Grass moved to explain its recent market action on Sept. 28, when Wynd Labs CEO Andrej released a long statement titled Abundant Intelligence. The post did not reveal any new customer list, and it did not lay out a material buyback program. What it did provide was a coherent framework for the latest price move in GRASS.
Over the past two weeks, GRASS climbed from around $0.35 to above $0.70, nearly doubling. In practice, the market had already priced in the shift before the statement arrived. The manifesto read more like a roadmap released after traders had begun treating Grass as something larger than a DePIN project that sells AI training datasets.
From packaged datasets to an internet access network for AI agents
For most of the past year, Grass had a fairly clear business profile in the market. It used residential IPs to collect public web data, assembled snapshots, and sold those packaged results to AI labs. The business could generate real cash flow, but many investors saw a ceiling. If high-quality training data is increasingly treated as a finite resource, then a company built around collecting and packaging that data can be viewed as transitional rather than structurally premium.
Andrej's new message was aimed squarely at that ceiling. The argument starts with a weakness in large models: once training ends and model weights are frozen, the system's knowledge is stuck at that point in time. If a user asks about breaking news or current prices, the model has to reach the live internet during inference and pull fresh public pages.
That is where the access problem begins. Many major websites run anti-bot defenses from providers such as Cloudflare and Datadome. Requests from data-center infrastructure are often blocked because they carry obvious server-side signatures. Grass is trying to turn that obstacle into its next business story.
The network it built originally runs on residential IPs shared by millions of ordinary users. In the old model, Grass mainly used that network in the background to collect and clean web pages, then sold the output as offline training material. In the new model, Andrej wants the same network to function as a live proxy layer that lets AI systems reach the web in real time.
The roadmap: search, retrieval, and multimodal access
The company described a tool stack meant to connect AI systems to the live internet more directly.
- Search API: this is meant to solve discovery. Because Grass has spent a long time collecting public web data, it says it has built up a large archive of public web copies. The API is positioned as a search layer for AI models, helping them locate the most relevant pages before retrieval starts.
- Contents API: this is meant to solve access. Once a target page is identified, the model would not hit the site from a data-center server. Instead, the request would be routed through Grass's residential nodes. To the target website's firewall, that request could resemble ordinary traffic from a home connection rather than a crawler coming from hosted infrastructure.
- Multimodal retrieval: this is meant to solve interpretation. The live internet is not just text. It includes images, PDF reports, and embedded video. Grass says this layer is designed to help models understand those forms of content as well.
Viewed through that lens, the comparison some overseas crypto analysts are making becomes easier to understand. If Cloudflare's business logic is to charge websites to keep automated crawlers out, Grass is trying to charge AI companies to get them in by routing traffic through residential networks.
That repositioning changes the valuation anchor. Grass no longer wants to trade as an off-chain data contractor valued on a conventional DePIN framework. It wants to be seen more like an AI real-time retrieval layer and agent gateway. The article argues that this is why a roadmap alone, even without any new signed customer announcement, was enough to fuel speculative interest in the secondary market.
The business already has revenue
The piece makes another point clearly: Grass is not being judged only as a concept story. According to figures disclosed by the company during a July tokenholder conference call, Grass generated $17 million in revenue for full-year 2025. In the first half of 2026, revenue reached $17 million again, up nearly 7x from the same period a year earlier.
Management expects revenue from the training-data business alone to reach $65 million to $75 million for full-year 2026. The company also said the business shows seasonality. One large order, worth roughly $15 million and originally expected in the first half, shifted into the third quarter because AI labs changed the timing of their training cycles.
The article says the company is already profitable on paper. After building its own compute and storage infrastructure in 2025, monthly operating cash spending fell to roughly $2 million to $3 million, mainly for infrastructure, which is below the pace of incoming revenue.
On structure, Grass is presented as trying to avoid a familiar Web3 problem in which the corporate entity captures most of the value while the token is left with little direct claim. Customer contracts and business revenue flow into Grass DataCo and the foundation, while Wynd Labs acts as an engineering and business services provider. In theory, that means protocol revenue belongs to the network rather than being siphoned away to a separate company.
The unresolved question is still token value capture
Strong business numbers do not automatically settle the issue for token holders. The article argues that Grass still faces its hardest test here: how profits return to GRASS.
Buybacks remain minimal
For public-market crypto traders, direct buybacks and burns are the clearest form of token support. Grass has shown very little urgency on that front. The piece notes that the project carried out only a small open-market buyback of about $350,000 in November 2025. Since then, large-scale repurchases have largely been absent.
The company's logic is to reinvest most profits into compute, storage, and network expansion rather than using cash to support the token in the secondary market. That may strengthen the operating moat over time, but it also removes the most visible near-term source of token demand.
Node incentives are under pressure
Grass also changed tokenomics in Stage 2. It stopped subsidizing nodes with newly issued GRASS and switched to paying rewards in USDC drawn from real protocol revenue. From an inflation standpoint, that reduces pressure from continuous token emissions and is an obvious positive. But it creates a practical tension on the supply side of the network.
Some ordinary bandwidth providers reportedly found that after months of participation they were earning only a few cents to a few dozen cents in stablecoins, leading some to shut off their nodes. That matters because the entire new story depends on a large network of real residential endpoints. If rewards are too low and the node base shrinks, the physical foundation of Grass's claimed access advantage weakens with it.
Stakers still have limited income visibility
The article adds that a July governance vote approved distributing part of USDC revenue to GRASS stakers. Even so, the actual payout scale and split ratios remain insufficiently transparent in public detail.
When token holders cannot clearly see those tens of millions of dollars in protocol revenue showing up as a straightforward APY or regular income stream, the token is left trading between business expectations and unlock anxiety. One specific overhang mentioned in the piece is a tail-end investor unlock scheduled for late October.
Narrative has changed, but execution still decides the outcome
Grass has now shifted its market language from a seller of static data packages to a provider of dynamic information access for AI. In crypto, older projects that move closer to a hot sector often get immediate attention from capital, and that appears to be a major factor behind the recent rally in GRASS.
But the long-term test remains more basic. The market still needs to see whether the company can deliver on the revenue guidance it has laid out for the next several quarters, and whether those profits will be translated into clearer and more direct support for the token. Until buybacks or profit distribution become more visible, the latest doubling in price looks, in the article's framing, more like a trade built on refreshed expectations than a full repricing of intrinsic value.
