AI is hungry for data. High-signal, real-world, multimodal data. And almost no one in the world can deliver it at the scale the largest AI labs now require.
Except one project: Grass, the DePIN network built by Wynd Labs, already running on 8M+ residential devices and pushing petabytes of real-world web data into AI pipelines. In this episode of the DePINed Podcast, Tom Trowbridge (co-founder & CEO of Fluence) sits down with Andrej Radonjic, co-founder & CEO of Wynd Labs, to unpack how Grass became one of the most important data engines in the AI ecosystem.
Below is a curated breakdown of the most important insights from their conversation.
The Data Bottleneck Behind AI’s Explosive Growth
Everyone sees the GPU shortage. Almost no one sees the data shortage. Modern AI models are running into a wall:
- They need more data than ever.
- They need new data types (video, multimodal, dynamic events).
- And they need it fast, often in real time.
What Grass discovered is simple but foundational:
Massive AI labs are limited not by modeling capacity, but by the laws of physics. Grass built infrastructure to solve that. And accidentally discovered a massive market gap: no one else was meeting these data needs at scale.
The Grass Network: 8M Devices, Ethical by Design
Grass operates one of the largest residential proxy networks in the world — but unlike the “dark proxy” industry, it’s fully transparent, opt-in, and tokenized.
Users contribute bandwidth from home devices and get rewarded. AI companies receive clean, ethically sourced, globally distributed web data at scale.
This is the opposite of centralized, opaque data vendors. And it’s exactly the model DePIN was designed for.
Forced Verticalization: Storage, Hardware, and Petabyte Pipelines
AI labs want high-signal data. Traditionally, they would:
- Buy a giant corpus
- Spend months filtering it with their own GPUs
- Throw away 90–95% of the low-signal material
But the industry hit a wall:
- Labs need data now, not in two months
- They don’t want to burn H100s on preprocessing
- Data transmission is physically capped at ~8 PB/day even at terabit speeds
So Grass had to verticalize up and down the stack:
Down the stack:
- Building their own distributed storage network
- Running their own data centers
- Optimizing infrastructure for petabyte-scale movement
Up the stack:
- Running classifiers and metadata extraction
- Using ML to identify high-signal subsets
- Delivering curated datasets instead of raw dumps
This turned Grass into not just a data network, but a data refinery.
The result? AI labs now often buy: 5% of the data, the same price
Revenue Arrived Early And It’s Bigger Than Expected
This was one of the biggest surprises in the episode.
Grass originally expected the multimodal data business to take years to monetize.
Instead, revenue hit millions per month early — and is growing quarter over quarter.
And there was a major hint: a token holder call
Tom calls this out as an inflection moment:
“When a founder says things are going better than expected — that’s generally a good thing.”
The Token Holder–First Philosophy
One of the strongest segments in the conversation was the discussion of token economics. Grass implemented unusual protections:
- Locked tokens cannot be staked, ensuring early investors don’t dilute community rewards
- Revenue sits at the Grass Foundation, not in a private company
- Wynd Labs operates as a contractor that can be replaced
This creates alignment rarely seen in crypto.
Live Context Retrieval: The Next Frontier
The most ambitious idea in the episode is Grass’s 3–5-year vision:
Live Context Retrieval (LCR) — feeding AI models real-time data from the entire public web.
Today’s models are trained offline and repeatedly become stale.
Tomorrow’s models will:
- Pull data from yesterday’s viral video
- Scrape product listings live
- Integrate events happening right now
- Access dynamic web content during inference
Grass wants to power 90%+ of all LLM calls in the world. That is the scale of their ambition.
How to Join Grass
Grass aims to make participation extremely simple:
- Visit grass.io
- Install the extension or app
- Contribute bandwidth
- Earn rewards in a transparent, community-first system
Their Discord hosts 500,000+ people, making it one of the largest communities in the DePIN industry.
About DePINed Podcast
DePINed is a podcast exploring the frontier of decentralized physical infrastructure, hosted by Tom Trowbridge, co-founder of Fluence. Each episode features in-depth conversations with founders, builders, and investors who are shaping the future of real-world Web3 networks.
️ Want to be a guest? Fill out the form here
Catch 50+ episodes featuring top DePIN founders and trailblazing protocols on:
