Grayscale Flags DATA, CHIP and PEARL as AI Infrastructure Plays

The meeting point of artificial intelligence and crypto has produced many bold claims, but few projects that can point to a clear job inside the AI supply chain. That is what makes the latest research note from Grayscale worth reading closely. Grayscale Head of Research Zach Pandl identified three cryptocurrencies in a research note published October 23 — DATA Network (DATA), USD.AI (CHIP), and Pearl (PEARL) — targeting specific gaps in the artificial intelligence infrastructure supply chain. Each addresses a different bottleneck. One deals with data, one with hardware finance, and one with compute itself. Together, they offer a useful map of where crypto builders think value might sit as AI systems grow larger and more expensive to run.

Grayscale is not a small voice in this conversation. As one of the largest managers of crypto investment products, its research is read by professional investors who want a framework, not a tip.

Why AI Needs New Infrastructure

Modern AI models are built on three inputs: data, compute and capital. Data trains the model. Compute runs the training and the work that follows. Capital pays for the hardware, the energy and the people. In the traditional technology world, all three are controlled by a small number of very large companies. They own the largest datasets, buy graphics processing units in bulk, and finance expansion from their own balance sheets. Smaller developers and open-source teams often struggle to get access on fair terms.

Crypto advocates argue that tokenised networks can open up each input. A network can reward people for contributing data they own. It can let lenders finance hardware directly, with the loan settled on-chain. It can let owners of computing power sell spare capacity to buyers who need it, without going through a central cloud provider.

DATA Network: Paying People For Training Data

DATA Network focuses on collecting and licensing AI training data — images, videos and voice recordings with explicit permission from contributors — and rewarding providers. That description contains two ideas often missing from debates about AI data. The first is consent. Much of the data used to train leading models was scraped from the public internet without a direct agreement with the person who created it. That has led to lawsuits and growing demand for licensed datasets. The second is payment. If a contributor’s photo, video or voice helps train a valuable model, the contributor should be paid, not merely thanked.

DATA’s model is a marketplace where contributors supply data with explicit permission and receive rewards. Buyers — AI developers who need high-quality, legally clean training data — can license that data knowing where it came from and that the contributor agreed.

The project received a notable vote of confidence when the Sui Foundation announced a $30 million investment in DATA Network on October 12. That is a substantial commitment, and it signals that Sui sees DATA as a meaningful part of its plans. DATA plans to migrate to the Sui blockchain with tokens transferring one-for-one, which means existing holders would move across without a change in the number of tokens they hold, while the network runs on Sui infrastructure. Despite the backing, DATA has declined 42.07% over the last 90 days to $0.15243, and fell 34.65% over the past month. A $30 million investment does not guarantee traders will value the token more highly the next day. It may reflect a long-term view, while the price reflects short-term sentiment. Readers should hold both facts at once: a serious backer has committed capital, and the token has fallen sharply.

USD.AI And CHIP: Financing The Hardware

If DATA is about the raw material for AI, USD.AI is about the machines that process it. CHIP (USD.AI) finances hardware through on-chain lending. Graphics processing units, known as GPUs, are the workhorses of AI training and inference. They are expensive, often in short supply, and buying them in quantity requires large amounts of upfront capital. Many smaller operators who want to run AI workloads cannot afford to buy the hardware outright, even if they have customers waiting.

USD.AI offers a different route: finance the hardware through lending arranged and settled on-chain. A borrower can access capital to acquire GPUs, and a lender can earn a return for providing that capital, with the terms recorded on a public ledger.

The model received a significant boost when Bullish announced a $100 million GPU-loan facility. A facility of that size suggests that at least one large, well-known crypto firm sees real demand for GPU financing and is willing to put substantial capital behind it. CHIP rose 47.98% in 90 days to $0.04386, a sharp contrast with DATA’s decline over the same period. That divergence shows the market is not treating all AI infrastructure tokens as one group. Investors appear to be distinguishing between models, backing and recent news. A token tied to a live financing facility may be rewarded, while a token tied to a longer-term data marketplace may be sold while the market waits for proof of adoption. The risks for CHIP are also clear. Lending against hardware carries credit risk: borrowers can default, hardware can lose value, and demand for compute can fall. On-chain lending does not remove those risks; it changes how they are managed.

Pearl: Mining Meets AI Compute

PEARL merges proof-of-work mining with AI compute revenue. That combination is the most unusual of the three. Proof-of-work mining is the original crypto security model, made famous by Bitcoin. Miners compete to solve computational puzzles, and the winner earns newly issued tokens.

Pearl’s idea is to take that mining infrastructure and give it a second job: selling AI compute. The same hardware and operational expertise that secures a network could also earn revenue by providing computing power to AI customers. If that works, the project would have two sources of value — the token issuance from mining, and the income from selling compute — rather than relying on token issuance alone. That could make the economics more sustainable, at least in theory, because real customer revenue would support the network.

The market response has been dramatic. PEARL surged 528.24% in 90 days to $2.00. The risks are substantial. The note points to estimated daily issuance of 1.2 million tokens and a $4.2 billion fully diluted valuation versus a $713 million market cap. Daily issuance of 1.2 million tokens means a large amount of new supply is entering the market every day. If demand does not grow at the same pace, that supply can weigh on the price. Fully diluted valuation imagines what the project would be worth if every token that will ever exist were already in circulation. Market capitalisation counts only the tokens actually circulating today. A $4.2 billion fully diluted valuation against a $713 million market cap means a large share of the total supply has not yet entered circulation. As those tokens are released, holders will face dilution unless demand grows fast enough to absorb them.

Three Models, Three Different Tests

What makes this trio useful is how different their tests are. DATA must prove that people will contribute high-quality data, that buyers will pay for it, and that a permissioned marketplace can compete with scraped data on price and scale. That is a marketplace test. CHIP must prove that on-chain GPU lending can be underwritten safely, that borrowers will repay, and that lenders will keep providing capital at a price borrowers can afford. That is a credit test. PEARL must prove that mining hardware can earn real AI compute revenue, that customers will buy that compute, and that token issuance does not overwhelm demand. That is an operations and supply test. Grayscale’s point that their market performance and adoption trajectories have diverged significantly over the past three months is therefore not surprising. Different tests produce different results at different speeds. A lending facility can be announced quickly. A data marketplace takes longer to build, because it needs many contributors and buyers. A compute business takes time to sign customers and prove reliability.

For DATA, the key milestones are the migration to Sui and growth of the marketplace. For CHIP, the key question is whether the $100 million facility translates into sustained lending. For PEARL, the key question is whether compute sales can grow fast enough to absorb daily issuance. None of these questions has an answer today. That is the point of research like this: to name the questions clearly, so readers can watch for the answers as they arrive.

FAQ

What did Grayscale’s Zach Pandl identify in the October 23 research note?

Grayscale Head of Research Zach Pandl identified three cryptocurrencies — DATA Network (DATA), USD.AI (CHIP), and Pearl (PEARL) — targeting specific gaps in the artificial intelligence infrastructure supply chain.

What does DATA Network do?

DATA Network focuses on collecting and licensing AI training data, including images, videos and voice recordings with explicit permission from contributors, and rewarding providers.

Why has CHIP performed differently from DATA over the past 90 days?

CHIP rose 47.98% in 90 days to $0.04386 after Bullish announced a $100 million GPU-loan facility, while DATA declined 42.07% over the same period to $0.15243, showing that their market performance and adoption trajectories have diverged significantly.

What is the main risk flagged for PEARL?

PEARL surged 528.24% in 90 days to $2.00, but estimated daily issuance of 1.2 million tokens and a $4.2 billion fully diluted valuation versus a $713 million market cap pose risks, because new supply can weigh on price unless demand grows to absorb it.

Source: The Currency Analytics

Related reading on MM Crypto Analysis Web

• AI Agent Payments Risks: Spending Limits May Not Stop Rogue Machine Spending
• Anthropic’s $80.7B Loss Sinks AI Crypto Tokens as Traders Question Valuations
• AI x Web3 Game Layer SynapLogic Hit in $186K Exploit on Base

More in AI Crypto on MM Crypto Analysis Web.