In the rapidly evolving cryptocurrency space, a new sector has emerged: the artificial intelligence (AI) agent crypto token market. This industry has grown significantly over the past year, reaching a market capital capitalization of US$13.5 billion by early 2025.

Though the sector’s early days were dominated by token speculation and hype, the sector is now entering a phase of maturity grounded in real value validation, data-backed achievements and sustainable revenue models. According to HTX Ventures, the investment arm of crypto exchange HTX, the sector’s next growth phase could be propelled by technological innovations from Chinese AI company DeepSeek.

AI agent crypto tokens are a novel concept that combines elements of blockchain technology and AI. These tokens are typically associated with projects or platforms that utilize AI agents within a decentralized ecosystem.

They often serve as utility tokens, granting access to AI-powered services or rewarding participants for contributing to network development. They may also facilitate seamless interaction between different AI systems and blockchain networks.

In some cases, AI agent crypto token holders may have voting power in decision-making processes, further integrating AI and blockchain technologies.

Reinforcement learning: a game-changer for cost and efficiency

Though AI agent crypto tokens initially gained traction through viral buzz, the sector is now at a crossroad, requiring renewed momentum. HTX Ventures believes that the groundbreaking advancements from DeepSeek can help trigger this next wave of innovation and drive growth for the sector.

In a new report released in March 2025, the investment firm delves into the technological developments made by the Chinese AI company, outlining how its technology is primed to lead a new era of crypto-tech development over the next few years.

Founded in 2023 as a spin-off from High-Flyer Capital Management, a Chinese hedge fund that uses AI for trading, DeepSeek is a pioneer in the development of advanced AI models.

One of its key advancement is its R1 reasoning model, which launched in early 2025. This model has attracted significant attention due to its remarkably low cost of operation and open-source nature, which allows other companies to test and build upon the model to improve it.

Unlike traditional AI models, which rely heavily on expensive supervised fine-tuning and human-labeled data, R1 uses reinforcement learning (RL), a type of machine learning (ML) where an AI model learns by interacting with an environment and receiving rewards or penalties based on its actions.

This approach brings several advantages. First, the model can be trained with significantly less data and computational resources, resulting in lower operational costs. This is particularly important in the AI space, where training large models often incurs high costs related to GPUs and cloud services.

Second, RL models are able to develop reasoning abilities and complex behaviors that may not be explicitly programmed since they learn from their interactions and rewards. This makes them highly reliable in fields like physics, science, and mathematics, where logical reasoning is crucial.

DeepSeek poised to boost the AI agent crypto market

According to HTX Ventures, these capabilities could revitalize the AI agent crypto sector by offering better technical tools and frameworks. These improvements would enable AI agents to perform more sophisticated tasks, such as automated trading, cross-chain interactions, and decentralized finance (DeFi) applications, effectively bridging the gap between AI and practical applications in the crypto space.

Ultimately, this evolution would enable AI agents to become commercially viable and sustainable business models that address real user needs, rather than relying on hype. This focus on sustainability over short-term gains will help maintain AI agents’ relevance in the crypto space, the report says.

HTX Ventures identifies several key trends to watch in the AI agent crypto space. In particular, the combination of multi-agent systems and DeFI + AI (DeFAI) showcases the potential of AI in complex decisions and financial automation.

In addition, upgrades to core AI frameworks like Eliza and ARC, are expected to unlock new use cases and business models, making them essential to watch.

Finally, new infrastructure projects, including AI models, data layers, and computational infrastructures designed specifically for crypto scenarios, could become bedrocks supporting the next-generation of AI agents, improving model efficiency and intelligence.

The AI agent crypto token craze

The AI agent crypto token craze began in mid-2014 when Truth Terminal, an autonomous chatbot developed by AI researcher Andy Ayrey, managed to convince venture capitalist (VC) Marc Andreessen to donate US$50,000 in bitcoin to its wallet.

This funding sparked Truth Terminal’s ambition to launch its own token, eventually leading to the creation of the meme coin Goatseus Maximus (GOAT) by an anonymous developer. GOAT quickly shot to a US$1 billion valuation, showcasing the potential of AI and viral marketing strategies in driving speculative trends in the crypto sphere.

Since then, a number of AI agent tokens have emerged, with popular ones that include:

• The Artificial Superintelligence (ASI) Alliance, a decentralized AI consortium founded by three leading blockchain-AI projects, namely Fetch.ai, SingularityNET, and Ocean Protocol;
• Virtuals Protocol, an ecosystem of AI agents designed generate services or products and autonomously engage in commerce onchain. These agents are tokenized, allowing for capital formation, permissionless participation, and aligned incentives among creators, investors, and agents;
• OriginTrail, a platform that uses both a blockchain and a technology called a decentralized knowledge graph (DKG) to manage data about real-world assets in the digital space; and
• AI16Z, a VC decentralized autonomous organization (DAO) led by AI agents. The project is led by an AI agent and aims to leverage AI and collective intelligence to make investment decisions.

Fetaured image credit: edited from freepik