130 crypto AI projects generated just $8.7 million in model fees over 30 days, despite severely inflated token counts. 2026-10-21 19:06:54 Key takeaways. * Venice accounted for 83% of fees across 130 crypto AI projects over a 30-day period, amounting to about $7.2 million, or approximately $87 million annualized. * Actual revenue per million tokens varies by more than 1,000x across crypto AI projects. Multiple cases, including Surplus Intelligence, show substantial overconsumption of tokens relative to actual revenue. * Payments made using the x402 standard account for a tiny share of large-model calls. Just 0.05% of Venice's $7.2 million in credit sales was paid via x402, and the share of crypto payments continues to decline. This study analyzed 130 Web3 projects that directly interact with model tokens, using on-chain verifiable data from September 13 to October 12. Excluding OpenRouter itself, verifiable or estimated model-related payments over 30 days totaled about $8.7 million. Venice alone accounted for about $7.2 million (83%), while the other 100-plus projects combined accounted for about $1.5 million. Research methodology and scope: only on-chain verifiable data. The $8.7 million total combines three components: receipts, burns, buybacks, credit activations, and x402 settlements directly verifiable on-chain; estimates based on OpenRouter providers' daily token volumes multiplied by effective unit prices; and settlement files published by projects. The statistical window covers September 13 to October 12, 2026, with data retrieved from October 13 to 15. The total excludes OpenRouter's own revenue, credit consumption such as DIEM, and projects with only self-reported figures. Resold gateway activity and overlapping upstream activity are counted only once. Off-chain card revenue is not visible, and OpenRouter provider figures represent a lower bound on the corresponding projects' usage. Venice generated about $7.2 million in 30 days; ranking of 100-plus other projects and comparison with OpenRouter's scale. Major projects' revenue outside Venice: Phala (TEE inference via OpenRouter), about $520k; Bittensor subnet SayGM, about $426k; Surplus Intelligence, about $128k; engy, about $112k; and Prime Intellect, about $86k. Most of the rest generated less than $10k per month. Pay-per-call gateways generated only about $3,800 combined over 30 days, while agent platforms and AI applications generated less than $1,000 combined. Venice's high revenue comes from its policy, in effect since July 17, of using $5 to buy back and burn VVV on the open market for every $100 of API credits sold. Based on 30-day data from the third-party venicestats dashboard, credit sales were about $7.2 million, or about $87 million annualized (on the same scale as Sacra's estimate of $110 million in annualized revenue). Venice's closed-source models cost 20% to 25% more than the official ones, but users still pay, primarily because of demand for privacy and freedom from censorship. The gap between token consumption and actual revenue: four typical cases of inflated usage. Actual revenue per million tokens varies by more than 1,000-fold across projects: about $2.3 for SayGM, less than $0.003 for Gonka and Twigpine, about $0.39 across OpenRouter, and about $0.84 for Venice. The following four cases illustrate typical gaps between usage and revenue: Surplus Intelligence: Processed 4.17 trillion tokens over 30 days, but paid sellers only $128k on-chain; 96% of buyer-side deposits came from fiat-to-USDC conversions via Bridge (owned by Stripe). AntSeed: Settled a total of $312k, with about 68% coming from wash-trading sellers excluded from its own rewards rules; weekly settlements fell from about $22k in July to $4,000-$5,000 in October. Cluster Protocol: Of its $54.8k in x402 revenue over 30 days, model conversations accounted for only $152; 722 wallets made calls daily using fixed combinations and all stopped at the same time on October 2, while token disclosures listed wallets that had used x402 as eligible for the airdrop. DGrid AI: The often-cited "$23 million in H1 revenue" was sales of membership passes; on-chain gateway deposits over 30 days were about $16k. x402 accounts for a tiny share of model calls: only a small portion of Venice payments use x402. x402 has become the de facto standard for agent payments, with about 40 organizations - including Stripe, Visa, and Google - as foundation members, but large-model calls account for only a tiny share. Of the 32,498 x402 resources listed on Coinbase Bazaar, only 3% were related to large models. Of about 940k calls over the past 30 days, large models accounted for 9%, of which 88% came from BlockRun alone. On x402scan, token-billed inference gateways generated only about $3,800 combined over 30 days. Of Venice's $7.2 million in credit sales, only about $3,100 was paid through x402 (less than 0.05%). OpenRouter's share of crypto payments fell from about 5%-6% in mid-2025 to about 1.3%-2.2% in Q1 2026. On April 2, 2026, OpenRouter disabled programmatic crypto deposits after Coinbase Commerce shut down its Charge API. Credit token utilization is low: arbitrage contracts account for 80% of trading in the Orbio CREDIT market. Venice DIEM is the most substantial credit-token usage case: its 30-day credit cap was about $834k, with $371k actually used (a utilization rate of about 44%), backed by Venice's paid business generating more than $7 million in monthly revenue. Orbio launched on Robinhood Chain on August 31. Its daily fees fell from about $37k on October 1 to $7,652 on October 12. Only about 34.3k CREDIT had been activated on-chain over the past four weeks, with about 15k coming from users who paid by card on the official website (activated on their behalf by Orbio). The CREDIT secondary market is dominated by arbitrage: the top five buyers by 30-day order-book trading volume were all arbitrage contracts, accounting for 80% of buyer trading volume. None of the CREDIT acquired by these contracts was activated; all of it was transferred to Uniswap v4 to be cashed out. Other credit tokens fared worse: SurplusLLM issued $10.6k in credits, of which $1,336 was activated; holdapi issued $697, of which $16 was used. The six credit tokens with trading pools on Robinhood Chain other than Orbio had a combined market cap of only 3% of Orbio's and included counterfeit contracts such as fake Zevio and fake c0mpute ZERO. Frequently asked questions. How was this study conducted, and how was the $8.7 million figure calculated? The $8.7 million total combines three components: receipts, burns, buybacks, credit activations, and x402 settlements directly verifiable on-chain; estimates based on OpenRouter providers' daily token volumes multiplied by effective unit prices; and settlement files published by projects. The statistical window covers September 13 to October 12, 2026. The total excludes OpenRouter's own revenue and projects with only self-reported figures, and does not use token counts reported by the projects themselves. Why did Venice capture 83% of market revenue, while other projects remained relatively small? Since July 17, Venice has bought back and burned $5 in VVV for every $100 in credit sales. Data from the third-party venicestats dashboard puts 30-day credit sales at about $7.2 million. Venice's closed-source models cost 20% to 25% more than the official ones, but users still pay for privacy and freedom from censorship. Most other projects rely on OpenRouter orders or token subsidies and lack a differentiated competitive advantage. Low prices for decentralized inference are maintained primarily through token emissions, and usage declines as subsidies decrease. Why is there such a large gap between token consumption and actual revenue? The gap has several causes: deposits for some projects mainly come from fiat-to-USDC conversions (96% of Surplus Intelligence's came from fiat); some tokens come from the projects' own tasks rather than external demand; some usage consists of machine calls using fixed combinations (such as Cluster Protocol's 722 wallets); and some usage is maintained by token subsidies or free credits (82% of Twigpine's usage came from models with free credits). The study recommends using actual revenue per million tokens as a more meaningful metric than total token volume. Disclaimer: The information on this page may come from third-party sources and is for reference only. It does not represent the views or opinions of Gate and does not constitute any financial, investment, or legal advice. Virtual asset trading involves high risk. Please do not rely solely on the information on this page when making decisions. For details, see the Disclaimer.

Antseed launches a decentralized marketplace for AI inference that dramatically lowers developer's costs. Developers now have a much cheaper way to access cutting-edge artificial intelligence inference capabilities with the launch of Antseed today. It's a new peer-to-peer marketplace that makes it possible to access almost any frontier model at a small fraction of its normal cost by bypassing centralized gateways in favor of independent and mainstream providers. The launch of Antseed came as its controlling entity, the Antseed Foundation, closed on a $2.4 million token funding round led by Spark Capital, an early investor in Anthropic PBC, with participation from Collider, DCG, North Island Ventures, Reciprocal Ventures, Venice.ai and Relay Capital. Antseed's marketplace was created to help developers combat what the startup says is the "skyrocketing cost" of AI inference, or the computing power required to run AI models and generate responses. It cites data from Gartner Inc., which projects that global spending on AI cloud infrastructure is likely to grow to $42 billion by the end of this year. The problem with inference is that developers have had few options until now but to access models through centralized middleman gateways and application programming interfaces that are controlled by just a handful of model providers. These companies have the ability to dictate the costs and usage limits, and they can adapt their pricing on the fly, leading to unpredictable price spikes for inference workloads. Antseed's open inference marketplace is meant to get around this challenge by giving anyone the ability to route AI requests across hundreds of independent and mainstream model providers, bypassing those centralized gateways. In effect, this forces model providers to compete for every request by lowering their prices. Antseed claims this is what allows users to access leading frontier models by up to 97% less than their official API prices. The marketplace relies on a self-hosted local router, through which providers can list their available inference capacity, utilizing hardware they own, and set their own prices. Antseed verifies that each provider does indeed provide access to the models it claims to serve. The marketplace is built on the idea of reputation: If a provider fails a verification check, its reputation will decrease, and its routing priority will decline. Antseed co-founder Amos Meiri compares its marketplace to the file-sharing service BitTorrent. He says it enables developers to maintain their existing AI workloads and configure routing for each request based on their cost, latency, capability and privacy requirements. "What BitTorrent did for files, Antseed is doing for AI inference, with no central server and no company deciding who gets access or what it costs," he said. "As agents make more model calls and inference becomes a material operating cost, developers need a simple way to route each request based on price, performance, capability and privacy. Antseed gives them one compatible local endpoint and access to a competitive market of independent providers, without locking their applications into a single centralized gateway." Antseed's decentralized approach to model access enables model pricing to be dictated by open market dynamics, and it's entirely anonymous too, so developers aren't required to sign up to the service or provide any details at all. Because the router runs locally on the provider's own systems, developers can also choose to go with hardware-verified providers that ensure strict prompt confidentiality. Developers can also access models through a pay-per-request pricing model, so there's no need to commit to rigid, subscription-based pricing tiers. Meiri said this strategy is already getting traction as a growing community of developers begin to realize the cost benefits Antseed providers. He points to on-chain data that shows how Antseed has processed almost 150 billion tokens across 202 active providers so far. Spark Capital co-founder and General Partner Santo Politi said inference is to AI what fuel is to airlines. "It's a core operating cost that companies need to manage as they scale, but right now only a handful of companies control the wells," he explained. "Antseed is the counterweight to that centralized control, serving as a decentralized open market that gives developers the freedom to choose how they source compute while creating a direct market for providers to compete." Images: Antseed. A message from John Furrier, co-founder of SiliconANGLE: Support its mission to keep content open and free by engaging with theCUBE community. Join theCUBE's Alumni Trust Network, where technology leaders connect, share intelligence and create opportunities. * 15M+ viewers of theCUBE videos, powering conversations across AI, cloud, cybersecurity and more * 11.4k+ theCUBE alumni - Connect with more than 11,400 tech and business leaders shaping the future through a unique trusted-based network Are you an AWS customer? Support SiliconANGLE financially by buying your AWS services from its Marketplace portal page and links: https://siliconangle.com/aws-marketplace/. About SiliconANGLE Media. SiliconANGLE Media is a recognized leader in digital media innovation, uniting breakthrough technology, strategic insights and real-time audience engagement. As the parent company of SiliconANGLE, theCUBE Network, theCUBE Research, CUBE365, theCUBE AI and theCUBE SuperStudios - with flagship locations in Silicon Valley and the New York Stock Exchange - SiliconANGLE Media operates at the intersection of media, technology and AI. Founded by tech visionaries John Furrier and Dave Vellante, SiliconANGLE Media has built a dynamic ecosystem of industry-leading digital media brands that reach 15+ million elite tech professionals. Its new proprietary theCUBE AI Video Cloud is breaking ground in audience interaction, leveraging theCUBEai.com neural network to help technology companies make data-driven decisions and stay at the forefront of industry conversations.

Venice AI chooses NEAR for encrypted AI inference. Partnership expands NEAR's role in privacy-focused artificial intelligence. Reading Time: 3 mins read Venice AI has selected NEAR Protocol as infrastructure for its privacy-focused artificial intelligence services, marking a new development in the integration of blockchain technology with confidential AI workloads. The announcement places NEAR at the center of an implementation involving verifiably encrypted inference, rather than a municipal blockchain initiative by the Italian city of Venice. The integration is designed to allow Venice AI users to run AI workloads with privacy protections that can be verified through technical mechanisms, including Trusted Execution Environments and end-to-end encryption. The development gives NEAR a direct use case in private AI infrastructure while extending the relationship between blockchain-based systems and AI applications. NEAR has increasingly positioned itself as infrastructure for AI-related applications, with its network supporting high-throughput blockchain operations and services focused on confidential computation. The protocol currently describes its infrastructure as supporting 600-millisecond block times, 1.2-second finality, and scalability of up to 1 million transactions per second. Confidential AI becomes a central focus. Venice AI has been developing privacy-oriented AI services in which sensitive prompts and generated content can receive additional protection during processing. Its integration with NEAR AI infrastructure provides access to private inference capabilities using hardware-enforced security environments. NEAR's website identifies Venice AI as an existing user of its private inference infrastructure, with prompts processed inside hardware-enforced enclaves. This approach is intended to reduce exposure of sensitive information while allowing AI applications to use advanced computational models. The latest development builds on an earlier September announcement involving Venice and NEAR. The integration has subsequently expanded into verifiably encrypted AI inference, with both Trusted Execution Environment and end-to-end encrypted models included in the implementation. The technical model could be significant for developers and businesses that need AI capabilities while maintaining stronger privacy controls. Rather than relying solely on policy commitments concerning data handling, the architecture uses technical safeguards designed to provide measurable protection during computation. NEAR expands its AI infrastructure strategy. The development also fits into NEAR's broader strategy of positioning the protocol as infrastructure for AI agents and applications. Its chain-abstraction architecture is designed to allow applications and AI systems to interact with assets and services across multiple blockchain networks. NEAR also operates Chain Signatures and NEAR Intents for cross-chain execution and transactions. NEAR has increasingly combined these blockchain capabilities with privacy and AI infrastructure. Its current platform emphasizes private inference, secure agent environments, and confidential transactions as components of its broader technology stack. For developers, the Venice integration provides an example of how private AI workloads can be connected with blockchain infrastructure without requiring users to expose sensitive information in conventional processing environments. Adoption will determine broader network impact. The partnership may also increase attention on NEAR's network activity as confidential AI applications expand. However, the available announcements do not establish a specific transaction-volume target or quantify how much additional network activity Venice's implementation will generate. That distinction is important because the technical significance of the integration does not automatically translate into measurable changes in token demand or network usage. The longer-term impact will depend on the scale of Venice's AI workloads and the extent to which other developers adopt similar infrastructure. NEAR has continued expanding its AI and privacy capabilities alongside broader blockchain developments. Its platform now highlights confidential inference, cross-chain transactions, and AI agents as interconnected parts of its infrastructure strategy. For Venice AI users and developers, the implementation provides a framework for running AI inference with verifiable encryption and hardware-based privacy protections, while giving NEAR a concrete application for its confidential AI infrastructure. The development therefore represents a shift from blockchain being used primarily as a transaction or settlement layer toward its potential role as supporting infrastructure for private and verifiable artificial intelligence services.