Yellow Capital CEO Diego Martin says the strongest demand forming in crypto is no longer coming from retail traders chasing the latest narrative, but from the emerging intersection of artificial intelligence and Web3, where autonomous agents may become recurring users of on-chain infrastructure.
In an interview with Yellow.com, Martin said AI-linked crypto demand looks different from prior market themes because it is beginning to form on the buyer side rather than simply through new token supply.
“This is the most real demand I’ve seen in years,” Martin said, noting that his view comes from watching order-book flow rather than only price charts.
His argument is that most crypto narratives have historically followed the same pattern: new tokens launch, liquidity rotates from one theme to another, and retail traders are pulled into short-term speculation. The AI-Web3 category, he said, appears to be developing differently because the end user may not always be human.
Machines Could Become The Next Buyers
Martin said the strongest part of the AI-Web3 thesis is that future demand could come from machines paying for services, not traders betting on price appreciation.
“The buyer might not even be a human that’s getting excited,” he said. “It’s a machine that needs its AI credits, for example, for Claude, or needs access to this kind of technology.”
That changes the market structure if it proves durable. AI agents may need to pay for compute, data, subscriptions, access rights, settlement services and other machine-to-machine functions. Those payments could create recurring usage rather than a one-time speculative bid.
Martin said agents may eventually “pay each other and hedge” using AI-linked utility tokens, credits or settlement layers. In that case, demand would grow as agent activity grows, rather than disappearing when retail attention moves to the next sector.
That is why he views AI-Web3 as one of the few themes in the current cycle with the possibility of structural demand.
“The capital is not only being sticky and not rotating too much, but it’s pooling and staying,” Martin said.
Bubble And Infrastructure Can Exist Together
Martin does not dismiss the possibility that AI-linked crypto assets are already in a speculative phase. Instead, he argues that bubbles and technological revolutions often happen at the same time.
He compared the current AI-crypto buildout to the dot-com era, when many internet companies failed even though the underlying technology became foundational.
“The bubble and the revolution are not in tension. They’re in the same event,” Martin said. “The internet didn’t die in 2001. Particular websites died.”
In his view, the next liquidity crunch will act as a filter. Projects that solve real problems and generate usage may survive. Projects that exist mainly because they can issue a token are likely to disappear.
His test for founders and investors is simple: would the project still exist if it could not print a token?
If the answer is yes, and the project has users, revenue and cash flow, Martin said it may represent infrastructure. If the answer is no, then the token is probably the product, and that makes it vulnerable when liquidity tightens.
“The vast majority of projects launching today don’t need a token at all,” he said. “They issued one because they need demand for it, not because the problem exists.”
Order Books Reveal What Price Charts Hide
Martin said one of the biggest mistakes investors make is relying too heavily on price charts without studying the market structure underneath.
Also Read: UK Carrier Group Russian Interception: How Prediction Markets Are Pricing the Arctic Standoff Some tokens may appear to be rallying, he said, while the order book shows a very different picture. In those cases, the market can look strong on the surface even though insiders or early participants are trying to exit.
“I see tokens that are ripping upward, but then in the book, nothing but sell orders,” Martin said. “Everyone’s trying to get out.”
He said this is often a sign of manufactured demand. A project or related market actor may be absorbing sell pressure to keep the chart looking healthy, even as the underlying book shows weakness.
“The chart is saying this is going to look great, but then the order book is saying this is just a ticking bomb,” he said.
That distinction matters as AI-Web3 tokens attract more attention. A rising price alone does not prove real demand. Martin said investors should look for usage, revenue, order-book depth and whether demand survives during periods of stress.
Next Liquidity Shock Will Be The Test
Martin pointed to the October 10 market selloff as an example of how quickly manufactured demand can disappear when global liquidity dries up. In his view, sudden market shocks expose the difference between infrastructure and speculation.
“When global liquidity dries up and the pipes freeze, the manufactured demand evaporates in just a single day,” he said. “Real usage is still there.”
That is why he believes the next crash will not end the AI-Web3 thesis, but will separate credible projects from token-only structures. The projects most likely to survive will be the ones solving real infrastructure problems, generating cash flow and serving users who need the product regardless of token price.
For Martin, the most important signals are not hype, branding or short-term performance. They are whether users continue to transact, whether revenue exists, and whether the order book shows genuine buying rather than controlled support.
“If you’re picking the winners off the chart, you’re going to buy the worst ones,” he said. “Watch the pipes, watch the order book instead.”
Retail Is No Longer The Main Volume Driver
Martin also said retail traders are becoming a smaller part of the crypto market’s total volume. In his view, most turnover today is generated by bots, market makers, arbitrage systems and other programmatic strategies.
“Most of the turnover is machines executing strategies without emotion,” he said. “And most of the pain is humans.”
That shift strengthens his broader argument. If crypto market activity is increasingly machine-driven, then AI agents becoming economic actors is not a distant idea. It may be the next stage of a market already dominated by automated execution.
Martin said retail investors still suffer most during liquidations because they enter positions emotionally and are often forced out by leverage systems faster than they can react. Bots do not panic, he said, but people do.
Over the next 12 to 15 months, he expects agents to become more visible as economic participants, buying services, transacting and contributing to market activity in ways that are less emotional and more utility-driven.
The implication is significant for crypto markets. If AI agents become recurring buyers of compute, access and settlement services, the next major demand cycle may not look like the retail-led rotations of previous years.
It may be quieter, more automated and more infrastructure-led.
The question, Martin said, is which projects are actually building for that future and which are simply using AI as the latest token narrative.
Read Next: Farage Under Fire As Hidden Crypto Gambler Cash From 2024 Surfaces