Skip to content

The CCS Interview

Sahara AI's Sean Ren on tokenizing models and data so AI contributors share revenue

Sean Ren, co-founder and CEO of Sahara AI, explains how the company uses a purpose-built blockchain to track who contributes to AI and pay them for it. Recorded in April 2025, the conversation covers the platform's enterprise traction, how it differs from Web2 AI and what he said was still to come, including the AI marketplace and token liquidity.

Interview by Ashton Addison Recorded April 2025 5 min read

Key takeaways

  • Sahara AI uses blockchain to create a provenance system that automatically records and traces contributions from data providers, model developers, and compute providers in AI development.
  • The platform enables revenue sharing through smart contracts that autonomously distribute earnings from AI models to all upstream contributors based on their ownership shares.
  • Sahara AI operates two layers: a purpose-built Layer 1 blockchain with AI-native functions for infrastructure, and application layer with a data service platform and developer platform.
  • The Sahara data service platform has 500,000 gig workers and 200,000 daily active users contributing various data types like audio transcription, video descriptions, and screen recordings.
  • Current centralized AI systems do not fairly compensate contributors whose data and feedback improve models, creating an unsustainable ecosystem that Sahara AI aims to address.
Since this interviewSince this interview, the SAHARA token listed on Binance, OKX and other exchanges in June 2025, and Sahara AI published a 2026 roadmap in March 2026 that targets a mainnet for its agent-focused chain in Q4 2026 or later. It also launched an Ethereum to BNB Chain bridge secured by Chainlink CCIP in June 2026.

From AI research to an on-chain copyright system

Ashton Addison: How did you go from AI research to building Sahara AI?▶ 00:27

Sean Ren: In late 2022, while I was at USC, I had many discussions with my co-founder Tyler, a former investor who taught me a lot about blockchain and crypto. I started to realize that AI is not only about productivity, or making models better than humans. It's also about building a more sustainable and healthier ecosystem. Everyone who interacts with AI is part of it, whether they contribute data, give feedback, provide model recipes or supply GPU compute. They should all get fair compensation for improving AI, and that is not what today's Web2 AI landscape looks like. So we started thinking about how to use blockchain to build a platform with an open-ledger copyright system that automatically records and traces everyone's contribution as AI is created. That gives us a provenance structure for new business models. For example, I can split a model into shares owned by the model developers, the compute providers and the data contributors. That is recorded on-chain, immutable, transparent and auditable by anyone, and smart contracts can attribute every dollar the model earns to those upstream contributors.

Ashton: If I use Sahara AI the way I use ChatGPT, what looks different?▶ 06:31

Sean: For users of AI, the first principle is that it's frictionless. It's like using any other AI app, and you won't notice much difference. For developers and builders who own the AI, it works differently. They have an account where they see revenue come in for every asset they co-own. Maybe they have a 10% split on a model they contributed to and 20% on a dataset they built. It's a continuous revenue stream as those assets are deployed and used by others. The ecosystem has two layers. The bottom is the blockchain infrastructure layer. The Sahara blockchain serves as the provenance infrastructure, like a copyright system for the whole platform. It's a purpose-built, EVM-compatible Layer 1 for AI, with AI-native precompile functions and access management protocols. On top is the application layer, which is our moat compared with many other crypto AI infrastructure projects. We build enterprise-grade applications natively.

In Web2 AI, all of the revenue and monetization opportunities are controlled by the app builders, the tech giants.

What is live, and who is paying

Ashton: Where does this stand today? Can people use it now?▶ 11:40

Sean: Our data service platform launched about three months ago. We started going to market with enterprise clients about a year ago, working with big names like Microsoft, Amazon and Snapchat, as well as schools and research labs. We generated over $5 million in revenue within six months. Now we're gradually opening the platform to retail users, starting with a whitelist of 200,000 who have a very high active ratio. In the next couple of months we'll probably make it generally accessible. On the developer platform, we started beta testing about a month ago with dozens of testers who are giving very good feedback. Everything we open to the public will already be on-chain. The last step of uploading a dataset or model is to register it on-chain, and you then claim an ownership NFT. It can be a co-ownership token if you jointly own the asset with others. Down the road this will integrate with licensing tokens that show how usage comes in and how revenue flows, and the ownership tokens split that revenue back to all the contributors.

Ashton: How did you land partnerships with the likes of Microsoft and Amazon, and where does the $5 million come from?▶ 14:47

Sean: The $5 million is from the data service platform alone. We've used these tools on some very challenging AI data projects for enterprise clients, across different modalities and use cases. It has worked for two reasons. First, our prices are very competitive, because our huge network of online gig workers, who come from different regions, is willing to take attractive pay. Second, that network covers a wide range of expertise that rarely exists in Web2 data service companies like Scale AI. We've competed with Scale on multiple enterprise projects and delivered both high quality and fast results. The developer platform has also won enterprise clients. A publicly listed car manufacturer in India is using it to build agents for design, supply chain optimization, dashboard monitoring and legal document review, while we record the provenance of each one. Our real moat is adding provenance and liquidity to the asset token. Today there's no way for Web2 developers to monetize their models and datasets, because Hugging Face and GitHub are open-source platforms. We're building a proprietary, privacy-protecting platform and bringing in liquidity from Web3 retail, so you get the best of both worlds.

Why decentralized AI changes the economics

Ashton: Does decentralized AI have an economic advantage over centralized AI companies?▶ 17:41

Sean: To understand it, first look at what happens in Web2. In Web2 AI, all of the revenue and monetization opportunities are controlled by the app builders, the tech giants like Meta, Google and OpenAI. They touch the application users, who give them revenue directly. With that revenue, they have a dominant say in how much they pay for GPU compute and how much they pay to hire AI talent. They control the entire landscape, and the rest of the ecosystem has little say because it doesn't have the revenue. The paradigm Sahara AI is creating is an open and transparent monetization architecture. If you own or co-own something, the revenue flows to you autonomously through the smart contract. There's no centralized accounting or bookkeeping that can stop you earning money. The second big difference is liquidity. Today, if you're a great AI developer, the best way to monetize yourself is to get hired by OpenAI with a $1 million package and equity. What if you could tokenize your models and datasets, get people to invest in those tokens and bring liquidity to them? Those are two huge opportunities that don't exist in Web2.

You know how many API calls are going to these models, so you can see which one is actually making more revenue.

How retail users earn, and what comes next

Ashton: How can non-developers earn on the platform?▶ 20:39

Sean: We're seeing two types of retail user. The first is online gig workers. They spend time online, create good datasets and label data. We have 500,000 of these users actively working on our data service platform. More projects will be available once we open the platform for anyone to create data projects. The second is the DeFi retail user, who has been investing in, say, memecoins or other opportunities. For them there's a brand-new space where datasets and models are being tokenized. The big difference from memecoins is that you can see the fundamentals of these assets. You know how many API calls are going to these models, so you can see which one is actually making more revenue than the others. You can base your speculation on that, and that's very different from the memecoin market.

Ashton: When can people expect to see the liquidity and marketplace features?▶ 22:37

Sean: You can already apply for whitelist access to our data service platform and developer platform, and we're opening up the whitelist every two to three weeks. The links are on our website, and everything will be announced on our Twitter. Down the road we'll launch our AI marketplace. Everyone can come to upload useful datasets, get them verified by the community and earn incentives for publishing good ones. You can also use those incentives, the Sahara points, to purchase good datasets for your own use cases and models. So stay tuned for our series of launches around the AI marketplace and the developer platform.

This interview has been edited for length and clarity. Recorded April 2025.

Sean Ren
Sean RenSean Ren, who publishes as Xiang Ren, is the co-founder and CEO of Sahara AI, a blockchain platform for AI data, models and agents, and an Associate Professor of Computer Science at the University of Southern California. He earned a bachelor's degree in computer science… Full profile →
Share this interviewXLinkedIn
▶ Watching the interview