Since the beginning of this year, the combination of AI and cryptocurrency has moved beyond mere "narrative" and has shown more observable progress along three main lines. The first is the decentralized supply of computing power: during the high-end GPU cycle dominated by Nvidia, new types of "AI clouds" and computing power contractors have rapidly expanded, highlighting the bargaining power and infrastructural status of computing power as a scarce resource, providing a realistic foundation for the on-chain computing power market. A recent Bloomberg report has spotlighted the relationship between this "neocloud" and high-priced GPUs, indirectly confirming the cost and supply logic of decentralized network entry.
In response, several native cryptocurrency projects have moved the "GPU supply - AI workload" onto the chain: Render Network continues to promote decentralized rendering and AI workloads, strengthening the matchmaking between nodes and creators/developers; Akash emphasizes in its 2025 roadmap and quarterly research that "agents can autonomously schedule computing power," and is advancing support for the next generation of GPU architectures, attempting to establish open computing power as a long-term effective market mechanism. These developments make the idea of "outsourcing parts of AI training and inference to the chain" more engineering feasible.
The second main line is the pilot projects and regulatory impacts of "machine identity" (proof-of-personhood/World ID, etc.) on mainstream platforms. Semafor disclosed that Reddit had explored introducing a World ID based on iris scanning for "both anonymous and unique" user verification, which means that in the vast ocean of AI-generated content, using "verifiable humanity" to filter and divert is becoming a demand that is moving towards contextualization. However, at the same time, China's national security department has issued warnings regarding biometric data collection similar to Worldcoin from a national security perspective, and earlier scrutiny in Europe and Kenya has not faded. The global multiple compliance thresholds for identity on-chain are becoming key to whether "human-machine distinction" infrastructure can scale in the AI era.
The third main line is the organization of "decentralized AI alliances and agent networks." The Artificial Superintelligence (ASI) alliance, which will complete its merger in 2024 and continue operations in 2025, is consolidating resources from Fetch.ai, SingularityNET, and Ocean Protocol, aiming to form a decentralized AI/data/computing power collaborative stack. The signal released by such attempts is that, at the model, data, and execution layers, an open network incentivized by tokens needs to complement or even compete with large centralized AI. Regardless of short-term market fluctuations, the merger itself provides a clearer structural boundary for the "multi-entity AI - data market - execution environment."
The regulatory and policy environment is a barometer for whether this wave of integration can develop in a stable state this year. The policy document released by the White House in January explicitly stated "support for the responsible growth of digital assets and blockchain," signaling a more positive tone from the federal level towards industrial innovation; however, sharp divisions remain in Congress regarding new legislation, with some lawmakers concerned that weakening the SEC's power and shifting to futures regulatory bodies could bring systemic risks. In other words, the industrial side of "technology - market - application" is accelerating, while the institutional side is still in a tug-of-war over "clarifying boundaries - redistributing rights and responsibilities."
For capital and developers, the intersection of these three main lines is sketching a more realistic path to implementation: — On the computing power side, decentralized networks are not meant to replace large cloud providers but to fill the gaps of "marginality, elasticity, and regional diversity," especially for mid-to-long tail inference and batch rendering tasks; however, the challenge lies in service consistency and SLA verifiability, requiring a more standardized measurement and settlement layer. Bloomberg's observation on "new clouds and GPU cost curves" reminds us that the supply side may remain in a cyclical fluctuation of "high prices - scarcity - integration" for a long time.
— On the identity side, to enter high-concurrency scenarios such as social networking, content distribution, and e-commerce, a compliance project of "zero-knowledge proof + minimal data collection" must precede. Regulatory signals from China and Europe indicate that global expansion requires "partitioned compliance" in product design, rather than a one-size-fits-all approach.
— On the agent and protocol side, the alliance integration reduces the friction of "reinventing the wheel," but cross-domain collaboration (data licensing, model invocation, settlement, and accountability) still requires stronger open standards and end-to-end verifiable toolchains.
Looking ahead to the second half of the year and into 2026, three "observation points" are worth tracking:
Hardware and supply chain: If high-end GPUs from companies like Nvidia continue to be in tight supply and demand, the cost-effectiveness window for decentralized computing power networks remains, but caution is needed regarding the "dark pool" risks caused by varying node quality.
Adoption by large platforms: If platforms like Reddit formally introduce "human-machine distinction" encrypted identity modules, it will become a key watershed in the governance of the AI content ecosystem.
Regulatory collaboration: The encouragement from the U.S. executive branch coexists with divisions in Congress; if a clearer dual regulatory framework can be formed around the boundaries of "goods/securities" and token distribution, the compliance space for open AI-crypto stacks will expand accordingly.
In conclusion: The integration of AI and cryptocurrency is transitioning from "story" to "infrastructure": the computing power market addresses supply, machine identity addresses trust, and agent networks connect applications. However, this path is not linear—supply chain cycles, regulatory redistribution, and privacy ethics will repeatedly pull at the rhythm. For participants, seeking "verifiable efficiency improvements" within factual constraints may be the most prudent strategic benchmark for 2025.
Related: 57% of Kalshi bettors predict Gemini will become the top AI text model in 2025.
Original article: “Three Pillars of AI on Chain: Compute Markets, Machine Identity, and Agent Networks”
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。
