AI Agent Token Revenue: Real Usage or Subsidies?

When looking at AI agent token revenue, first assume there are heavy subsidies, then identify how much genuinely comes from real usage.

The design of AI tokenomics naturally blends subsidies with real revenue. Only by clarifying how much revenue comes from actual use can you avoid being misled by superficial on-chain transaction volume.

Prerequisites

• Access to the project's disclosed revenue source explanations or whitepaper.
• Know whether the token has a "buyback and burn" mechanism and its trigger conditions.
• Distinguish between "protocol revenue" (fees users pay to the protocol) and "token price" (a result of market trading).

Step 1: Directly Check the Revenue Distribution Mechanism

The core of token revenue lies in how value is distributed — this is the dividing line between real usage and subsidies.

How to Do It

Go to the project whitepaper or official documentation and search for "revenue" or "fee distribution". Take Virtuals Protocol as an example: in its tokenomics, when a user interacts with an AI agent by paying $VIRTUAL, the revenue enters the agent's wallet. The protocol then uses that money to buy back and burn the agent's token from the market.

Key Judgments

• If the revenue distribution description includes phrases like "the protocol subsidizes the shortfall to ensure developer cash flow", it means a portion of the revenue is supplemented, not directly paid by users.
• If it explicitly states "60% goes to the agent wallet, 30% to buyback and burn, 10% to the treasury", this indicates a real revenue-share mechanism, not airdrops or subsidies.

Completion standard: You confirm the project's revenue source structure — whether it comes from user-paid usage fees or from a protocol subsidy pool.

Step 2: Distinguish "Real Revenue" from "Protocol Subsidies"

Virtuals charges a 10% fee from ACP transactions and places it into an ecosystem incentive pool, allocating a monthly budget of $1 million to support AI agent developers. The subsidies are distributed based on "real service output", using a leaderboard system combined with on-chain data to automatically exclude artificial activity.

How to Do It

See whether the project discloses "subsidies as a proportion of total revenue". If a project has never published this figure, its revenue composition is opaque, and subsidies likely account for a large share. We currently cannot find any project that publicly discloses the precise ratio of subsidies to real revenue.

Completion standard: You assess whether the project's subsidies are sustainable — if subsidies are the majority and not funded from the protocol's own revenue, once subsidies stop, token revenue will cliff-drop.

Step 3: Look at Changes in Model Providers' Billing Models

The true cost of AI usage is itself becoming transparent. In May 2026, Anthropic announced it would split subscription billing into two pools: "interactive chat" and "Agent SDK usage". Under the Max 20x plan ($200/month), the Agent SDK allowance is only $200. GitHub Copilot also switched to usage-based billing on June 1, 2026.

What this means: The real cost of using AI agents is rapidly being exposed, and the subsidies previously hidden by flat subscription fees are retreating. If a crypto AI project's token revenue still heavily relies on a "fixed-rate subsidy" logic while actual user costs are already metered, its revenue sustainability must be re-evaluated.

Common Pitfall

Equating "token buyback and burn" with "real usage revenue". Buyback and burn can be funded by a protocol subsidy pool or by genuine user payments. The former is artificially manufactured buy pressure; the latter is sustainable value capture.

Risk Warning

AI agent tokens are currently in a "speculation first, revenue verification later" phase. If a project's main revenue source is trading fees (users buying and selling tokens on secondary markets) rather than agent service fees (users paying for actual AI services), its revenue is essentially speculative activity, not cash flow from real usage. When inference subsidies end, agent tokens lacking real usage support will face significant value re-evaluation risk.

How to Verify After Operation

Open the project's official dashboard (if available) and look at the "30-day revenue" and "30-day trading volume" metrics. If revenue is far lower than volume (e.g., below 1%), the project's income heavily relies on token trading fees rather than real service usage. If you cannot find a breakdown of revenue composition, treat the project as opaque. Verification channels: project official whitepaper, Dune dashboards (if any), financial reports or monthly summaries proactively disclosed by the project.