ArkStream Capital: Can AI Agents Be the Lifeline for Web3+AI?
TechFlow Selected TechFlow Selected
ArkStream Capital: Can AI Agents Be the Lifeline for Web3+AI?
For Web3 projects, integrating AI technology into non-AI-core application products could become a strategic advantage.
Navigating Web3 tides with focused insightsAuthor: James, ArkStream Capital
TL;DR
-
In the Web2 startup ecosystem, AI Agent projects are primarily focused on enterprise services, while in Web3, model training and platform aggregation projects dominate due to their critical role in building ecosystems.
-
Currently, Web3 AI Agent projects account for only 8% of all AI projects but represent 23% of the market capitalization within the AI sector, demonstrating strong market competitiveness. We anticipate that as technology matures and market recognition grows, multiple projects will emerge with valuations exceeding $1 billion.
-
For non-core AI applications in Web3, integrating AI technology could become a strategic advantage. For native AI Agent projects, emphasis should be placed on holistic ecosystem construction and token economic design to promote decentralization and network effects.
The AI Wave: Current State of Project Proliferation and Valuation Growth
Since ChatGPT's debut in November 2022, it attracted over 100 million users within just two months. By May 2024, ChatGPT had reached a staggering monthly revenue of $20.3 million. Following its launch, OpenAI rapidly introduced iterative versions such as GPT-4 and GPT-4o. This rapid advancement has led major tech giants to recognize the importance of cutting-edge AI models like large language models (LLMs), prompting them to release their own AI models and applications. For example, Google launched the PaLM2 large language model, Meta introduced Llama3, and Chinese companies released models like Wenxin Yiyan and Zhipu Qingyan. Clearly, the AI domain has become a fiercely contested battleground.
The competition among tech giants has not only accelerated commercial application development but also spurred growth in open-source AI research. According to the 2024 AI Index Report, the number of AI-related projects on GitHub surged from 845 in 2011 to approximately 1.8 million in 2023. Notably, project numbers grew by 59.3% year-over-year in 2023 following the release of GPT, reflecting intense global developer engagement in AI research.
This enthusiasm for AI technology is directly reflected in investment trends. The AI investment market showed robust growth, experiencing explosive expansion in Q2 2024. Globally, there were 16 AI-related investments exceeding $150 million—twice the number in Q1. Total funding for AI startups soared to $24 billion, more than doubling year-on-year. Among these, xAI, led by Elon Musk, raised $6 billion and achieved a valuation of $24 billion, making it the second-highest valued AI startup after OpenAI.
Top 10 AI Sector Fundings in Q2 2024. Source: EqualOcean, https://www.iyiou.com/data/202407171072366
The rapid advancement of AI technology is reshaping the technological landscape at an unprecedented pace. From fierce competition among tech giants and the flourishing of open-source community projects to strong investor appetite for AI concepts, new projects continue to emerge, investment amounts keep breaking records, and valuations rise accordingly. Overall, the AI market is currently in a golden period of high-speed development. Large language models and retrieval-augmented generation (RAG) technologies have made significant progress in natural language processing. Nevertheless, challenges remain when translating technical advantages into practical products—such as uncertainty in model outputs, hallucination risks involving inaccurate information generation, and issues related to model transparency. These concerns are particularly critical in applications requiring high reliability.
Against this backdrop, we began studying AI Agents because they emphasize comprehensive problem-solving and environmental interaction. This shift marks AI’s evolution from pure language models toward intelligent systems capable of truly understanding, learning, and solving real-world problems. Thus, we see promise in AI Agents gradually bridging the gap between AI technology and practical problem-solving. While AI continues to reshape the architecture of productivity, Web3 technology is redefining the production relationships of the digital economy. When AI’s three core elements—data, models, and computing power—converge with Web3’s core principles of decentralization, token economics, and smart contracts, we foresee a wave of innovative applications. In this promising intersection, AI Agents, with their ability to autonomously execute tasks, demonstrate immense potential for large-scale adoption.
To this end, we initiated an in-depth study of AI Agents’ diverse applications across Web3—from infrastructure, middleware, and application layers to data and model markets—aiming to identify and evaluate the most promising project types and use cases to better understand the deep integration of AI and Web3.
Concept Clarification: Introduction and Classification Overview of AI Agents
Basic Introduction
Before introducing AI Agents, to help readers better grasp the distinction between definitions and model architectures, let us consider a real-life scenario: suppose you're planning a trip. A traditional large language model might provide destination details and travel suggestions. Retrieval-augmented generation (RAG) can offer richer, more specific content about destinations. An AI Agent, however, is akin to JARVIS from the Iron Man movies—it understands your needs and proactively searches for flights and hotels, executes bookings, and adds your itinerary to your calendar—all based on a simple instruction.
Currently, the industry widely defines an AI Agent as an intelligent system capable of perceiving its environment and taking corresponding actions. It uses sensors to gather environmental data, processes it, and then influences the environment through actuators (Stuart Russell & Peter Norvig, 2020). We believe an AI Agent integrates capabilities including LLMs, RAG, memory, task planning, and tool usage—an assistant that goes beyond mere information provision to plan, decompose, and actually execute tasks.
Based on this definition and characteristics, we observe that AI Agents have already permeated our daily lives and are applied in various scenarios. Examples include AlphaGo, Siri, and Tesla’s Level 5+ autonomous driving systems—all instances of AI Agents. These systems share a common trait: they perceive user inputs and respond with actions that impact the real world.
Using ChatGPT as an example for clarification: it’s important to note that Transformer is the underlying architectural framework of AI models, GPT is the model series built upon this architecture, and GPT-1, GPT-4, and GPT-4o represent different developmental stages of the model. ChatGPT, therefore, is an AI Agent evolved from the GPT model.
Classification Overview
The AI Agent market currently lacks a unified classification standard. Through tagging 204 AI Agent projects across both Web2 and Web3 markets according to their prominent features, we categorized them into primary and secondary classifications. The primary categories are Infrastructure, Content Generation, and User Interaction, which are further subdivided based on actual use cases:
-
Infrastructure: Focuses on foundational aspects of the Agent field, including platforms, models, data, development tools, and mature B2B service solutions.
-
Development Tools: Provides developers with auxiliary frameworks and tools for building AI Agents.
-
Data Processing: Handles and analyzes data in various formats, primarily supporting decision-making and serving as training data sources.
-
Model Training: Offers AI model training services, including inference, model creation, and configuration.
-
B2B Services: Targets enterprise users, offering vertical, automated enterprise solutions.
-
Platform Aggregation: Platforms that integrate multiple AI Agent services and tools.
-
Interaction: Similar to content generation, but distinguished by continuous two-way interaction. Interactive Agents not only receive and understand user requests but also provide feedback using NLP and other technologies, enabling bidirectional communication.
-
Emotional Companionship: AI Agents designed to offer emotional support and companionship.
-
GPT-based: AI Agents built on GPT (Generative Pre-trained Transformer) models.
-
Search-focused: Agents specialized in search functions, emphasizing accurate information retrieval.
-
Content Generation: Projects focused on content creation, leveraging large models to generate various forms of content based on user instructions, including text, images, videos, and audio.
Current Status Analysis of Web2 AI Agents
According to our analysis, AI Agent development in the traditional Web2 internet space shows a clear concentration trend. Approximately two-thirds of projects fall under the infrastructure category, predominantly B2B services and development tools. We conducted some analysis on this phenomenon.
Impact of Technological Maturity: The dominance of infrastructure projects is first attributed to their higher level of technological maturity. These projects typically rely on proven technologies and frameworks, reducing development complexity and risk. They serve as the "shovels" of the AI domain, providing a solid foundation for AI Agent development and deployment.
Driven by Market Demand: Another key factor is market demand. Compared to consumer markets, enterprises have a stronger urgency for AI solutions, especially those aimed at improving operational efficiency and reducing costs. Additionally, stable cash flows from enterprise clients benefit developers in sustaining long-term project development.
Limited Application Scenarios: At the same time, we observe that content-generating AI applications have relatively limited use cases in B2B environments. Due to output instability, enterprises tend to favor applications that consistently enhance productivity. This results in fewer content-generation-focused AI projects in the overall portfolio.
This trend reflects realistic considerations around technological maturity, market demand, and application feasibility. As AI technology advances and market demands become clearer, we expect this landscape may evolve—but infrastructure-level projects will likely remain the cornerstone of AI Agent development.
Analysis of Leading Web2 AI Agent Projects
Overview of Leading Web2 AI Agent Projects. Source: ArkStream Project Database
We delve into several current AI Agent projects in the Web2 market, analyzing them through examples such as Character AI, Perplexity AI, and Midjourney.
Character AI:
-
Product Overview: Character.AI offers AI-powered conversational systems and virtual character creation tools. Its platform enables users to create, train, and interact with virtual characters capable of natural language conversations and performing specific tasks.
-
Data Analysis: Character.AI recorded 277 million visits in May, with over 3.5 million daily active users, mostly aged between 18 and 34, indicating a young user base. The company performed well in fundraising, securing $150 million in funding with a $1 billion valuation, led by a16z.
-
Technical Analysis: Character AI signed a non-exclusive licensing agreement with Alphabet, Google’s parent company, to use its large language models, suggesting the company relies on proprietary technology. Notably, co-founders Noam Shazeer and Daniel De Freitas previously contributed to Google’s conversational language model Llama.
Perplexity AI:
-
Product Overview: Perplexity retrieves and delivers detailed answers from the internet. By citing references and links, it ensures information accuracy and reliability. It also guides users in refining queries and keywords, meeting diverse search needs.
-
Data Analysis: Perplexity boasts 10 million monthly active users. Its mobile and desktop app traffic increased by 8.6% in February, attracting around 50 million users. In the capital market, Perplexity AI recently secured $62.7 million in funding, reaching a $1.04 billion valuation, led by Daniel Gross with participation from Stan Druckenmiller and NVIDIA.
-
Technical Analysis: Perplexity primarily uses fine-tuned GPT-3.5 and two large open-source models customized for its needs: pplx-7b-online and pplx-70b-online. These models are suitable for academic research and vertical domain queries, ensuring high factual accuracy and reliability.
Midjourney:
-
Product Overview: Users can generate images of various styles and themes via prompts on Midjourney, covering a broad spectrum from photorealistic to abstract creations. The platform also supports image blending and editing, allowing overlay and style transfer, with real-time generation delivering results in seconds to minutes.
-
Data Analysis: The platform has 15 million registered users, with 1.5 to 2.5 million active users. Publicly available information indicates Midjourney has not taken institutional investment, achieving self-sustained growth through founder David’s entrepreneurial reputation and resources.
-
Technical Analysis: Midjourney uses its own closed-source model. Since launching Midjourney V4 in August 2022, it has employed diffusion-based generative AI. The model reportedly contains between 30–40 billion parameters, providing a solid foundation for image diversity and quality.
Commercialization Challenges
After experiencing multiple Web2 AI Agents, we observed a common product evolution path: starting with a focus on narrow, specific tasks and later expanding capabilities to handle complex multi-task scenarios. This trend highlights the potential of AI Agents in enhancing work efficiency and innovation, signaling their increasingly vital future roles. Based on preliminary statistics of 125 Web2 AI Agent projects, we found most concentrate on content generation (e.g., Jasper AI), development tools (e.g., Replit), and especially B2B services (e.g., Cresta). This finding contradicted our initial expectation—that with maturing AI model technology, C2C AI Agent applications would explode. However, analysis revealed that commercializing C2C AI Agents is far more challenging than anticipated.
Take Character.AI as an example: despite having excellent traffic performance, its business model relies solely on a $9.9 subscription fee. With minimal subscription revenue against heavy inference cost consumption, the team ultimately faced monetization difficulties and cash flow issues, leading to its acquisition by Google. This case illustrates that even with strong traffic and funding, C2C AI Agent applications face significant commercialization hurdles. Most products still fail to replace or effectively assist human labor, resulting in weak willingness among C-end users to pay. Our actual research uncovered that many startups encounter similar issues as Character.AI—C2C AI Agent development is not smooth sailing and requires deeper exploration in technological maturity, product value, and business model innovation to unlock its full market potential.
By comparing the valuations of most AI Agent projects with ceiling-level players like OpenAI and xAI, there remains room for 10–50x growth. Undeniably, the ceiling for C-end Agent applications remains high, proving it’s still a promising sector. However, considering the above analysis, we believe B2B markets may be the ultimate landing point for AI Agents. Enterprises can build platforms integrating AI Agents into vertical domains, CRM, office OA, and other management software, boosting operational efficiency while opening broader application spaces for AI Agents. Therefore, we believe B2B services will be the primary direction for AI Agent development in the traditional Web2 internet space in the near term.
Web3 AI Agent Development Status and Prospects
Project Overview
As previously analyzed, even top-funded AI Agent applications with solid user traffic face monetization challenges. Next, we analyze the current state of AI Agent projects within Web3. By evaluating representative projects—including their technological innovation, market performance, user feedback, and growth potential—we aim to extract insightful recommendations. The figure below shows several representative projects in the current market with issued tokens and relatively high market caps:
Overview of Leading Web2 AI Agent Projects. Source: ArkStream Project Database
According to our statistics on the Web3 AI Agent market, project development types also show a clear concentration pattern. The vast majority of projects fall under the infrastructure category, with notably few content-generation-focused ones. Most attempt to address model training needs by incentivizing users to contribute distributed data and computing power. Others aim to build one-stop platforms embedding various AI Agent services and tools—from development tools to front-end interactive and generative applications. Traditional AI Agent industries currently limit themselves to open-source parameter tuning or applying existing models, failing to generate significant network effects at either enterprise or individual levels.
Status Analysis
We believe this phenomenon may stem from several factors:
Misalignment Between Market and Technology: The convergence point between Web3 and AI Agents currently lacks obvious advantages compared to traditional markets. Their true strength lies in improving production relations—optimizing resource allocation and collaboration through decentralization. This may leave interactive and generative applications less competitive against technologically and financially stronger traditional rivals.
Limited Application Scenarios: Within Web3 environments, there may simply not be sufficient practical demand for generating images, videos, or texts. Instead, Web3’s decentralized and distributed nature is more often used to reduce costs and improve efficiency in traditional AI fields rather than creating entirely new application scenarios.
The root cause, we believe, traces back to the current stage of AI industry development and its future trajectory. Perhaps current AI technology remains in its infancy—akin to the transitional phase of the Industrial Revolution where steam engines gave way to electric motors—before reaching the electrified era of widespread adoption.
We have reason to believe future AI development may follow a similar path. General-purpose models will gradually stabilize, while fine-tuned models will diversify. AI applications will disperse widely across enterprises and individuals, shifting focus toward interconnectivity and interaction among models. This trend aligns closely with Web3’s ethos. Web3, known for composability and permissionless innovation, resonates with the concept of decentralized model fine-tuning. Developers gain greater freedom to combine and adjust various models. Moreover, decentralization brings unique advantages in data privacy protection and computational resource distribution, benefiting model training in ways difficult to achieve in centralized systems.
With technological advancements—especially the emergence of new techniques like LoRA (Low-Rank Adaptation)—the cost and technical barriers to model fine-tuning have significantly decreased. This makes developing public models for specific scenarios or fulfilling personalized user needs much easier. Web3-based AI Agent projects can fully leverage these advances to explore novel training methods, innovative incentive mechanisms, and new models of model sharing and collaboration—elements often unattainable in traditional centralized systems.
Furthermore, the concentration of Web3 projects in model training reflects strategic positioning within the broader AI ecosystem. Hence, the clustering of Web3 AI Agent projects in model training represents a natural convergence of technological trends, market demand, and Web3’s inherent strengths. Below, we examine several model-training-focused projects from both Web2 and Web3 for comparison.
Model Training Projects
Humans.ai
-
Project Overview: Humans.ai is a diversified AI algorithm model library and training/deployment environment covering image, video, audio, and text domains. The platform allows developers to further train, optimize, share, and trade their models. A notable innovation is Humans.ai’s use of NFTs as media to store AI models and users’ biometric data, making AI-generated content more personalized and secure.
-
Data Analysis: The market cap of Humans.ai’s token Heart is approximately $68 million. It has 56k Twitter followers, though user data remains undisclosed.
-
Technical Analysis: Humans.ai does not develop its own models but adopts a modular approach, packaging all provided models into NFTs, offering users a flexible and scalable AI solution.
FLock.io
-
Project Overview: FLock.io is an AI co-creation platform based on federated learning—a decentralized machine learning method emphasizing data privacy. It aims to solve pain points in the AI space such as low public participation, insufficient privacy protection, and monopolization by large corporations. The platform enables users to contribute data while preserving privacy, promoting democratization and decentralization of AI technology.
-
Data Analysis: Raised $6 million in seed funding in early 2024, led by Lightspeed Faction and Tagus Capital, with participation from DCG and OKX Ventures.
-
Technical Analysis: FLock.io’s architecture is built on federated learning, a privacy-preserving, decentralized method. Additionally, it employs zkFL, homomorphic encryption, and secure multi-party computation (SMPC) to provide enhanced data privacy protection.
These are model training projects from the Web3 AI Agent space. Similarly, in Web2, platforms like Predibase offer comparable model training services.
Predibase
-
Project Overview: Predibase specializes in AI and large language model optimization, allowing users to fine-tune and deploy open-source LLMs such as Llama, CodeLlama, and Phi. The platform supports various optimization techniques, including quantization, low-rank adaptation, and memory-efficient distributed training.
-
Data Analysis: Announced a $12.2 million Series A round led by Felicis. Major enterprise users include Uber, Apple, and Meta, alongside startups like Paradigm and Koble.ai.
-
Technical Analysis: Users on Predibase have trained over 250 models. The platform leverages LoRAX architecture and Ludwig framework: LoRAX enables serving thousands of fine-tuned LLMs on a single GPU, drastically reducing costs without compromising throughput or latency. Ludwig is a declarative framework used by Predibase to develop, train, fine-tune, and deploy state-of-the-art deep learning and LLMs.
-
Project Analysis: Predibase offers user-friendly features, providing customizable AI application-building services for users of all levels—whether C-end or B-end, novice or experienced.
For beginners, Predibase’s one-click automation simplifies model building and training, automatically handling complex setup and deployment steps. For advanced users, it offers deeper customization options, granting access to and control over professional parameter settings. When comparing traditional AI model training platforms with Web3-based AI projects, although their overall frameworks and logic may appear similar, we find notable differences in technical architecture and business models.
-
Technical Depth and Innovation: Traditional AI model training platforms often employ deeper technical moats, such as proprietary architectures like LoRAX and Ludwig. These frameworks provide powerful functionalities for handling complex AI training tasks. In contrast, Web3 projects may prioritize decentralization and openness over deep technical innovation.
-
Business Model Flexibility: A common bottleneck in traditional AI model training is inflexible business models. Platforms require payment for model training, limiting sustainability—especially during early stages requiring broad user participation and data collection. Conversely, Web3 projects offer more flexible models, such as token economies driven by community participation.
-
Privacy Protection Challenges: Privacy remains another critical issue. For instance, while Predibase provides virtual private cloud services on AWS, reliance on third-party infrastructure inherently carries data leakage risks.
These differentiated aspects collectively represent bottlenecks in traditional AI industries. Due to the nature of the internet, efficiently addressing these issues proves difficult. Yet, they also present opportunities and challenges for Web3—projects that successfully resolve them could pioneer the industry.
Other Types of Web3 Agent Projects
Having discussed model-training-focused AI Agent projects, we now broaden our scope to other types of AI Agent projects in the Web3 space. Although not exclusively focused on model training, these projects stand out in fundraising, exchange listings, and token market cap. Below are several representative and influential AI Agent projects in their respective domains:
Myshell
-
Product Overview: Offers a comprehensive AI Agent platform where users can create, share, and personalize AI agents. These agents provide companionship and boost work efficiency. The platform covers diverse AI agent styles—including anime and traditional aesthetics—with interactions spanning audio, video, and text. MyShell uniquely aggregates multiple existing models such as GPT-4o, GPT-4, and Claude, delivering a premium experience comparable to paid Web2 AI Agents. Additionally, it introduces an FT bonding curve trading system to incentivize creators to develop high-value AI models and allow users to invest and share in their success.
-
Data Analysis: Last funding round valued MyShell at ~$80 million, led by Dragonfly, with participation from Binance, Hashkey, Folius, and others. While exact traffic data isn’t available, MyShell has nearly 180K Twitter followers. Though Discord online users rarely exceed 10% of total followers, this suggests a loyal user and developer base.
-
Technical Analysis: MyShell does not independently develop AI models but acts as an integration platform, aggregating cutting-edge models like Claude, GPT-4, and GPT-4o, claiming compatibility with other closed-source models. This strategy allows MyShell to leverage existing technologies to deliver a unified, advanced AI experience.
-
Subjective Experience: MyShell empowers users to freely create and customize AI agents tailored to personal or professional needs across audio, video, and other modalities. Even users not deploying agents can access integrated Web2 paid models at lower costs. Moreover, by incorporating FT economic concepts, MyShell enables users not only to consume AI services but also to invest in promising agents via bonding curves, amplifying wealth effects.
Delysium
-
Product Overview: Delysium offers an intent-centric AI Agent network, aiming to deliver a seamless Web3 experience. Currently, Delysium has launched two AI Agents: Lucy and Jerry. Lucy is a connected AI Agent intended to provide utility assistance (e.g., querying top token holder addresses), though chain-intent execution is not yet live—currently limited to basic commands like staking AGI or exchanging for USDT within the ecosystem. Jerry functions as Delysium’s internal GPT, primarily answering questions about the ecosystem, such as token distribution.
-
Data Analysis: Raised $4 million in initial funding in 2022, followed by a $10 million strategic round. Its token AGI currently has an FDV of ~$130 million. Latest user data is unavailable, but official statistics indicate Lucy had accumulated over 1.4 million unique wallet connections by June 2023.
Sleepless AI
-
Product Overview: A Web3 and AI Agent-powered emotional companionship gaming platform offering virtual companion games HIM and HER. Using AIGC and LLMs, it immerses users in dynamic interactions with virtual characters. Through ongoing dialogue, users can modify character attributes and outfits. Compatible with large language models, each character iteratively learns from every conversation, becoming increasingly attuned to the user.
-
Data Analysis: Raised $3.7 million from investors including Binance Labs, Foresight Ventures, and Folius Ventures. The token’s total market cap is around $400 million. With 116K Twitter followers and 190K pre-registrations, active users reach 43K—indicating strong user stickiness.
-
Technical Analysis: Although Sleepless AI hasn’t disclosed which mainstream LLM powers its product, it trains a separate model for each character to ensure progressive understanding of the user. Combined with vector databases and personality parameter systems, characters retain memory across conversations.
-
Subjective Experience: Sleepless AI enters via AI Boyfriend and AI Girlfriend games with a Free-to-Play model, going beyond simple chatbot interfaces. High-cost artwork, continuously refined language models, high-quality voice acting, and features like alarms, sleep aids, menstrual tracking, and study companionship greatly enhance the realism of virtual beings—offering emotional value unmatched by other apps. Additionally, Sleepless AI establishes a sustainable, balanced content monetization model. Users can sell NFTs without falling into P2E or Ponzi traps, balancing player rewards and gameplay experience.
Prospect Analysis
AI Agent projects in the Web3 space span public blockchains, data management, privacy protection, social networks, platform services, and computing power. In terms of token market cap, AI Agent projects collectively reach nearly $3.8 billion, while the entire AI sector totals around $16.2 billion. This means AI Agent projects account for approximately 23% of the AI sector’s total market cap.
Despite numbering only around a dozen—few compared to the broader AI sector—their market cap share approaches one-quarter. This again confirms our belief in the substantial growth potential of this niche segment.
After analysis, we posed a core question: what traits enable certain Agent projects to secure strong funding and get listed on top exchanges? To answer this, we examined successful projects like Fetch.ai, Olas Network, SingularityNET, and Myshell.
A clear pattern emerges: these projects all belong to the platform aggregation category within infrastructure, acting as bridges—one end connecting B2B or C2B users needing Agents, the other serving developers and validators responsible for model tuning and training. Whether their products are on-chain or off-chain seems less crucial. This leads us to a preliminary conclusion: in Web3, the Web2 logic of prioritizing practical applications may not fully apply. For leading Web3 AI Agent products, building a complete ecosystem with diverse functionalities may be more critical than the standalone quality or performance of a single product. In other words, a project’s success depends not just on what it offers, but how it integrates resources, fosters collaboration, and generates network effects within its ecosystem. This ecosystem-building capability may be the key differentiator for AI Agent projects in the Web3 space.
The right way to integrate AI Agents into Web3 is not to focus narrowly on deep development of a single application, but to adopt an inclusive approach—migrating and integrating the diverse product frameworks and types from the Web2 era into the Web3 environment to build a self-sustaining ecosystem. This shift is mirrored in OpenAI’s strategic pivot this year—launching an application platform instead of merely updating models.
In summary, we believe AI Agent projects should focus on the following aspects:
-
Ecosystem Construction: Move beyond single applications to build an ecosystem encompassing multiple services and functions, fostering interaction and added value among components.
-
Token Economic Design: Develop sound tokenomics to incentivize user participation in network building, contributing data and computing power.
-
Cross-domain Integration: Explore the application potential of AI Agents across domains, creating new use cases and value through cross-sector integration.
After summarizing these three areas, we offer forward-looking suggestions tailored to different types of project teams: those developing non-core AI application-layer products and those focusing on native AI Agent projects.
For Non-Core AI Application-Layer Products:
Adopt a long-term mindset—focus on your core product while integrating AI technology, aligning with the times and waiting for the right moment. Under current technological and market trends, using AI as a traffic driver to attract users and enhance product competitiveness has become a key strategic move. While the actual long-term contribution of AI to project development remains uncertain, we believe early adopters gain a valuable window of opportunity—provided they already possess a strong core product.
In the long run, if AI achieves breakthroughs, projects that have already integrated AI will be able to iterate faster and seize opportunities to become industry leaders. This mirrors how, over recent years, live-stream e-commerce gradually replaced offline sales as a new traffic channel. Merchants with strong products who embraced this trend early quickly stood out when live commerce exploded.
We believe that amid market uncertainty, for non-core AI application-layer products, timely integration of AI Agents may be a strategic decision—not only boosting current market visibility but also unlocking new growth avenues as AI technology evolves.
For Native AI Agent Projects:
Balancing technological innovation and market demand is key to success. Native AI Agent projects must look beyond R&D to market trends. Currently, some Web3-integrated Agent projects overly focus on narrow technical paths or grand visions without matching product development. Both extremes hinder long-term viability.
Therefore, we recommend project teams ensure product quality while staying attuned to market dynamics, recognizing that traditional internet AI logic doesn’t directly apply to Web3. Instead, learn from successful Web3 projects—study their defining traits, such as model training, platform aggregation, and compelling narratives like AI modularity and multi-Agent collaboration. Crafting compelling narratives may be the key to market breakthroughs.
In conclusion, whether for non-AI-core products or native AI Agent projects, the most critical factor is identifying the right timing and technological pathway to maintain competitiveness and innovation in an ever-changing market. Teams should uphold product quality, monitor market trends, learn from successes, and innovate continuously to achieve sustained growth.
Conclusion
In closing, we analyze the Web3 AI Agent sector from multiple angles:
-
Capital Investment and Market Attention: Although Web3 AI Agent projects currently lag in listing numbers, they account for nearly 50% of market valuation, indicating strong capital market confidence. With increasing investment and attention, the emergence of more high-valued projects in this sector is inevitable.
-
Competitive Landscape and Innovation Capacity: The competitive landscape in the Web3 AI Agent space is still forming. At the application layer, no breakout leader like ChatGPT has emerged, leaving ample room for innovation and growth. As technology matures and projects innovate, the sector is poised to develop more competitive offerings, driving up valuations.
-
Token Economics and User Incentives: The essence of Web3 lies in reshaping production relations—decentralizing the traditionally centralized process of deploying and training AI models. Through thoughtful tokenomic design and user incentives, idle computing power and personal datasets can be pooled and redistributed. Combined with ZKML and other privacy-preserving solutions, this reduces data and compute costs and enables broader individual participation in AI development.
In summary, we hold a positive outlook on the AI Agent sector. We believe multiple projects in this space will surpass $1 billion in valuation. Horizontal comparisons show the narrative is compelling and the market potential vast. Current valuations are generally low. Considering the rapid advancement of AI technology, growing market demand, increasing capital inflow, and the innovative potential within the sector, we expect numerous projects to emerge with valuations exceeding $1 billion as technology matures and market recognition strengthens.
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
Arkstream Capital
