Digital assets tied to artificial intelligence posted the strongest relative returns of any major crypto category last month, according to research published by Grayscale. In a note dated October 2, 2026, the firm’s head of research, Zach Pandl, reported that Grayscale’s Artificial Intelligence Crypto Sector advanced 54 percent in September 2026.
That result more than doubled the 24 percent gain recorded by the broader crypto market under the same six-sector classification the firm uses to organize the asset class.
The outperformance was concentrated in a handful of tokens that Grayscale tracks as proxies for blockchain infrastructure serving AI.
NEAR rose 183 percent, the largest move among the names highlighted.
Venice’s VVV token climbed 70 percent, World’s WLD gained 47 percent, and Bittensor’s TAO advanced 37 percent.
Grayscale describes NEAR as a platform oriented toward agentic commerce and notes that one of its co-founders contributed to the research behind the transformer architecture.
Venice is characterized as a consumer application that offers private access to models without retaining prompts or responses.
World is presented as a proof-of-human network co-founded by OpenAI’s chief executive.
Bittensor is framed as an open set of specialized networks covering inference, agents, data, and compute.Even after the September rally, the category remains the smallest of Grayscale’s six sectors, with an estimated market value of roughly $15 billion.
Pandl’s note treats that scale as both a constraint and an opening: the firm’s view is that the segment could still produce one or more substantial winners over the next five years if public blockchains become useful rails for payments between software agents, identity, private computation, and verifiable records.
The research stops short of claiming that September’s prices already reflect proven revenue; it presents the move as early investor interest in that infrastructure thesis.
The market reaction arrives while commercial AI development continues to expand through the closing months of the year.
Calls from some researchers and policymakers to slow the pace of model development have not produced a visible pause among the largest labs.
OpenAI and Anthropic, along with other heavily capitalized AI efforts, have kept adding capacity, product lines, and operational reach as planning cycles shift toward 2027.
Enterprise adoption of generative tools, agent-style workflows, and specialized models has continued to broaden, supporting demand for both centralized compute and any complementary systems that can handle machine-to-machine settlement or attestation.
That backdrop helps explain why a still-small crypto segment could lead monthly returns.
Investors appear to be assigning some probability to blockchains as supporting infrastructure for an economy in which software agents transact and verify activity without constant human mediation. Whether those use cases scale, and whether any of the current tokens capture durable value from them, remains an open question.
Grayscale’s figures measure only September performance and do not imply that the relative strength will persist.
The sector’s limited capitalization also means individual token moves can dominate the category return, so the 54 percent aggregate figure should be read alongside the dispersion among NEAR, Venice, World, and Bittensor.
As the year ends, the practical question for market participants is whether accelerating AI deployment creates lasting demand for the functions these networks claim to supply, or whether the September gains largely reflected narrative momentum in a thin corner of the market.
Grayscale’s research supplies the performance data and the infrastructure hypothesis; subsequent adoption and revenue evidence will determine how much of that hypothesis holds.