Last night I opened MetaMask and one screen made me stop scrolling.

It was a trader's profile. 72% win rate, over $272,000 profit in seven days, and an average hold time of 2.9 hours. Under three hours per trade. I stared at it longer than I want to admit. I wanted it to be true.

Then I checked the positions. A token literally named "AI" was up almost 600%. Another meme-style token was up more than 18,000%. One position was down 27%, and I was almost glad to see it. Not everything on these screens is green. That one red number felt like the only honest thing on the page.

My first thought was "is this the future?" My second was "is any of this even about artificial intelligence?" That's why I'm writing this.

TL;DR: AI tokens are only about 0.7% of the crypto market, most of the money goes to a few names, and plenty of tokens with "AI" in the name have nothing to do with AI. Some can still grow, but mostly the ones with real users. And that $272K? Probably not cash. Probably a screenshot.

What AI actually does in crypto

It's already in the market. Bots and prediction models make a big part of the trades, so prices react faster and short-term moves are sharper. I felt this myself last week. I watched a token move 15% in four minutes while I was still reading the chart. By the time I understood what happened, it was over. Nobody can keep up with that by hand.

Security goes both ways. AI catches suspicious transactions and smart contract bugs, and scammers use it too for better phishing and fake videos. I've gotten fake support DMs that were so polished I almost replied. A year ago they had typos. Now they don't. That worries me more than a bad trade.

The part I like most is AI agents. Software that can pay and buy on its own. It can't open a bank account, but it can use a crypto wallet, and that fit makes sense to me. This is the only part of the whole story that makes me lean forward instead of squinting.

The tokens are not all the same thing

This is where I got confused at first, and I think most people do.

Bittensor runs a network where people compete to provide AI services. I like the idea. It's basically a market for machine intelligence. Virtuals and FET are built around agents. Grass pays people for data, which sounds boring until you realize how much companies pay for data. Render rents out GPU power, and if you've ever tried to rent a GPU, you know why that matters.

Just putting "AI" in the name says nothing about which one you're looking at. I've seen tokens called "AI Dog" and "Pepe AI." Those are just memes with AI in the name.

The size surprised me

One tracker puts the whole sector at about $18.5 billion, roughly 0.7% of all crypto. Others count more tokens and land higher, up to around $30 billion. Either way, it's small.

When I first saw that number, I didn't believe it. Then I thought: all this noise, all these tweets, all these YouTube thumbnails with robots on them, and it's under 1%? That tells you something. The noise is way bigger than the market.

One market researcher said money is going to fewer names. Nearly a thousand tokens carry the AI tag, and only a handful get real money. Most of the others have almost no activity.

The main names (approximate, September 2026)

I'm keeping this short, because a list of prices is not why I'm writing this.

NEAR: around $2.30 a few days earlier, then a jump to about $4.25. That's more than 80% in a week, and now over $5 billion. I didn't buy it. I just watched it and felt that familiar mix of "I missed it" and "I'm glad I didn't chase it."

Bittensor (TAO): around $265, up about 13% this week. Hard cap of 21 million coins. I like the hard cap. It's honest math.

Venice (VVV): about $30, up roughly 40% this week. It sells private access to AI models. That's a business I can actually understand, which is rare in this space.

Render (RENDER): around $1.70, up about 20% this week.

FET: around $0.17, roughly flat this week.

Internet Computer (ICP): around $2.90.

Virtuals (VIRTUAL): about $0.67. It hit around $5 once. I remember when people were calling it the future. That's how fast things change here.

One token nearly doubled while another barely moved. So when someone says "the AI sector is up," it's usually two or three tokens with their own reasons. The rest are just watching.

The war between the AIs

Behind all this there's a bigger fight. OpenAI, Google, Anthropic, xAI, Meta, and Chinese labs like DeepSeek and Alibaba are all racing each other. Several released new models in the first ten days of September, and the top spot on the leaderboards keeps changing. DeepSeek keeps pushing prices down with cheap open models.
I use these models. I test them. The pace is exhausting. Every week there's a new "best model ever," and by Friday it's already second place.
AI is ending up in very few hands, and crypto's pitch is the opposite: open networks where anyone can contribute computing power or data and get paid. That's what Bittensor, Render and Venice try to do. I want that to work. I'm just not sure it will at the scale people promise.
My pick to win is Anthropic. It's my opinion, nothing more. Benchmark trackers currently put its newest Claude models at or near the top, and I like that the company built its name on safety and trust. That will matter more when AI agents start handling real money. OpenAI and Google still have huge reach though, and one cheap open model could change everything overnight. I could easily be wrong.
The crypto side is not an even fight. These networks are tiny next to big tech. Money is pouring into AI everywhere, and very little of it reaches AI tokens. The technology can win while the token loses, and that gap is where people get hurt. I've seen people get burned by this.
There's also a quieter war, AI against AI. Trading bots compete for tiny price gaps. Scam tools get smarter, and so do the systems built to catch them. And in the middle is a regular person looking at a green number on a phone screen, wondering if they missed something.

Can they still grow?

Some of them, yes. I actually believe that. If real usage keeps growing (GPU jobs, data, agent payments) and the token is actually needed for it, the best projects could grow a lot, since the whole sector is under 1% of the market.
But I think we're in the middle scenario. A few projects with real users do fine and most go nowhere. A big crypto drop would pull everything down anyway. Nobody can give you a real number here, and if someone does, don't trust them. I don't have one either.
What worries me is the screenshot. That $272K is probably unrealized gains on open positions, not cash. I've held bags that looked amazing on screen and turned into nothing. Holding for a few hours is very risky, and one bad trade can wipe out weeks of profit. Copying a wallet doesn't give you the same result, because the price has often moved by the time you see the trade. Then there are token unlocks, hacks, legal risk for projects that scrape web data, and projects that succeed while their token gets none of the value. Small tokens can jump 300% in a week and lose most of it just as fast.
I've learned this the hard way more than once.

The part I keep coming back to

AI and crypto are getting closer, that's real. But the market seems to be moving from stories to projects people actually use, and I think that's good.
I don't know what happens next. Nobody does. If this space survives, it will be because of projects people actually use.
I'll keep watching. I'll probably keep writing about it too.
This is my personal opinion, for information only, not financial advice. Crypto is very volatile, so do your own research.