Grayscale’s head of research has outlined three cryptocurrency projects positioned at distinct points in the artificial intelligence infrastructure chain. In an October 23 analysis, Zach Pandl connected DATA Network to the supply of training data, USD.AI to the financing of graphics processing units, and Pearl to the generation of revenue from AI computation itself.

The note frames the projects as responses to concrete bottlenecks. AI models require large volumes of images, videos and voice recordings for training, while the specialized processors that run those workloads remain costly and constrained. Pandl’s assessment treats each layer as separately addressable rather than as competing alternatives.

DATA Network Focuses on Licensed Training Material

DATA Network aims to gather human-contributed photographs, videos and voice recordings with the permission of creators, then license the resulting datasets to AI developers while compensating contributors. Tools for collection, licensing and tracking form the core of the design.

On October 12 the Sui Foundation announced a $30 million investment in the network and a planned migration of the protocol onto the Sui blockchain. Existing tokens would convert on a one-for-one basis, after which Walrus would serve as the default storage layer for applications built by the DATA Foundation. No completion date has been set. Trace, the contribution system, has registered more than 250 million items since its June launch and processes an average of more than three million receipts daily. Separate Grayscale figures place verified records near 280 million.

Despite the institutional commitment, DATA has declined 42.07 percent over the past 90 days to approximately $0.15243, including a 34.65 percent drop in the most recent month.

USD.AI Channels Capital Into GPU-Backed Loans

USD.AI operates blockchain-based lending pools that allow computing firms to borrow against their GPU fleets and the expected income from those machines. Loans can cover up to 80 percent of deployment costs, subject to credit review, with borrowers supplying the remaining equity and repaying principal over three to five years. The CHIP token serves utility and governance functions and does not represent equity or a claim on protocol assets.

In August the crypto exchange Bullish provided a $100 million financing facility dedicated to these GPU-backed loans. Grayscale notes that more than $400 million has already been deployed through the platform. CHIP has risen 47.98 percent over 90 days to about $0.04386, though the token remains roughly 69 percent below its April peak.

Pearl Combines AI Workloads With Proof-of-Work Mining

Pearl proposes that the same GPUs perform AI matrix-multiplication tasks and secure a proof-of-work blockchain simultaneously. Operators would collect payments from AI customers while earning network rewards for mining. A June 2026 preprint examining dominant mining software found limited evidence of inference code in practice, though those findings do not describe current network-wide performance.

PEARL has been the strongest performer among the three, advancing 528 percent to 580 percent over three months to approximately $2.00 and more than 240 percent year-to-date. The token has since retreated about 19 percent from a record high near $2.46 set on October 21. An estimated 1.2 million new tokens enter circulation each day, and the project’s fully diluted valuation of $4.2 billion substantially exceeds its $713 million market capitalization. Grayscale data also record a 93.5 percent rise in hash rate.

The research assigns no investment rating, price target or fund inclusion to any of the three tokens. Success, according to Pandl, hinges on whether the projects achieve meaningful adoption and maintain viable economics against competing solutions.

Together the three projects illustrate how blockchain designs are being applied to discrete stages of the AI supply chain. Whether any of them achieves lasting commercial traction remains an open question that Grayscale continues to monitor.