Grayscale breaks down AI token infrastructure roles

Grayscale published a research note on October 23 that examines specific infrastructure roles within the intersection of cryptocurrency and artificial intelligence. The report, authored by research head Zach Pandl, moves beyond general speculation on AI software to analyze who supplies training data, who finances hardware, and who receives payment for computing capacity.

The note highlights three specific tokens as examples of different infrastructure roles: DATA, USD.AI, and Pearl. Grayscale describes these as selected examples and expectations rather than independently proven outcomes for the entire sector.

Three Infrastructure Roles

The report distinguishes between three primary business models. DATA is described as working on contributed training data. USD.AI is identified as engaging in GPU-backed lending. Pearl is described as attempting to combine AI computing with blockchain rewards.

Each model requires a different type of customer to function. A data network requires customers willing to license its datasets. A financing protocol requires borrowers who can service debt. A computing network requires useful workloads and a method to measure delivery. The report notes that combining these distinct models under a single AI-token heading can obscure more than it reveals.

Credit Mechanics and Token Rights

USD.AI’s documentation describes loans secured by GPUs and associated cash flows. Recourse is generally limited to the collateral and specified exceptions. The protocol’s yield-bearing instrument draws on loan payments and reserve-related income.

This structure introduces traditional underwriting questions into a new distribution format. The report raises issues regarding equipment value loss, customer concentration, and collateral recovery if a borrower fails. A blockchain record can improve visibility into a claim, but it does not answer all underwriting questions.

The report also notes that a useful service and a valuable token are separate propositions. USD.AI’s depositor guide separates the non-yield-bearing USDai instrument from the income-accruing sUSDai instrument and a CHIP governance token. Naming the protocol alone does not identify which specific exposure a holder receives.

For data networks, sustainable demand depends on whether buyers find the data lawful, relevant, and difficult to obtain elsewhere. Token incentives can finance initial activity but do not prove that customers will continue paying once those incentives shrink. The report suggests that adoption, unit economics, and enforceable claims provide stronger evidence of durable businesses than token price action alone.