TechFlow Logo
Login/ Sign up
ETH Gas
Gwei
Fear
gas
Behind the $GRASS Two-Week Doubling: From Selling Data Packets to “AI Agent Internet Access,” Legacy DePIN Projects Stay Alive by Pivoting Their Narrative

Behind the $GRASS Two-Week Doubling: From Selling Data Packets to “AI Agent Internet Access,” Legacy DePIN Projects Stay Alive by Pivoting Their Narrative

2026.09.29
Share
CEO releases a long-form article, shifting valuation anchors toward the 100-billion-scale AI market, prompting the secondary market to detect a new catalyst.

Author: Gemini, TechFlow

On September 28, Grass officially released Wynd Labs CEO Andrej's long-form manifesto Abundant Intelligence. While the manifesto did not announce a new client list or a substantive buyback plan, it provides a self-consistent narrative logic for the recent price volatility of the GRASS token.

Over the past two weeks, the price of GRASS has surged from around $0.35 to above $0.70, nearly doubling.

Funds have already priced this into the market, while the long article serves more as a "post-hoc explanation": Grass is attempting to pivot its narrative, shifting from a DePIN project selling training data to a vastly larger AI real-time information layer.

Whether this new narrative can sustain the current market cap in the long run remains to be seen through the lens of its tens of millions in real cash flow versus the long-term disconnect in its token capture mechanisms.

Shifting the Narrative: From “Data Package Outsourcer” to “AI Agent Network”

For the past year, the market's understanding of Grass's business has had a clear ceiling: scraping public web pages using residential IPs and packaging them into snapshots for sale to AI labs.

While this business generates cash flow, within an industry consensus that "high-quality training data is running out," the market often views it as a transitional, phase-based venture.

I understand the core objective of yesterday's long post by the Grass CEO as breaking this ceiling expectation and forging a new narrative. To grasp the new roadmap proposed by Andrej, we must first clarify a fatal flaw in current large models: they become outdated the moment they leave production.

No matter how many trillions of tokens are fed to it during training, once weights are frozen and the model is deployed, its understanding of the world freezes at that exact moment. When users ask for real-time news or compare the latest product prices in a chat interface, the model must independently connect to the internet during the "Inference stage" to scrape the most up-to-date public web pages at that very moment.

However, large language models cannot browse the internet on their own.

The majority of today's mainstream websites deploy stringent anti-bot firewalls like Cloudflare and Datadome. When AI labs attempt to scrape web pages in real time using data center servers, requests are ruthlessly blocked by these firewalls due to the distinctive fingerprints of data center IPs. Cloudflare grew into a near-$100 billion giant precisely by "keeping machine bots out."

This is precisely the commercial story Grass aims to rewrite.

The foundational asset Grass was built upon is the real residential IP network shared by millions of everyday users running nodes at home. In the past, Grass simply used this network to batch-scrape and clean web pages in the background, packaging them into offline corpus bundles for AI labs; now, Andrej plans to transform this network directly into a "real-time proxy internet access channel" for AI models.

In this roadmap, the official team has outlined a toolkit chain designed to give AI complete access to the real world:

  • Search API (Solving “Where to Find”): Having accumulated massive replicas of public web content through years of crawling, this API acts as a dedicated search engine for AI, allowing models to horizontally index the most relevant target pages when facing complex queries.
  • Contents API (Solving “How to View”): Once a target page is located, the model no longer smashes against walls with data center requests. Instead, it accesses the page through millions of authentic residential nodes behind Grass. To the firewall of the target site, this appears as a normal browser visit from a regular household, bypassing interception mechanisms like Cloudflare entirely to open and fully render the webpage content before returning it to the model.
  • Multimodal Retrieval (Solving “How to Understand Complex Content”): The real internet contains more than just plain text; it is flooded with images, PDF reports, and embedded videos. This module aims to enable real-time models to directly comprehend non-textual information.

Following this logic, it becomes easy to understand why some overseas crypto analysts currently hold a bullish view on GRASS.

If Cloudflare’s business logic is “charge websites and keep AI bots out,” then the business model Grass aims to build is its exact inverse: “charge AI companies and use human residential networks as disguised channels to let AI in.”

With this, Grass’s capital narrative undergoes a complete refloating: It is no longer a “off-chain data outsourcing” company trading at a steep discount based on traditional DePIN hardware multiples, but rather packages itself as an “AI real-time retrieval layer and agent gateway,” similar to Exa and Parallel, which command high valuation multiples in primary markets.

This is also why, even without any announcements of substantive new client contracts, this logically consistent roadmap alone has been enough to spark secondary market speculation around this narrative shift.

Is There Real Revenue?

In the current market cycle, investors strongly favor projects backed by genuine revenue and internally logical frameworks.

By this principle, examining GRASS's current fundamentals reveals a growth-stage data company with tens of millions in actual annual revenue. According to data disclosed by the official team in the July Token Holder Call, its ability to generate cash has already moved past the early proof-of-concept stage:

  1. Full-year revenue for 2025 was $17 million. For just the first half of 2026 (H1), revenue hit another $17 million, marking nearly a 7x year-over-year growth.
  2. The company expects revenue from its “training data” segment alone to reach $65 million to $75 million for the full year of 2026. The business exhibits clear seasonality, with some major deals originally slated for H1 (approximately $15 million) pushed into Q3 following shifts in AI labs’ training cycles.
  3. The company is currently profitable on paper. By building out its own compute and storage infrastructure in 2025, monthly cash operating expenses (OpEx) have been compressed to between $2 million and $3 million (primarily for infrastructure), far lower than its revenue intake.

In terms of architectural design, Grass avoids the traditional Web3 pitfall of “equity companies eating meat while token holders sip broth.” Its customer contracts and business revenues flow directly into Grass DataCo and the foundation, while the underlying development team, Wynd Labs, exists merely as an engineering and business services provider.

Theoretically, this ensures that all protocol revenue belongs entirely to the network itself.

Who Will Pay for GRASS?

The business model works and the architecture is relatively clean, but “how profits will flow back to the token” remains Grass’s greatest test.

Under the current mechanism, Grass operates on typical “growth-stage tech startup” logic, which creates severe friction with the “crypto logic” of secondary markets craving liquidity.

  1. Highly Restrained Buyback Scale

For secondary market traders, tangible buy-and-burn programs remain the only hardcore mechanism for value capture. Yet, the official Grass team has shown extreme reluctance in this area. Aside from a minor ~$350,000 open-market buyback in November 2025, large-scale repurchases have essentially stalled.

The official rationale is that reinvesting the vast majority of profits into compute, storage, and network expansion yields a much higher long-term return than propping up prices directly in the secondary market. While this heavy-asset investment solidifies the business moat, it leaves the token without the most direct bid-side support in the short term.

  1. The Reality of Node Incentives

In the network's second stage (Stage 2), Grass implemented a critical tokenomic shift: abandoning token inflation to subsidize nodes, and instead funding rewards with real USDC revenue generated by the protocol.

From a token inflation perspective, this cuts off bottomless sell pressure on the circulating supply, representing an absolute bull case. However, from a physical supply side of the network, contradictions are emerging.

Some ordinary residential node operators offering broadband have discovered they only receive pennies to a few dozen cents worth of stablecoins every few months, prompting them to start shutting down. The foundation of Grass’s new narrative about “breaking through Cloudflare blocks” relies entirely on these millions of real residential nodes. If low returns cause the underlying network to contract, its business moat will suffer physical erosion.

  1. Insufficient Yield Experience for Stakers

Although a July governance vote approved allocating a portion of USDC revenue to GRASS stakers, the actual payout size and revenue-sharing ratios remain opaque in public disclosures.

When token holders cannot feel those tens of millions in protocol revenue through intuitive Annual Percentage Yield (APY) dashboards, the token price is left oscillating repeatedly between earnings expectations and sell-offs triggered by early allocation unlock panics (such as tail-end investor unlocks at the end of October).

Thus, Grass has completed a contextual shift from a “static data package company” to an “AI dynamic information layer” via a new roadmap.

In the capital games of crypto markets, this kind of narrative upgrade—where an established project pivots toward a trending sector—often serves as the most effective short-term catalyst.

But once short-term sentiment dissipates, real-world commercial fundamentals will ultimately dictate GRASS’s long-term pricing. The market’s ultimate test will not be how cutting-edge the APIs described in the long article are, but whether operations can deliver on those tens of millions in financial guidance over the next six months. The core contradiction lies in determining when this highly cash-generating entity will be willing to convert its paper profits into tangible buy-side pressure for GRASS tokens in the secondary market.

Until the narrative and hard cash fully align, the CEO’s statement merely gives capital a reason to re-enter. If clear token buybacks and profit distributions are delayed indefinitely, this price doubling ignited by expectations will ultimately prove to be just a brief capital game, rather than a genuine revaluation of the asset's value.