In 2026, DePIN's evolution in the AI computing field is shifting from token subsidies to protocol revenue and engineering stability verification.
Written by: Zeuspace Yao Kun
In 2026, DePIN's evolution in the AI computing field is shifting from token subsidies to protocol revenue and engineering stability verification. The transition of AI computing from training to inference has opened a structural window for DePIN, but the sector's market value has retraced over 80% from its historical peak, and issues with enterprise-level SLA and scheduling friction remain unresolved, indicating that this track is between "proof of concept" and "scalable commercial use."
The DePIN sector is undergoing a valuation paradigm shift: token prices have plummeted, but real revenues have rebounded. As of mid-July 2026, the total market value of the sector is approximately $3.46 billion, down about 83% from the peak of $20.2 billion in March 2024, with a cumulative decline of 23% within the year. In contrast to the secondary market performance, industry protocol revenues have not shrunk in tandem—by January 2026, DePIN projects had achieved monthly on-chain revenues of $150 million, derived from actual fees paid by enterprises for computing, storage, and connectivity, rather than token emissions or speculative incentives.
Among them, GPU infrastructure provider Aethir leads with $55 million, Render Network contributes $38 million, and Helium contributes $24 million. The market value retracement does not indicate the failure of the track but rather a valuation correction as the market shifts from "expectation-driven" to "revenue-driven."
The structure of AI computing is shifting from training-dominated to inference-dominated, which is highly compatible with DePIN's distributed form. Yang Yuanqing, Chairman and CEO of Lenovo Group, pointed out at the performance conference in May 2026 that currently about 70%-80% of AI computing power is used for training, while 20%-30% is used for inference, but "in the future, this trend will reverse, with AI computing power for inference accounting for over 70%." Training tasks rely on ultra-low latency and highly centralized computing clusters, which remain the moat of traditional clouds; inference tasks, on the other hand, are characterized by high dispersion, high concurrency, and geographical sensitivity, which naturally aligns with DePIN's node form.
Small and medium-sized AI studios, game rendering teams, and agent service providers can access computing power on demand by the second without signing annual commitment contracts. However, favorable demand does not automatically translate—whether the supply side can deliver stably is key to commercialization.
Nominal price advantages do not equate to real cost advantages; SLA, development friction, and procurement barriers constitute three engineering thresholds. By mid-2026, the hourly rental prices for H100 on the DePIN platform are generally lower than those of traditional cloud providers: Akash Network is about $2.30-$3.68/hour, io.net is about $1.85-$2.69/hour, and Aethir is about $1.90-$3.10/hour. In contrast, AWS's H100 (P5 series) rates in the U.S. region are about $5.19/hour, while non-U.S. regions are $4.72/hour; Google Cloud's H100 is about $11.06/hour; traditional cloud providers' price ranges are about $3.93-$11.06/hour. However, enterprises do not only look at unit prices when purchasing.
First, there is a gap in SLA guarantees: Although leading projects have emphasized enterprise-level service goals, most distributed nodes face significant risks of task interruption due to unstable home bandwidth or power supply, which is still significantly higher than the 99.99% guarantee of traditional clouds, requiring developers to reserve redundant nodes to offset price savings. Second, scheduling and engineering friction have not been eliminated: Heterogeneous GPU load balancing, cross-regional data transfer, etc., need to be handled independently, consuming additional R&D hours. Finally, institutional barriers in the procurement process remain evident: Enterprises are accustomed to fiat currency settlements, standard contracts, and auditable invoices, while crypto-native networks are still catching up. Therefore, DePIN's realistic positioning is as a "cloud supplement" rather than a "cloud replacement," especially suitable for edge computing and burst inference.
Leading protocols are shifting from "high emissions for supply" to a new paradigm of "revenue-driven token destruction," but value capture still needs validation. io.net launched the Incentive Dynamic Engine (IDE) in June 2026, promising to use at least half of the network revenue (after deducting provider shares) for the permanent destruction of IO tokens, with an expected destruction of at least 12 million tokens within the next year. The project has signed an enterprise contract worth $8 million, contributing about $650,000 in monthly on-chain revenue; simultaneously, it processes over 4 billion AI inference tokens daily through OpenRouter.
Aethir has disclosed an annualized revenue scale of over $100 million (with different sources reporting figures ranging from $113 million to $147 million), covering 93 countries. The $260 million B300 cluster contract signed by Nasdaq-listed Axe Compute is also delivered by the Aethir network. Tokens are transitioning from "subsidy tools" to "value capture interfaces"—driven by revenue destruction rather than new emissions, the supply-demand relationship is shifting from dilution cycles to deflationary spirals. However, risks remain: whether tokens can truly capture value ultimately depends on revenue sustainability and customer retention rates.
DePIN is neither a defunct bubble nor has it become a mature enterprise-level infrastructure. The market value has dropped from $20.2 billion to $3.46 billion, reflecting the market's correction of valuations lacking revenue support. Monthly on-chain revenues of $150 million, annualized revenues exceeding $100 million for leading projects, and multi-million dollar enterprise contracts prove that decentralized computing is becoming a real supply layer in the AI industry chain.
However, this "reality" is still conditional—it has found the most suitable scenarios for inference tasks, but to bridge the gap of enterprise-level deployment, continuous efforts are needed in SLA standardization, development toolchains, and procurement compliance. The next stage will no longer be about who is cheaper, but whether stability and sustainability can approach enterprise standards. When developers only care about whether computing power is cheap, stable, and user-friendly, without concern for whether it comes from centralized or decentralized networks, DePIN will have truly completed its transformation. In 2026, this process has just passed halfway.
This content is provided for general informational purposes only and doesn't constitute financial, investment, legal, or tax advice. Any events, rewards, online promotions, or related information mentioned herein should not be considered a recommendation, solicitation, or invitation to purchase, sell, trade, or otherwise deal in any crypto assets. Crypto assets are highly volatile and may result in loss. The availability of WEEX services, products, and related events may vary by region. You are responsible for ensuring that your participation is in accordance with applicable local laws and regulations.



![[Editorial] The Rails Are Laid Before the World Notices](/public-static/7_ca1b7746d1.png?format=avif)















Since 2018, WEEX has built its trading experience on three pillars — security, cross-asset access, and liquidity. Here's how a 1,000 BTC Protection Fund, 330+ TradFi pairs, and Top-2 BTC futures liquidity translate into faster fills, safer trades, and more markets in one account.










