Crypto vs. AI Stocks: Where the Bigger Upside May Lie
Could crypto benefit more than AI stocks as agents grow? Examine Jordi Visser’s thesis, the risks behind it, and a practical way to assess the opportunity.

Could Crypto Benefit More Than AI Stocks From the AI Boom?
AI stocks have had a strong run, but the investment case is starting to look different. The companies building models and data centers still have plenty of runway, yet rising infrastructure costs and crowded positioning make it harder to expect another round of outsized returns. That leads to a tougher question: could crypto capture more of the next phase of AI adoption than the businesses selling the computing power?
Wall Street veteran Jordi Visser argues that AI agents could make crypto more useful by initiating payments and interacting with tokenized assets. His point isn't that "AI is good for crypto" in some broad, automatic way. It's that agents may need financial rails that can handle transactions programmatically - creating demand that goes beyond the usual AI infrastructure trade.
That's an idea worth pressure-testing, not a foregone conclusion. Agents could rely on traditional payment systems, crypto markets are still volatile, and a sharp correction in AI-linked equities could hit risk assets broadly. This article breaks down Visser's argument, the evidence he points to, the portfolio signals he watches, and the risks you should check before treating the thesis like an investment case.
Why does Visser think AI-stock gains may be harder to repeat?
Visser's argument starts with a distinction between an industry growing and an investment delivering strong returns. AI adoption can keep expanding while the stocks tied to it deliver more modest gains. Share prices reflect expectations, not just performance. When expectations are already high, companies often have to keep exceeding them to maintain the same momentum.
In an interview with Cointelegraph, Visser described the AI trade as crowded and said he began cutting exposure to Micron in mid-May. He pointed to strong revenue growth in AI-related businesses, including Microsoft's reported annualized AI revenue rate of $37 billion and Micron's fiscal second-quarter revenue of $23.86 billion. Those numbers suggest demand is real - but they don't guarantee the stocks can keep climbing at the same pace.
A key pressure point is the cost of building the infrastructure. Training and running AI systems takes expensive chips, data centers, power, and networking. Visser highlighted Oracle's reported fiscal 2026 capital expenditures of $55.7 billion versus $32 billion in operating cash flow, along with $43 billion in debt financing. Bigger investments can make sense if future revenue catches up, but the gap creates execution and funding risk.
Visser said the "fireworks show" in AI stocks may be over - not that AI development is finished. That matters. A company can keep growing while its valuation compresses, especially if investors start doubting margins, capital needs, or the ability to turn demand into free cash flow.
Growth is not the same as investment return
For investors, the real question isn't whether AI matters. It's whether the future cash flows implied by a stock's current price are realistic. A tech trend can be true and still lead to weak returns if the market has already priced in too much success.
Visser also said the AI portfolio he tracks had become more volatile than Bitcoin on a rolling 126-day measure, even though it was up about 40% for the year. That comparison doesn't prove crypto is safer or likely to outperform. It does challenge the simple idea that AI equities are automatically the lower-volatility way to invest in AI adoption.
A disciplined comparison should look at valuation, cash generation, funding requirements, and downside sensitivity - not just the size of the opportunity. Those checks help separate a durable business from a compelling story that the stock market may have already priced in.
How could AI agents create demand for crypto?
Visser's crypto thesis depends on what changes when AI systems move from answering questions to taking actions. An agent might be told to book a service, manage a portfolio, or purchase a digital asset. If it runs those steps without a person approving every move, it needs a reliable way to find counterparties, verify permissions, and settle transactions.
He compares AI agents to the smartphone and App Store era for crypto. The point is about access and use: smartphones put apps in front of a mass audience, while agents could make digital transactions easier to kick off. In that scenario, a user issues an instruction and software handles the steps underneath. If that scales, it could mean more activity in crypto payments and tokenized assets.
The "fit" isn't hard to imagine. Public blockchains can support programmable transfers, and digital assets can be moved or interacted with via software. Tokenized assets can also package certain financial instruments in forms agents can process. If that becomes practical at scale, demand might spread beyond the chip and cloud suppliers.
Still, adoption isn't one switch. People have to trust agents to act within limits. Developers need secure key and permission management. Payment rails have to be fast, affordable, and usable where the activity would actually happen. And businesses and regulators have to accept the transaction model. Better software alone doesn't automatically create more activity on-chain.
The key test is whether crypto is necessary
An agent can initiate payments without a blockchain. It could use a bank account, card network, or existing payment provider. Crypto becomes relevant only if it offers a real advantage - like better programmability, easier cross-border access, or settlement for assets that already live on-chain - not simply because the payment is triggered by AI.
So investors should look for observable usage, not agent-related announcements. Practical signals include repeat transaction volume, active users, stablecoin settlement, fees paid for network services, and integrations that lead to real economic activity. Even then, context matters: rising activity can reflect speculation, incentives, or automated trading - not durable demand.
The strongest version of Visser's argument is conditional: agents could expand crypto's audience and use cases if they make digital transactions easier and if crypto rails solve the problem better than existing systems. The weakest version assumes every agent will need a token or blockchain. That's not something the technology alone proves.
What does Visser's broader crypto strategy signal?
Visser said his holdings used to focus on Bitcoin, Ether, and Solana, then expanded into a broader set of tokens. He also described maintaining an equally weighted index of 42 tokens plus four publicly traded stocks, saying the portfolio has outperformed Bitcoin since June. He planned to hold the index constituents.
This is one way to express a broader market view without betting everything on one asset. It comes with a trade-off, though: equal weighting gives smaller constituents real influence, even when liquidity, network activity, and business models vary a lot. A basket can catch winners, but it can also dilute exposure to the projects that see the strongest adoption or best risk management.
Visser also pointed to Dogecoin as a measure of returning retail enthusiasm. The idea is that retail participation can add fuel to a wider rally. That can be a useful sentiment check - but it shouldn't be treated as proof of fundamentals. Meme-coin activity may signal that traders are willing to take risk; it doesn't prove crypto payments or tokenized assets are getting more useful.
Privacy is another theme he highlighted. He argued AI could increase the value of privacy and identity protection as fraud and abuse grow, and he cited Zcash and Monero as examples tied to that narrative. That's a market-and-technology story, not evidence that those assets will gain adoption or appreciation. Privacy features can also face regulatory pressure, exchange restrictions, and compliance challenges.
How to evaluate a broad token basket
If you're using a diversified basket as a model, start by defining what it's supposed to capture. Is it exposure to payment activity, smart-contract usage, privacy tools, retail sentiment, or just a broad rebound across crypto? Without that purpose, a "diversified" list can still leave you exposed to the same underlying risk - speculative demand falling everywhere at once.
A practical checklist might include:
- Liquidity: Can you enter and exit without moving the market too much?
- Usage: Do you see repeat users and transactions beyond promotions?
- Token function: Does the asset have an actual role in the network or product?
- Concentration: How much of the basket's risk comes from a small number of correlated assets?
- Regulatory exposure: Could rules or access restrictions meaningfully limit real-world use?
- Rebalancing rules: When will you review positions, and what evidence would trigger changes?
These questions don't predict returns. They do make the risks visible - and they help keep a thematic allocation from turning into an unexamined grab bag of tokens.
What could undermine the crypto thesis?
The most direct risk is a serious downturn in AI-related equities. Visser acknowledged that a major AI sell-off could pull down broader markets and test recent lows - even if it doesn't end the crypto bull market. That distinction matters, but it doesn't shield crypto prices from short-term correlation. In stressed markets, investors often sell volatile assets to cut risk, regardless of the longer-term story.
There's also a mismatch between how quickly technology improves and how quickly investment returns show up. Agents could become more capable without becoming widely trusted or economically useful. People might still want human approval for financial decisions. Companies could limit agents' ability to make payments, or existing payment networks might add the needed functionality before blockchain-based alternatives gain traction.
Crypto has its own specific obstacles too. Wallet security and key management are hard for ordinary users. Transactions can be irreversible, and an agent with too many permissions can cause losses fast. Protocols and apps can have bugs, liquidity can disappear, and token prices can be driven by leverage or sentiment rather than real usage.
Regulation adds another layer of uncertainty, especially for privacy-focused assets, tokenized financial products, and payment services. Rules vary by jurisdiction. Even if something is technically possible, it may still be limited by who can offer it, who can use it, or what checks are required to onboard users. Those constraints aren't minor details - they can determine whether a use case reaches mainstream users at all.
A scenario-based way to test the bet
Instead of asking whether crypto will outperform AI stocks in general, break the thesis into scenarios:
- Adoption grows and crypto rails matter. Agent usage expands, and on-chain payments or tokenized assets show repeat demand. This is the strongest case for crypto-linked exposure.
- AI grows, but existing rails win. Agents become useful, but banks, cards, and established payment providers handle most transactions. AI stocks may still benefit while the crypto adoption thesis weakens.
- The investment cycle corrects. Infrastructure spending slows or financing concerns rise. AI equities and crypto could both fall, even if long-term adoption continues.
- Crypto activity stays speculative. Token prices or trading volumes rise without sustained user demand. The market may rally, but the agent-driven crypto thesis hasn't been validated.
For each scenario, spell out what evidence would change your view. That turns a headline into something you can actually test. It also makes the comparison fair: crypto and AI stocks have different business models, liquidity profiles, and risk drivers, so "outperformance" depends on time horizon and which assets you're comparing.
Key takeaways
- Visser's argument is that AI agents could create new demand for crypto payments and tokenized assets, not that AI automatically guarantees crypto gains.
- AI businesses may keep growing even if stock returns get harder to sustain, since valuations, capital spending, and funding needs still matter.
- Crypto has a stronger case when it solves a specific settlement or programmability problem better than traditional payment rails.
- Agent adoption, retail enthusiasm, and token price action aren't the same thing as real economic use.
- A broad token basket can spread risk, but it doesn't remove liquidity, regulatory, or market-wide downside.
Conclusion: follow utility, not the headline
Visser's comparison is useful because it separates exposure to AI infrastructure from exposure to what AI systems might enable. Data-center spending and chip demand are one part of the story. If agents start executing transactions at scale, payment networks and tokenized assets could become another.
But the gap between possibility and investment outcomes is still wide. Crypto has to offer a meaningful advantage, people have to trust the software, and activity has to persist beyond hype. At the same time, a correction in AI stocks could weigh on crypto even if the long-term adoption story holds up.
Treat the thesis as something to test, not something to copy. Watch whether agents generate repeat transactions, whether those transactions truly need crypto rails, and whether the assets involved capture lasting value. If the evidence isn't there, AI's growth alone isn't enough to assume crypto will outperform.