GM. This is Milk Road, the newsletter that takes you deep into crypto rabbit holes so you can make informed decisions on the other side.
Hereβs what weβve got for you today:
- βοΈ AI labs are paying this crypto network for web data.
- ποΈ The Milk Road Show: $71K or $205K? The Bitcoin Setup That Could Decide This Cycle.
- πͺ DRV launches zkVM-based v3 settling directly on Ethereum mainnet.
Nexo is a premier digital asset platform that helps clients build, manage, and preserve their wealth. Get started with Nexo today.
Prices as of 2:00 p.m. ET. Powered by CoinGecko.
AI LABS ARE PAYING THIS CRYPTO NETWORK FOR WEB DATA π±
GRASS is up ~300% since its February low.
That's a big move for a token most people have never heard of⦠so naturally, we went digging.
Turns out AI labs have paid Grass ~$32M since the start of 2025, and Multicoin Capital (one of crypto's best-known investment firms) bought in last week.
The team also has a much bigger plan: go from selling old internet data to AI companies to becoming the way AI agents browse the live web.
My take: Grass has one of the few businesses in crypto where outside customers pay real dollars for the product.
Whether owning the token gets you a slice of those dollars is a separate question, and it's the one this whole play is built around.
But first, what the heck is Grass?
It's an app that uses your spare home internet in the background to visit public websites, then packages up what it collects (text, images and a whole lot of video) and sells it to AI labs to train their models.
Home internet is an important piece of the puzzle here, because websites routinely block traffic from big data centers, while letting normal households through.
That puts Grass in the DePIN sector (Decentralized Physical Infrastructure Networks), where a token pays regular people to build a network instead of one company paying for it all.
DePIN has a bad reputation. Back in January, Meltem Demirors (Crucible Capital) posted that DePIN is dead, arguing these networks pump out so many new tokens (often 10%+ of supply a year) that they can't compete with the cheap money funding regular AI infrastructure.
For most of the sector she's right, but Grass is one of the rare cases where somebody is paying, and the market has noticed. π
(It's also still ~83% below its November 2024 high of ~$3.90, so plenty of early holders are underwater.)
Grass's revenue by half-year:
- H1 2025: $2.7M.
- H2 2025: $14.3M.
- H1 2026: $14.5M verified ($17-18M by Grass's own count).
That's ~5x a year earlier, and Grass says nearly every AI lab that's bought from it has come back for more.
Since July, Grass has also paid contributors in USDC out of that revenue, instead of handing out newly created GRASS.
That cuts out a lot of selling, which is exactly the DePIN problem Demirors was talking about.
Multicoin (an early backer of Solana and Helium, one of the first DePIN networks) bought through both its token fund and its venture fund, a sign it plans to hold for a while.
Its pitch goes something like thisβ¦
Training data is a one-off sale, and a model's knowledge freezes the day its training ends.
But when you ask an AI agent about something happening today, it has to check the live web before it answers.
Every one of those lookups is a fresh, paid request, and Multicoin is betting Grass becomes the network agents use to make them.
The business is growing. But if you own GRASS, you don't own the businessβ¦
ONE ACCOUNT FOR ALL YOUR CRYPTO NEEDS
Crypto is still a weirdly fragmented experience.
One app to buy. Another to earn yield. A third to borrow against your stack.
But Nexo is now bringing it all under one roof: Trade, Earn, and Borrow.
Nexo is a premier digital asset platform that helps clients build, manage, and preserve their wealth.
Here's what makes them worth trusting:
- Official Crypto Partner of Tennis Australia
- First Digital Asset Partner of the Audi Revolut Formula 1 Team
- Operating since 2018
- $7B+ in AUM
- SOC 2 & SOC 3 certified
- 24/7 client care
AI LABS ARE PAYING THIS CRYPTO NETWORK (P2) π±
Before we get to the token, here's why Multicoin thinks live web search is the bigger prize π
(Btw - "tokens" here means the chunks of text an AI model processes, not crypto.)
On OpenRouter, a marketplace that sends requests to different AI models, agents passed humans in February and their usage has grown ~14x since.
Agents check the web constantly, and Grass is building three products to serve them:
- Contents API: lets an AI open any public web page through Grass's network of home connections.
- Search API: an index of the web that tells the AI where to look in the first place.
- Multimodal: video, images and PDFs, so agents can pull more than text.
(An API is a plug that lets one piece of software use another and pay per request.)
That's why Grass stands out to me from the pile of AI tokens - most are selling a roadmap, while Grass is selling data to labs today, and using the cash to build the next product.
At ~$0.66, GRASS has a fully diluted value (what it'd be worth if every token that will ever exist were trading today) of ~$660M.
That's around 9-10x Grass's 2026 revenue guidance, which is a reasonable price if two things hold up: the guidance, and token holders eventually getting paid.
Does any of this money reach GRASS holders?
Everything sits under the Grass Foundation, which has no shareholders, and Wynd Labs (the original builders) handed over all its intellectual property and now works as a paid contractor.
That avoids the usual crypto setup where a company's shareholders keep the revenue and the token gets leftovers.
But the token itself still has no claim on the cash.
GRASS is used for staking, voting and rewards, and its only buybacks to date (~$350K in late 2025) add up to about 1% of the revenue Grass has recorded.
Multicoin argues that reinvesting profit into a web-scale search index will do more for the token long term than buybacks would.
I think they're probably right about the business, but it means buying GRASS today is a bet that the Foundation will eventually share the money, and right now nothing forces it to.
The revenue story also has a few holes to poke at:
1. We can't see the customers. Grass hasn't named a single one, and since none of the money moves onchain, crypto Twitter has spent months asking how much of it to trust. Regen Financial (the firm Grass hired to check its numbers) traced the deals to cash, but says up front its work stops short of a full audit.
2. H2 needs to be huge. Guidance of $65-75M for 2026 means roughly $50-60M in the second half, ~3.5-4x the first. That counts on ~$15M of deals that slipped out of H1 plus ~$35M that aren't signed yet, so pretty much the whole pipeline has to land.
3. One type of customer. Almost all revenue comes from AI training data, so if labs cut back, Grass feels it immediately.
4. More supply is coming. The last chunk of early investor tokens (investors hold ~25% of all GRASS) unlocks around the end of October, right after a ~300% run.
Where that leaves me:
Grass is the strongest evidence I've seen that DePIN can work, and it's firmly on my watchlist.
But a 300% run has already priced in plenty of good news, and the next two quarters will show whether it was earned.
What I'm watching:
- Q3 revenue, and whether that ~$15M of delayed deals actually closed.
- The Contents and Search products shipping and earning money of their own.
- Any concrete plan to send revenue to GRASS holders (buybacks, burns or direct payouts).
If all three show up, today's price could look cheap in hindsight.
If Q3 comes in light, I'd expect a good chunk of that 300% to give itself back.
CALLING ALL CRYPTO NERDS π₯
If you wake up and check Bitcoin before the weather, spend too much time on Crypto Twitter, and can actually write, we might have a job for you.
Milk Road is looking for a crypto-focused writer/content creator to join the team.
BITE-SIZED COOKIES FOR THE ROAD πͺ
Bitcoin miners made money by owning the machines. What if you could do the same with AI? B3 Labs lets investors own Nvidia GPUs and get paid when AI labs use the compute.*
DRV launches zkVM-based v3 settling directly on Ethereum mainnet.
John Gillen says you donβt need leverage to win this cycle.
ETHFI is launching an Ethena-powered stablecoin.
*this is sponsored content.
