost alien. So initially, they are most easily attracted to those AI agents that are understandable, fun, and entertaining—like chatbots that can talk and tell jokes. These "entertainment-type AI agents" do attract followers, but over time, retail investors also begin to realize that these things seem to have no real utility? Coupled with the recent poor market conditions, a large number of useless projects have gradually been eliminated, while those useful agents that can provide functionality, although their valuations have also dropped, are still alive.
This wave of "cleansing" has made more and more people realize: only AI projects with practical use cases and core product capabilities have a future. As a result, project parties have begun to shift in two directions. Either they develop real AI products themselves to solve practical problems; or they collaborate with truly technical and valuable DeAI projects, such as @AlloraNetwork and @opentensor (Bittensor).
This shift has two positive implications: it makes everyone start to pay attention to the underlying infrastructure that was originally "incomprehensible"; and it transforms AI agents from mere performance tools into products that can deliver real results. Projects like @AskBillyBets and @thedkingdao have already become typical cases—not only are they powerful in functionality, but they also bring cool DeAI technologies like Bittensor 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 points about the aforementioned project Bittensor is that it is a decentralized AI ecosystem that ordinary people can also participate in investing. Currently, most DeAI projects are still in their early stages, and only VCs or "strategic partners within the circle" can invest, with many not even having issued tokens yet. But Bittensor is different. Users can directly use $TAO to vote and support the subnet they believe in, equivalent to "switching positions" into these DeAI projects' sub-tokens, getting on board early.
Although I have previously complained about cross-chain bridges and the trading experience being a bit annoying… their underlying technology, product logic, and overall atmosphere are really strong. Especially with the presence of @rayon_labs, the entire ecosystem's UX/UI design is evolving towards a "user-friendly" direction. In the mechanism of Bittensor, each subnet must rely on market recognition to earn more rewards (mining incentives)—those who are useful and valuable can receive more distribution.
Therefore, for these subnets, "making users understand what you are doing" becomes super important. Rayon Labs is doing just that. Their product direction is very clear: optimizing UI/UX for ordinary users. They not only have several practical subnets (such as Gradients: an extremely convenient AutoML platform where users can directly train models and get them running with just a few clicks. Their latest flagship product is also very cool: the Squad AI Agent platform, where you can drag and drop modules to create AI Agents, truly achieving "zero-code AI agent building." This experience is somewhat like a Web3 version of a "dummy-proof AI factory," making it suitable for users who are not tech-savvy to get started).
"Overall," the Bittensor ecosystem is now not only one of the most technologically advanced DeAI projects but also leads in user participation friendliness. This clear product logic and user-friendly team is the key role that makes this ecosystem vibrant.
We are in a transformative era led 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 enterprises are using AI to optimize their business in Web2 or retail investors are experiencing the convenience of new agents in Web3, future data sovereignty and user participation will become crucial. The peak of Web3 AI is still far from being reached. The real show has just begun.
All content on the Coinspire platform is for reference only and does not constitute an offer or advice for any investment strategy. Any personal decisions made based on the content of this article are the sole responsibility of the investor, and Coinspire is not responsible for any profits or losses incurred as a result. Investment involves risks, and decisions should be made cautiously.
免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。
