“When I see a bubble forming, I rush in to buy, adding fuel to the fire.”

— George Soros

A new crypto bubble inflates: AI agents

The first thing to know about the current craze for crypto AI agents is that they're not agents.

By Google’s definition, an AI agent is “an application that attempts to achieve a goal by observing the world and acting upon it.” 

With a couple of possible exceptions, crypto AI agents do not meet that standard. 

Instead, they’re more like “wrapped LLMs,” “chatbots with memecoins attached,” or just plain “slop,” as they’ve justifiably been described by critics — there’s nothing particularly revolutionary about an AI that posts on X (often with human curation).

But when did facts ever get in the way of a good story?

Especially when it’s a good crypto story: Bitcoin isn't really money, Ethereum isn’t really a world computer, most of DeFi isn't really decentralized and few DAOs are either decentralized, autonomous or even organized.

But the opportunity cost of being persnickety about these things has been high — historically, the way to make money in crypto has been to suspend disbelief and buy into the current narrative (or “meta,” as the kids say).

This is nicely summed up in the crypto catchphrase: “Believe in something.”

In 2020, for example, people who believed that DeFi would disintermediate the entire financial sector made out like bandits in the tokens of early DeFi protocols.

They were wrong, of course — DeFi has turned out to be a niche activity and no threat to TradFi at all (so far). 

But if they were early enough buying in and then prescient enough to sell out at any point during DeFi’s run from 0% to 5% of crypto’s total market cap, they were also rich.

Today, AI agents are near 0% of crypto’s market cap (just $15 billion currently) and if crypto history is any guide, it’s easy to see why they might get to 5%.

Or bigger, even: In 2021 we could tell a story about DeFi replacing the financial system, but in 2025 we can tell a story about AI replacing pretty much every system.

This is why the crypto investor Ejaaz Ahamadeen is taking a page out of the Soros playbook by rushing in to buy the crypto AI agent bubble: “Over the long term, this sector is going to [hundreds] of billions of dollars,” he told Bankless. “And secretly I think it’s going way higher than that.”

Ejaaz concedes that his numbers are “boy math,” but that’s no reason to dismiss them — at this point, we’re not betting on how useful AI agents will ultimately be, but how useful people might think they’ll be.

Which is to say, we’re betting on how big the story might get.

Ejaaz thinks it will get unprecedentedly big: “The AI-agent meta is going to form potentially the biggest bubble we’ve ever seen in crypto.”

It’s not just a story, however. 

Like any good investing bubble, this one has a core element of truth to it. 

In a blog post yesterday, Sam Altman predicted that “in 2025, we may see the first AI agents ‘join the workforce’ and materially change the output of companies.”

With that in mind, seeing that the DeFi bubble got as big as it did without anyone outside of crypto having ever heard of it, how big do you think a bubble in AI agents could get if they’re everyone’s co-workers?

VanEck predicts over one million of these agents will be onchain in 2025 and some of these might make money by selling services to us — or to each other, even.

But crypto “investors” (aka, traders) think the real money will be in the picks and shovels of the AI agent ecosystem — the two largest tokens are infrastructure plays: ai16z, worth $2.4 billion, and Virtuals, worth $4.1 billion.

It’s easy to see why: If AI agents do start paying each other, they may do so through ai16z’s Eliza framework, with fees being returned to tokenholders — and Virtuals tokenholders are already benefiting from fees paid on the Virtuals launchpad, some of which are used to buy back tokens.

Traditionally minded investors might not value those earnings too highly, because it seems unlikely that either of those businesses have much of a moat — the proliferation of launchpads suggests that businesses like Virtuals might already be commoditized and the founder of Virtuals, Jeffy Yu, says frameworks like Eliza will be commoditized “very, very quickly.”

But I don't know what Solana’s business moat is either, and Solana seems to be unbothered by its proliferating competitors (so far, at least) — so it’s reasonable to think that the token of an agent platform like ai16z has a chance at becoming as big as SOL.

This is part of Ejaaz Ahamadeen’s investment case for AI agents, as well: He thinks ai16z and Virtuals “can (and will) be priced as L1s this cycle.”

We might even be able to dream bigger than that.

Ai16z founder Shaw Walters thinks that “AI agents basically replace websites in the social media age” — and that sounds to me like a bigger story than the 2020 narrative of DeFi replacing TradFi (or the 2021 bubble in proliferating L1 blockchains).

Shaw, currently the main character in the AI agent story, is a technologist who wants to build real things: “Any AI agent whose only functionality is shilling a memecoin is probably going to zero,” he told Laura Shin. “It’s going to require a lot more focus on actual capability.”

Better yet, he wants those capabilities to benefit everyone: “I’m not rich until everybody’s rich.”

By “everybody,” Shaw means everyone in the world, but he’s starting with ai16z’s tokenholders: “The profits that we make from various things will be going into the Treasury to reward tokenholders and we’re also doing a lot of things outside of that for buy pressure.”

One of those things is to start a Shark Tank-like incubator where ai16z is “investing into things that we’re then kind of blowing up with attention.”

I don’t know how to value something like that, but I can’t think of anything more likely to inflate a crypto bubble.

Should you be adding fuel to that fire?

I truly have no idea — but don’t be surprised if the fire gets a lot bigger.

— Byron Gilliam

1. Seeker and Firedancer expected to ship Solana products in 2025

2. Celebrate good times, come on

3. BTC hits milestone as possible big month begins

4. What is ETH worth?

5. [Sponsored] T3 Financial Crime Unit marks enforcement victory: USDT 100 million in criminal assets frozen across five continents

Looking Back on 2024

0xResearch analysts discuss their biggest mistakes and what they would have done differently over the year. Tune in for their commentary on airdrop feedback loops and chasing beta.

Listen to 0xResearch on Spotify, Apple Podcasts or YouTube.