Cycles and narratives have always been core topics in the global crypto market. Historically, the industry often referenced Bitcoin halving cycles to gauge trends and uncover major narratives. However, with the approval of Bitcoin and Ethereum spot ETFs, the crypto market has become highly coupled with global financial markets, and the variables affecting crypto market trends are increasing.
In this context of increased uncertainty, it’s crucial to have a clearer understanding of cycles and to identify future narrative trends. Investment institutions, as pioneers in capturing innovative narratives, have always had a keen sense of these developments. With this in mind, OKX created the Crypto Evolution Theory series, inviting leading global crypto investment institutions to discuss topics including current market cycles, emerging narratives, and popular sub-sectors, providing insights to explore further.
The following is our third issue in the series, where OKX Ventures, Polychain, and Delphi Digital discuss the integration of AI and crypto. We hope the insights and perspectives shared here inspire you.
About OKX Ventures
OKX Ventures is the investment arm of OKX, a leading crypto asset trading platform and Web3 technology company, with an initial capital commitment of $100 million. OKX Ventures focuses on exploring the best blockchain projects globally, supporting cutting-edge blockchain technology innovation, promoting the healthy development of the global blockchain industry, and investing in long-term structural value.
Through our commitment to supporting entrepreneurs who advance the blockchain industry, we help to build innovative companies and bring global resources and historical experience to blockchain projects.
About Polychain
Polychain is an investment firm committed to exceptional returns for investors through actively managed portfolios of these blockchain assets.
About Delphi
Delphi Digital is a research-driven firm dedicated to advancing the understanding and development of the growing digital asset market. The firm supports the ecosystem through four lines of business: Delphi Research, Delphi Ventures, Delphi Creative, and Delphi Labs
I. When Crypto Meets AI
OKX Ventures (Researcher): The development of AI technology is currently heavily reliant on a push from major players like OpenAI, Google, and Nvidia. Nvidia controls the "power" of the entire AI era, while OpenAI and Google hold the most crucial data and technological solutions. This centralized, heavily giant-dependent state can limit innovation and development in the industry.
However, the decentralized, permissionless nature of crypto can break the shackles of these giants, potentially fostering technological innovation and bringing new prosperity to the industry. Currently, common focus areas include computing power, data, models, and applications.
Computing power
Distributed or decentralized computing power markets like io.net and Prodia use idle computing power from global markets, breaking the giants' monopoly over computing power as a product.
We’re eager to see what will happen if the total supply of distributed computing power surpasses that of centralized computing power in the future. Meanwhile, due to the scarcity and high profitability of AI computing power assets, RWA (real world asset) projects like Compute Labs have emerged. These projects tokenize computing power assets and develop related derivatives, creating an AI-Fi ecosystem.
Data
Crypto's economic models can effectively incentivize user participation in the AI data field. For example, various Depin projects can incentivize users to contribute, label, or validate data through token economic models, providing data sources for AI model training. Projects like 0g.ai have built a scalable data availability layer and storage system. Additionally, Crypto's privacy protection features can better preserve the security and privacy of user data. Projects like Flock.io and Privasea.ai emphasize the importance of protecting user data privacy during model training.
Models
Open model markets have the potential to break the monopoly tech giants hold over AI models. Users can not only provide computing resources to support AI model training and inference but also offer data or models for direct interaction via network protocols. However, distributed model training remains a challenging area, and we’re particularly eager to see technological breakthroughs in this space. We also hope to see entrepreneurial teams fill this gap in the near future.
Applications
Looking at applications, the combination of AI and crypto creates new opportunities for content creation. Users can independently build virtual characters and chatbots with customized personalities, like Myshell, where users train models by uploading data to create their own AI Smart Agents. This also allows data providers and model trainers to benefit from the platform's development, forming a positive data flywheel.
Polychain Capital: The landscape of AI is undergoing a significant transformation, moving from closed-source models to increasingly sophisticated open-source alternatives. While this shift has democratized access to AI capabilities, it’s also introduced new challenges, particularly in value capture for model creators. The financialization of open-source models is one of the areas where the intersection of crypto and AI presents new opportunities.
Blockchain technology enables provenance, ownership, and verifiability. When combined with a token, the technology creates a foundation for value accrual that wasn’t possible before. For example, Ora's Initial Model Offering (IMO) shows how tokens can represent AI models, allowing for token owners to make gains when these models are tradeable. This approach not only incentivizes open-source development but also ensures fair compensation for creators and contributors.
Beyond financialization, the convergence of crypto and AI is driving innovation in public governance and system transparency. As concerns grow over potential biases and centralized control in AI models, blockchain-based solutions offer decentralized training, inference, and governance mechanisms. These systems can provide transparent decision-making processes and allow for community input in model development and deployment.
However, the true leading edge of innovation lies in the development of the underlying infrastructure. We're seeing advancements in distributed compute networks, training algorithms designed for distributed networks, novel data ownership mechanisms, and new token standards that enable model ownership with revenue sharing. These infrastructural developments are laying the groundwork for more sophisticated applications of crypto and AI.
One particularly promising direction is the emergence of AI agents and executable task systems. This concept envisions AI agents or networks of agents as extensions of individuals, capable of performing complex sequences of tasks autonomously. The potential applications range from personalized digital assistants to advanced automation in DeFi and ultimately fully autonomous beings that can create their own value. However, the realization of this vision hinges on several critical foundations, namely robust privacy protections for user data, verifiable computation systems, especially for tasks with governance or financial implications, and seamless integration with the infrastructure layers.
The development of projects that combine crypto and AI is still in its early stages. But, they’re making rapid progress through testing and improvements. While definitive best practices aren’t yet established, the potential for crypto to address some of AI's most pressing challenges is becoming increasingly evident.
As we move forward, we expect to see more refined applications that leverage the strengths of both crypto and AI. This convergence is likely to yield systems that are not only more transparent and accountable, but also significantly enhance the usability and functionality of both AI and blockchain technologies. The journey ahead promises exciting developments as we continue to explore and expand the boundaries of what's possible at the intersection of these transformative technologies.
II. Unraveling the Investment Methodology for Crypto and AI
OKX Ventures: We can approach this topic by looking at the current development trends in this sector.
From hype to substance
The sector is currently moving from hype to substance. Over the past year, many crypto and AI projects have emerged, mostly in the infrastructure space, with fewer applications. The application layer has often been superficial, lacking innovation, and low in technical content, with a significant portion driven by hype rather than genuine progress. Hype and bubbles are typical in the early stages of technological innovation, but as market resources are optimized, we’ll see truly tech-savvy entrepreneurial teams enter the crypto and AI field. The market will begin to favor projects that can provide real value, scalability, and usability, rather than those relying solely on hype and marketing.
From speculation to demand
The market will shift from speculation-driven to demand-driven, with the focus moving from potential speculative value to actual usage and adoption. Entrepreneurs can no longer rely solely on narratives to attract investors. The market is now more careful and conservative towards purely narrative-driven projects. In the future, projects with real market demand and business revenue will become a necessity for investors. This is a fundamental logic we follow when investing in the crypto and AI sector.
Based on the trends above, we’ve organized our investment logic into three core points.
Market demand orientation
Many AI startups only realize after launching their product that the market doesn’t buy into it, meaning users have no demand for their product. This often stems from inadequate market research at the outset of the startup, where the product wasn’t oriented toward actual market demand or was based on an incorrect assumption of unverified pseudo-demands. Therefore, when positioning ourselves in the crypto and AI sector, we place great emphasis on meeting market demand. Firstly, we assess the general direction to determine which specific sub-sector of crypto and AI the project belongs to, the market size, the potential for future growth, and the competitive landscape. Secondly, we consider what problem the project solves and what demand it meets. Even if it addresses a small issue, solving a market pain point is a feasible approach.
Not just narratives
The crypto and AI field is often criticized for being overly narrative-driven with little real application. While we don’t entirely agree with this perspective, it’s true that the market is increasingly unwilling to buy into pure narratives. Therefore, real business scenarios and business models are particularly important. The startup must generate business revenue to be financially sustainable. Many startups rely solely on NFT or token sales as their only source of income, which is unsustainable. The team must have a clear business model, knowing exactly where their revenue will come from, rather than depending solely on narratives to attract market investment.
The team needs an AI background
The rapid rise of AI has quickly ignited the enthusiasm of the Web2 market and VCs, and this wave has naturally spread to the crypto world. Many crypto startups have begun to ride the AI trend and repackage their offering, leading to a surge in crypto and AI projects. However, because many teams lack a technical AI background, these projects are often superficial, lack competitiveness, and are quickly eliminated by the market. AI has a high technical threshold, especially when combining crypto and AI. A deep understanding of both fields is necessary to effectively merge them. Otherwise, it’s difficult to gain market acceptance.
In summary, the basic investment approach is to identify market demands and problems within a high-potential sector, and to find the most suitable team. By providing support to entrepreneurs, we aim to help them grow from zero to one, meeting market demand and solving problems.
Polychain Capital (Sven): The current landscape of crypto and AI projects is undeniably narrative-driven, a characteristic typical of early-stage, transformative technologies. This narrative focus isn’t just marketing, but a necessary part of the ecosystem's evolution. It serves to attract attention, foster community and engagement, and drive the initial stages of adoption.
However, we know it’s important to look beyond the narrative to see the technology behind these projects and how it can be applied in real life. Our investment thesis in the crypto and AI space comes from deep research and a comprehensive understanding of both technologies and their potential synergies. We place a premium on projects that not only captivate with their vision but also demonstrate a clear path to market adoption and a solid technological foundation. This approach requires a broad yet deep research methodology, which has allowed us to filter the noise and identify projects we think are truly innovative.
The current investment landscape spans various layers of the crypto and AI stack. At the infrastructure level, we've seen promising developments in GPU networks, distributed training, inference and intelligence networks, verifiable and private computation, and data management solutions. While many of these projects are still in their building phase, they lay the crucial groundwork for the next wave of innovation.
Looking ahead, we anticipate several key trends shaping the future of crypto and AI. We have a strong conviction that blockchain systems and cryptocurrencies provide the ideal framework for AI agents to operate autonomously and execute complex tasks. Data privacy continues to be a central problem, and as cryptographic techniques such as Homomorphic Encryption, Multi-Party Computation, and Zero-Knowledge Proofs improve in efficiency, they start to become real options to support privacy-preserving AI solutions alongside more machine learning approaches, such as Differential Privacy and Federated Learning.
We expect continued growth in decentralized data marketplaces, verifiable inference networks, decentralized training and fine-tuning platforms, and AI agent infrastructure. These developments could democratize access to AI capabilities and create fair, transparent, and efficient systems. The integration of AI into existing blockchain platforms has the potential to drive the next wave of crypto adoption. This could manifest in AI-powered analytics for DeFi, predictive models for active asset management, risk models, and enhanced governance mechanisms for DAOs. The trend towards smaller, more efficient models trained on high-quality datasets is likely to continue. This shift could enable more personalized AI experiences and reduce friction in various applications
It's important to note that while we can identify these trends and investment criteria, the crypto and AI space remains highly dynamic. The infrastructure supporting many of these innovations is still under development, and we expect to see numerous iterations and improvements. This evolving landscape, while potentially noisy, presents unique investment opportunities for those willing to dive deep and maintain a long-term perspective.
The convergence of crypto and AI offers fertile ground for transformative innovation and investment opportunities. As the space matures, we expect a shift from purely narrative-driven projects to those that demonstrate real-world impact, robust technological foundations, and sustainable growth models. It’s in this transition that the most compelling investment opportunities are likely to emerge.
Delphi Digital (Pondering): Software is eating the world, and AI is eating software. The essence of AI lies in data and computation. Therefore, those who can most effectively acquire these two key inputs (infrastructure), coordinate them (middleware), or use them to meet user needs (applications) could capture immense value.
Delphi Ventures is actively investing in projects across each layer of the DeAI stack. Firstly, at the infrastructure level, DeAI relies on data and computation, particularly through cryptocurrency incentive mechanisms to efficiently acquire these resources. This is the most challenging yet potentially rewarding part of the stack. Currently, distributed training protocols and the GPU market provide cost-effective solutions by coordinating heterogeneous hardware, while DePIN networks, with their ability to build hardware networks at low cost, are poised to play a significant role in the future intelligent economy.
Secondly, at the middleware level, DeAI aims to achieve efficient composable computing, similar to the "Lego" model of DeFi. We’re particularly bullish on efficient routing mechanisms, for example, how to select the most cost-effective and performant model for the right use case. We’re also optimistic about Graph Neural Networks, Co-Processors to scale data and compute in restrained onchain environments, and crypto-based mechanisms that solve the incentive problem for developers in open source. If applied correctly, DeAI middleware presents compelling possibilities for a modular approach to AI, which may eventually outcompete the integrated, closed-source versions of today’s tech giants.
Finally, at the application level, onchain agent protocols may be key to improving the user experience in the crypto space. By connecting computing networks and users, these protocols not only reduce costs but could also drive demand for Web3 infrastructure, fostering new economic models.
Overall, AI will profoundly change our economic landscape. While the current narrative around DeAI might be overly optimistic, the scale of opportunities is enormous. For those with patience and insight, the true vision of composable computing in DeAI may well validate the vision of blockchain itself.
III. Discussing Future Opportunities
OKX Ventures (Researcher): Technological breakthroughs and innovations are eternal opportunities.
In the AI field, there’s a significant tech monopoly. Data and core technologies are largely being controlled by tech giants, which greatly squeezes the survival space for startups. Addressing the technological monopoly of these tech giants is a primary challenge that entrepreneurs need to confront. We eagerly anticipate seeing more entrepreneurial teams step out of the follower role and, through the integration of crypto and AI, break the monopoly of centralized tech giants. In doing so, they can achieve technological breakthroughs and innovations, and truly implement narratives and products to meet market demands.
How to stay in the game is a question that entrepreneurial teams need to consider.
Business model and sustainability: Entrepreneurial teams need to explore their business model and its sustainability. Projects based solely on narrative are no longer accepted by the market. Teams need stable business revenue or a clear and viable monetizable business model for the future.
Financial management and cost control: Entrepreneurial teams need to have sound financial management and cost control skills to preserve the long-term stability of the project. Financial issues are among the most common problems for startups, and many projects fail due to poor financial management.
Flexibility and agility: Entrepreneurial teams need sufficient flexibility and agility. The market evolves rapidly, and a single technological breakthrough can cause many startups to disappear. Teams must be adaptable and adjust their strategies and directions in response to market changes, learning to time their moves and take opportunities effectively.
Polychain Capital (Sven): In terms of opportunities, we're witnessing a significant shift in sentiment across both the AI and crypto sectors. The crypto market has experienced improved perception from institutions and regulators, as shown earlier in 2024 with the approval of Bitcoin and Ethereum ETFs in the U.S. and more recently the outcome of the U.S. elections. This progress signals a sentiment shift for the crypto community as mainstream acceptance grows and opens doors for further innovation.
At the same time, the AI landscape is experiencing its own transformation. Concerns surrounding leading AI companies have emerged, as seen with the founding members of OpenAI leaving to fulfill the idea of "superalignment" and the arms race for the physical infrastructure that powers AI. This has created an opening for new, innovative approaches to AI development and governance. This shift aligns well with the ethos of decentralization central to many crypto projects, presenting a unique opportunity for synergy.
The demand for AI solutions remains high across various sectors, from corporations to individual users. However, the field is still nascent and wide open for innovation, with no clear dominant strategy established. Combined with the idea of a fairer and more open AI, this creates fertile ground for projects that can effectively combine the strengths of both AI and blockchain technologies. The idea of “superalignment” tackles concerns around AI's accelerating intelligence and making sure that these systems act in a way that aligns with human values and goals. Alongside alignment, there’s an increasing recognition of the need for user contribution and ownership in AI development. Crypto projects that allow for this ownership and alignment are already gaining significant mindshare.
However, substantial challenges counterbalance these opportunities. The global economy faces significant headwinds, including ongoing conflicts, recession warnings, high inflation, and high interest rates. These factors contribute to a careful spending environment and may impact investment in alternative assets like cryptocurrencies. Interestingly, this challenging environment might also drive mindshare towards crypto as an alternative to the traditional financial system. This can be seen in Bitcoin, as it is increasingly viewed as ‘digital gold’ and a potential store of value during uncertain times.
Regulatory uncertainty always remains a concern. The legal landscape for both crypto and AI is in flux, with different jurisdictions implementing their own approaches. Projects in either space have to navigate this uncertain terrain and always be agile. Talent acquisition presents another critical challenge for projects. The scarcity of individuals with deep expertise in both AI and blockchain technologies has led to fierce competition. This talent crunch can significantly impact project development and innovation, potentially slowing progress in this rapidly evolving field.
Looking ahead, the current market cycle is likely to serve as a filtering mechanism for crypto and AI projects. Those that can effectively address real-world needs, adapt to regulatory challenges, and genuinely integrate these technologies are poised to emerge as leaders in the next phase of industry development. As the market matures, we can expect to see a shift towards more sustainable and practical applications of crypto and AI technologies. This evolution will likely be characterized by increased focus on user ownership and data rights in AI systems, the development of robust decentralized AI infrastructure, integration of AI capabilities into existing blockchain ecosystems, and the emergence of new economic models enabled by the synthesis of AI and blockchain technologies.
Delphi Digital (Pondering): The biggest challenge for DeAI lies in the infrastructure layer, particularly in the capital intensity required to build foundational models and the returns of scale in data and compute.
Large tech companies have a clear advantage in this regard. By leveraging monopoly profits from the second generation of the internet to build substantial capital reserves and reinvesting these during a decade of low interest rates into cloud infrastructure, they’re now attempting to monopolize the essential inputs — data and compute — in the market for intelligence.
Due to the capital demands and high bandwidth requirements of large-scale training, superclusters remain the optimal choice, providing large tech companies with the most powerful and usually closed-source models. Many companies plan to lease these models for monopoly profits and reinvest the returns. However, the pool in AI is shallower than the network effects seen in Web2. The value of cutting-edge models is rapidly depreciating, especially with Meta's shift to a scorched earth strategy, investing billions of dollars in developing high-performance open-source models like Llama 3.1.
With the rise of low-latency decentralized training methods, the trend towards the commodification of cutting-edge models is becoming apparent. This shift moves competition from hardware superclusters controlled by large tech companies to an environment more favorable to open-source and crypto software innovation. Simultaneously, the price of accessing intelligence is decreasing rapidly.
We explore the tension between big tech and DeAI in our DeAI series reports, which interested readers can access for free.
Given the computational efficiency of “Mixture of Expert” architectures and large language model integration and routing, the future is unlikely to be dominated by a few supermodels. Rather, we expect influence to be held by an intelligent network composed of millions of models and agents of varying shapes, sizes, and use cases. This presents significant coordination challenges, but equally provides a major opportunity for blockchain and crypto incentives to fill the void, enabling the vision of a truly open and composable network of the resources set to govern the 21st century.
OKX Ventures Disclaimer: Please see OKX Disclaimer.
Delphi Digital Disclaimer: Please see Delphi Digital Disclaimer.
© 2024 OKX. Este artículo puede reproducirse o distribuirse en su totalidad, o pueden utilizarse fragmentos de 100 palabras o menos de este artículo, siempre que dicho uso no sea comercial. Cualquier reproducción o distribución del artículo completo debe indicar también claramente, lo siguiente: "Este artículo es © de OKX 2024 y se utiliza con permiso". Los fragmentos permitidos deben citar el nombre del artículo e incluir su atribución, por ejemplo "Nombre del artículo, [nombre del autor, en su caso], © 2024 OKX". No se permiten obras derivadas ni otros usos de este artículo.
