Grayscale's Head of Research Names 3 Cryptocurrencies Tied to the AI Buildout
Grayscale's Zach Pandl named DATA, USD.AI and Pearl as crypto plays on AI infrastructure, with 90-day price moves from -42% to +528%.
AI SummaryAI
- Grayscale head of research Zach Pandl named DATA, USD.AI and Pearl in an October 23 note.
- DATA fell 42.07% over 90 days to $0.15243 despite a $30 million Sui Foundation investment.
- Bullish announced a $100 million financing facility backing USD.AI's GPU loans in August.
- CHIP gained 47.98% over 90 days to $0.04386, still roughly 69% below its April peak.
Update 00:50 UTC: On USD.AI, the note adds that the platform's loans are non-recourse and stablecoin-denominated, and that a single $98.1 million facility was struck in June against 2,304 NVIDIA B300 GPUs.
Pandl's Three-Token AI Thesis
Few desks in digital assets sit closer to institutional capital than Grayscale's research team, and on October 23 that team published a thesis with a number attached: three tokens it tracks for the buildout of artificial intelligence infrastructure. Zach Pandl, the firm's head of research, identified DATA Network (DATA), USD.AI (CHIP) and Pearl in the note, and an official X post from Grayscale summarized the same list for its followers the same day. The premise behind the picks is demand. Training data is the raw input, chips are the bottleneck and compute is the revenue line, and the note treats each layer as investable in its own right. AI models consume enormous volumes of images, videos and voice recordings during training, and the graphics processors that run those workloads are scarce and expensive. Pandl's argument is that crypto is converging with frontier technology exactly where those needs bite, in the verification of training data, the financing of hardware and the monetization of computing capacity itself. The three tokens divide that work rather than competing for it. DATA Network plans to collect training material with the permission of its creators, license it to AI developers and reward the people who contribute it. USD.AI runs blockchain-based lending pools that let computing companies borrow against their GPU fleets and the income those fleets are expected to generate. Pearl takes a different split of the same machines: providers would run AI workloads and secure a blockchain at the same time, earning revenue from AI customers while also collecting tokens for proof-of-work mining. In the note's framing, each project occupies a different layer of the buildout, so the list reads as a map of the trend rather than three bets on the same service. The demand behind it is mechanical rather than speculative: every new model needs more licensed material to learn from and more processors to train on.
@Grayscale · X post
Official X post.
View on X
Underneath the thesis, the price record of the three names could hardly be more divided. DATA is down 42.07% across 90 days at $0.15243, with a 34.65% drop inside the past month alone. That weakness came despite fresh institutional money: on October 12, the Sui Foundation put $30 million into the network, alongside a planned migration to Sui in which existing tokens transfer one-for-one, so holders keep their balances through the move. USD.AI carries institutional backing of its own. The backing matters for a lending design, because GPU-backed loans need committed capital before the pools can scale. In August, Bullish committed a $100 million financing facility behind the project's GPU loans, a dedicated credit line for the onchain lending model. The market has warmed to it: CHIP has gained 47.98% over 90 days to $0.04386, though the token still sits roughly 69% below its April peak, so the rally only partly retraces a much deeper drawdown. PEARL is the outlier on the upside. The token has climbed 528.24% across three months to $2.00 and is up more than 240% year-to-date per CoinGecko price data. It has also already given some of that back, retreating about 19% from a record high set on October 21, two days before the note and after the strongest 90-day run of any name on the list. Supply is the risk the research itself surfaces: an estimated 1.2 million new PEARL tokens come into circulation every day, and the project's $4.2 billion fully diluted valuation far exceeds its $713 million market cap, a gap that represents a large quantity of tokens still to arrive. Pandl closed the note with caution rather than a call. “Their long-term relevance will depend on adoption and competitive economics. We are keeping an eye on them,” he wrote, framing all three as watchlist names whose success remains uncertain rather than rated products.
A follow-up reading of Grayscale's note, published October 24, adds operating figures the original article's price table did not carry. It cites CoinGecko data as of October 22 showing PEARL up 580% over three months, a higher reading than the 528.24% figure already on record, and puts Pearl's compute capacity up 93.5%. On the funding side, the note reports DATA holding 280 million verified records, while USD.AI has deployed more than $400 million of capital. The lending terms also get their first detail: GPU loans can cover up to 80% of deployment cost after credit review, with borrowers funding at least 20% themselves and repaying principal over three to five years.
The rebrand and structure behind the two smaller names also get fuller treatment in the follow-up coverage. DATA Network is the former Story, an IP-focused blockchain that pivoted to AI training data when it rebranded in June and swapped its old token for DATA one-for-one between June 5 and June 7; once the migration to Sui completes, the project will drop its own layer 1 and rely on Sui's decentralized storage layer Walrus. On USD.AI, the note adds that the platform's loans are non-recourse and stablecoin-denominated, and that a single $98.1 million facility was struck in June against 2,304 NVIDIA B300 GPUs. Pearl's 528% run also carries an unresolved technical question: a research paper argues its mining performs random matrix multiplications that mimic AI compute without the protocol verifying the work comes from real inference or training.
Issuance Math Behind PEARL
The spread in the numbers is the real content of the list: a 42% drawdown, a 48% rally that still leaves a token 69% under its April high, and a 528% surge already 19% off its October 21 high. For PEARL, the arithmetic is the pressure point. At $2.00, an estimated 1.2 million new tokens a day amounts to roughly $2.4 million of recurring issuance, nothing like a one-time airdrop, while the $4.2 billion fully diluted valuation against a $713 million market cap is the token version of a P/E ratio paid on coins not yet circulating. Market makers must absorb that daily flow, and that is where the sharpest FUD around the name sits. Pandl's caution that adoption and competitive economics will decide outcomes matches the price record; keeping watch, not calling a winner, is the stance the note takes.
Primary sources
- x.com
- CoinGecko price data · coingecko.com
AI-generated, AI-reviewed, under COINOTAG editorial oversight.

