AI Tokens Rose 54% in September, Then Fell With Chip Stocks
AI tokens gained 54% in September, then slid with chip stocks on Oct. 13. What drove both moves, and why most still sit far below their old highs.
AI tokens had a big September in 2026: Grayscale's AI crypto sector gained 54%, against 24% for the broader crypto market, led by NEAR at +183%. Then on October 13, AI-linked tokens fell as much as 10% in a day as chip stocks sold off in Asia. Both moves point the same way. AI tokens still trade mostly as a bet on the AI story, and most of them remain far below their old highs.
Key takeaways
- Grayscale's AI sector rose 54% in September 2026, yet it is still the smallest of Grayscale's six crypto sectors, at about $15 billion.
- On October 13, a semiconductor selloff took South Korea's KOSPI down 6.5%, and AI-linked tokens fell as much as 10%.
- On our tracker, 70% of AI tokens worth $100 million or more trade at least 70% below their all-time highs (October 14, 2026).
- A strong month is not a recovery. Momentum counts least of the five questions the AI Score asks.
What happened in September?
Grayscale's Head of Research, Zach Pandl, reported on October 2 that the firm's Artificial Intelligence crypto sector "meaningfully outperformed its peers in September," gaining 54% while the broader market gained 24%. Four tokens led it:
- NEAR (+183%): Grayscale describes it as a platform for agentic commerce, co-founded by a researcher behind the transformer architecture.
- Venice (VVV, +70%): a privacy-focused AI chatbot that gives access to AI models without storing users' prompts or responses.
- World (WLD, +47%): a proof-of-human network.
- Bittensor (TAO, +37%): an open ecosystem of specialized AI networks.
Grayscale's explanation is that investors are starting to price in a role for public blockchains as infrastructure for AI agents: payments, identity, private compute and verifiable records. That is a reasonable thesis. It is also, for now, a thesis. A single month of prices tells you what investors expected, not what any of these networks earned.
Why did AI tokens fall with chip stocks?
On October 13, South Korea's KOSPI index fell 6.5% to about 6,002 after Chinese chipmaker CXMT announced a new memory-chip design, and AI-linked crypto tokens fell as much as 10%, TokenPost reported. Nothing about the crypto networks themselves had changed that day.
The link is the buyer, not the technology. Many investors hold AI tokens as a way to own "AI" alongside chip and cloud stocks. When a shock hits AI hardware, those investors cut AI exposure across the board, and tokens usually get sold faster than stocks because they trade around the clock and have thinner order books. On our tracker, Render fell 7.7% and NEAR 4.9% in the 24 hours to early October 14.
Correlation like this is not new. Our guide to why Bitcoin's volatility drags the whole market explains the same pattern with Bitcoin as the anchor; here the anchor was AI equities.
How far below their highs are AI tokens?
A 54% month sounds large until you set it against the starting point. FXStreet noted on September 25 that most major AI tokens were still 70% to 90% below their 2024–25 highs. Our own data from October 14 agrees:
| Measure (AI tokens worth $100M+ on AiTokens, Oct 14, 2026) | Result |
|---|---|
| Tokens in the group | 33 |
| Median distance below all-time high | 86% |
| Share at least 50% below their high | 82% |
| Share at least 70% below their high | 70% |
| Share at least 90% below their high | 45% |
Even NEAR, up 114% over 30 days on our tracker, sat about 75% below its January 2022 high. TAO was about 63% below its March 2024 high, Render about 85% and FET about 94%.
Why does the gap persist while AI companies boom? FXStreet's answer is that money is flowing to AI businesses with measurable revenue, while many tokens have weak links between usage and token value. That matches what we see: a token can sit inside a booming industry and still capture little of it.
Why do different sources give different sector sizes?
Grayscale puts the AI sector at about $15 billion. Our tracker counts every AI-labeled token we follow, 676 of them including AI memecoins, worth about $27 billion combined on October 14. Neither number is wrong; they measure different things. When you see a headline figure for "AI crypto", check which tokens it includes before comparing it with another.
What the AI Score shows
The AI Score rates every AI token from 0 to 100 on five questions: is anyone building it, can you get in and out, will unlocks dilute you, how badly could it hurt you, and is the market agreeing. Development counts most and momentum counts least, on purpose. A month like September lifts momentum, but it can't carry a score on its own, which is the point: a price surge is the easiest signal to get and the hardest one to keep.
What to watch next
- Grayscale's next monthly note, usually in the first days of the month, for whether the sector held its gains in October.
- AI equities and chipmakers. Until AI tokens show demand of their own, expect them to move with AI stocks on big days.
- Usage that rises with price. Fees, burns and paid workloads are the evidence that a rally has something underneath it.
- Supply. Rallies are when unlocks get sold into. Check a token's supply before reading too much into its chart; our tokenomics guide explains what to look for.
Check any token's AI Score on AiTokens.app before you decide.
Frequently Asked Questions
Why did AI tokens rise 54% in September 2026? Grayscale attributes the move to investors beginning to price in blockchains as infrastructure for AI agents: payments, identity, private compute and verifiable records. NEAR, Venice, World and Bittensor led the sector. The figure is Grayscale's index return, not every AI token's.
Why do AI tokens fall when chip stocks fall? Many investors hold AI tokens as one more way to own the AI theme. When AI hardware stocks sell off, those investors reduce AI exposure everywhere, and tokens, which trade around the clock with thinner liquidity, often fall faster.
Are AI tokens back to their highs? Most are not. On October 14, 2026, the median AI token worth $100 million or more on our tracker sat 86% below its all-time high, and 70% of them were at least 70% below.
Is the September rally a reason to buy AI tokens? No single month is. A rally shows what investors expected, not what a network earns. This article is research, not a recommendation to buy or sell anything.
Sources
- Grayscale Research, Zach Pandl: The AI Trade Reaches Crypto with Enthusiasm, October 2, 2026
- TokenPost: South Korea's KOSPI Falls 6.5% as AI Crypto Tokens Slide, October 13, 2026
- FXStreet, John Isige: AI is booming but AI crypto tokens are missing the rally, September 25, 2026
- AiTokens.app: AI token tracker, market data from CoinGecko, snapshot of October 14, 2026, 05:44 UTC
This article is for educational purposes and is not financial advice.
Research and education only — not financial advice. Prices and figures change; check live data before acting on anything here.
