The AI supercycle is the thesis that artificial intelligence has replaced the old business cycle of capital versus labour with a new one of compute versus energy — a buildout so large that shortages, not recessions, become the thing that slows it down. In a conversation recorded June 4, 2026, investor Jordi Visser told Raoul Pal on The Journey Man that “there is almost no way for this not to be a supercycle,” because even the bottlenecks require more spending to clear.

That single reframing — bottlenecks as an accelerant rather than a brake — is what makes the AI supercycle worth taking apart. Below is the argument stripped to its structure: what changed, where the rotations go, and why both Visser and Pal keep landing on crypto as the trade the market has ignored.

Key takeaways

  • The AI supercycle swaps the old “capital versus labour” business cycle for compute versus energy — growth now scales with chips and power, not hiring and office space.
  • Bottlenecks replace recessions. When chips or electricity run short, Visser argues the fix is more capex, so shortages slow earnings without ending the cycle.
  • Rotations are the tell. Capital moves from Nvidia to whatever bottleneck is binding — copper, batteries, power — and then to the applications layer, from GLP-1 drugs to crypto.
  • Personal AI “vaults” and a permanent memory layer are, in Pal’s framing, the coming consumer edge of the same buildout.
  • Crypto is the “third wave” — a delayed catch-up trade that Visser expects once the AI capex boom pauses to digest, likely a 3–6 month process.

What is the AI supercycle?

The AI supercycle is Visser’s name for an economy where the growth equation is no longer capital versus labour but compute versus energy. In the old model, described from his and Pal’s years at investment banks, you grew a business by borrowing money, hiring people, and opening offices. In the new one, output scales with how many chips you can make and how much power you can plug into them.

That shift matters because it changes what a slowdown looks like. Visser argues the AI cycle is now governed by “bottlenecks and shortages” — his phrase for supply and demand getting out of whack — rather than the credit-driven booms and busts of the past. He has been building what he calls an index of “intelligence per unit of energy,” a log chart that ran along Moore’s Law, then bent upward with GPUs, and has now, he says, gone exponential on a log scale. It is the same exponential-age framing we unpack in our analysis of Reed’s Law and the exponential age of crypto.

Why bottlenecks replace recessions

In the AI supercycle, a bottleneck is not a reason for the cycle to end — it is a reason for more money to be spent. That is the load-bearing idea in Visser’s argument. If you can’t make enough chips or generate enough power, you don’t get a recession; you get a demand-versus-supply mismatch that concentrates capital into solving whatever is binding.

He points to physical evidence: data centres, he says, are roughly 30% built versus what was announced, held back by delays in turbines, transformers, and grid capacity. Because no single AI lab and no single country is allowed to “win” the race outright, the buildout can’t be switched off. The result, in his words, is “the largest capex cycle I think humanity will ever see.” The uncomfortable corollary is that the bottlenecks themselves may slow the earnings of the biggest AI companies — “not because the demand is not there, because the demand is too big.”

The rotation beyond Nvidia

The AI supercycle does not mean owning Nvidia forever; it means following capital as it rotates into each new bottleneck. Both investors expect the leadership to broaden well beyond the current AI champions. Pal frames the mechanism plainly: if the universe is solving for intelligence per unit of energy, capital and attention will “clear all roadblocks to get there.”

Power is the clearest example. Visser flags battery innovation — a reported Nvidia and Siemens effort with battery maker Fluence, a focus on solid-state batteries, and heavy silver demand — as the kind of adjacent trade the supercycle creates. Pal adds the copper story: when Tesla hit a global copper shortage, Elon Musk changed the Cybertruck’s system from 12 to 24 volts and cut copper use by about 70% — his example of how engineering routes around scarcity. The other rotation is the applications layer. Visser argues GLP-1 weight-loss drugs, led by Eli Lilly, are financing the next stage of “human software” the way advertising once financed Google’s buildout.

Personal AI vaults and the permanent memory layer

One of the supercycle’s most concrete consumer ideas is the personal AI “vault” — a private database of everything you own, say, and do that an AI can query instantly. Pal describes building a “GMI brain” from 21 years of his own writing, video transcripts, and X posts in a vector database, plus a first-principles tool he calls “the lens” that reasons only through his exponential-age framework.

The bottleneck here is memory, not intelligence. Both men describe the same daily frustration: context windows fill up, chats are forgotten, and information is lost because the models compress what they can’t hold. The fix, Pal argues, is a persistent “database memory layer” — the foundation before any agent. It’s the private, self-sovereign side of AI we explore in our piece on AI privacy and self-sovereignty, where owning your own data becomes the point rather than an afterthought.

The invisible economy: tokenized data and the agentic marketplace

The largest market the AI supercycle creates, Visser argues, is invisible to humans — an “agentic economy” of AI agents trading data with other AI agents. His reasoning: to keep getting smarter, AI needs to consume ever more information, so every digitisable scientific dataset, university archive, and corporate feed will get digitised and monetised. He cites Ribbit Capital’s framing of a token as a “machine-readable packet of information,” and notes Google processed on the order of tens of trillions of tokens in a year, with the industry now counting in the quadrillions.

This is where tokenization enters. Visser says two-thirds of the world’s assets — real estate, private credit, private equity, art — are illiquid and don’t move, and that tokenization brings velocity to those dormant assets. That thesis connects the data economy to the on-chain asset economy we cover in our breakdown of Ondo Finance and tokenized Treasuries. The claim is that agent-to-agent commerce plus tokenized real-world assets, not human consumer spending, becomes the biggest marketplace on Earth.

Why crypto could be the third wave

Crypto, in Visser’s framework, is the “third wave” — a catch-up trade that arrives only after the AI infrastructure buildout pauses. His logic runs through two conditions. First, when AI earnings are this strong, capital has no reason to chase narrative-based assets like Bitcoin; there is simply too much else to buy, which is why crypto has felt stuck. Second, the market has to recognise that the “financial guardrails” — transactions, settlement, the velocity of money — are the durable layer even if we don’t yet know which AI companies survive.

The trigger, he argues, is the coming “digestion phase.” A wave of massive AI IPOs — Visser points to Google raising roughly $85 billion in public equity and an estimated $4 trillion of listings on the way, with Cerebras already down about 50% from its debut — likely marks a peak in the capex trade for now. When that trade stalls, capital rotates. As Pal puts it, if Nvidia and the AI leaders stop going up, “the probability of crypto going up goes up much faster.” Global liquidity, he notes, is expanding roughly 10% a year, so forward-loaded demand — like the initial Bitcoin ETF surge — simply takes time to catch back up.

What this means for investors

The practical read on the AI supercycle is patience plus rotation, not all-in conviction on one name. Visser is explicit that the next few months may be “sloppy and frustrating” as the market digests an “all-you-can-eat” AI capex buffet over roughly three to six months. That digestion is a consolidation, he stresses — “not the end of the market.”

The discipline, then, is to watch where capital rotates rather than to time a top. The bottleneck-to-application chain — power and batteries, then longevity and human software, then crypto — is the map both investors are using. For the macro backdrop of debasement and abundance that sits underneath all of it, our analysis of Raoul Pal’s economic singularity thesis covers the 2030 timing call in detail, and Pal’s own research is laid out at Global Macro Investor. As of July 2026, none of these rotations is confirmed — they are a framework for what to watch, not a promise of what will pay.

The bottom line

The AI supercycle is a bet that the defining constraint of this decade is physical — compute and energy — and that the constraint funds itself, because every bottleneck demands more capex to clear. Strip away the futurism and the testable claim is about rotation: leadership moves from chips to power to applications to, eventually, crypto’s financial rails.

Treat it as a lens, not a signal. The framework tells you which curves to watch and why crypto sits on the delayed side of them. It doesn’t tell you what any of it is worth today — and the honest version of the thesis, the one Visser and Pal keep returning to, admits that patience is the hardest part of the trade.

Frequently asked questions

What is the AI supercycle?

The AI supercycle is the thesis, argued by investor Jordi Visser on Raoul Pal’s Journey Man in June 2026, that AI has replaced the old capital-versus-labour business cycle with a compute-versus-energy one. Growth now scales with chips and power rather than hiring, and shortages of both — not recessions — are what slow it down. Visser calls it “the largest capex cycle” in human history.

Why do AI bottlenecks not end the cycle?

Because clearing a bottleneck requires more spending, not less. When chips or electricity run short, the fix is to build more power, more storage, and more infrastructure — so a shortage concentrates capital into solving it rather than triggering a downturn. The side effect, Visser notes, is that bottlenecks can slow the biggest AI companies’ earnings even while demand stays overwhelming.

Why do Raoul Pal and Jordi Visser think crypto is the third wave?

They argue crypto is a delayed catch-up trade that arrives once the AI capex boom pauses to digest. While AI earnings are strong, capital ignores narrative assets like Bitcoin; when the buildout stalls — signalled by a wave of large AI IPOs — capital rotates, and the crypto “financial guardrails” plus tokenized real-world assets become the next place to play. Visser frames it as an Elliott-wave “third wave.”

This article is analysis and commentary, not investment advice. Do your own research.