Jordi Visser Weekly Market Report: The End of Human Market Structure Is the Real AI Trade
Source anchor: Jordi Visser, “Bubbles, Parabolas and Speed Crashes: The End of Human Market Structure,” latest video on the Jordi Visser Labs YouTube channel, published May 10, 2026.
Main takeoff points
- The argument is not that every AI-linked chart should be chased. The argument is that calling every parabola a “bubble” is lazy when the buyer base, speed of information, and structure of market participation have changed.
- Jordi’s central point this week is market-structure regime shift: human investors still anchor to old crash memories, while agents, passive flows, retail risk-takers, and benchmark constraints increasingly move capital without the same emotional baggage.
- AI demand is still outrunning normal planning assumptions. The next bottlenecks are less about whether the theme is “real” and more about compute, power, electrical equipment, memory, cooling, copper, grid capacity, and the physical stack required to run the agent economy.
- The risk is not no volatility. It is speed crashes inside a durable structural move. Semis, memory, and high-momentum winners can correct hard even if the larger AI capex regime remains intact.
- Inflation is the macro constraint that can make the second half harder: oil, food, power, copper, and long-end yields matter because the AI buildout is physical, energy-hungry, and capital-intensive.
- Positioning still matters. The more professionals dismiss the move as a bubble, the more vulnerable they become to benchmark arbitrage, trapped shorts, and forced catch-up into the wrong parts of the stack.
The bubble argument is becoming too easy
Jordi Visser’s latest market note starts with a warning about language. “Bubble” has become the word people use when a price moves faster than their framework. It sounds cautious. It often feels intelligent. But in this tape it can also be a way to avoid doing the harder work: separating speculation from a genuine change in the system that allocates capital.
That distinction matters because the AI trade is no longer just a story about one chip company, one earnings cycle, or one group of retail traders. It is becoming a market-structure story. The participants are different. The speed is different. The information loop is different. The benchmark pressure is different. The emotional anchors are different.
Older investors remember the dot-com crash, 2008, the COVID collapse, the inflation shock, and every other episode where a parabola ended badly. Those memories are not useless. They are risk controls. But they are also anchors. AI agents do not carry those memories. Passive flows do not sit around reading Kindleberger. Retail risk-takers are not trying to protect a career benchmark. And many professional investors who are certain they are being disciplined are also underexposed to the parts of the economy that are actually getting the new spending.
That is the real point of this week’s report: the market is not simply replaying an old bubble. It is trying to price a new capital cycle through an old mental model.
Parabolas can be real and still be dangerous
There is a false choice embedded in most AI commentary. Either the move is “real,” in which case every leader should keep going, or it is a “bubble,” in which case the whole thing is fake. Jordi’s framework is more useful because it allows for both truths at once.
A structural move can produce parabolic charts. A parabolic chart can still have a speed crash. A company can be central to the next economy and still become crowded, over-owned, or tactically stretched. That is why Jordi’s discussion of Micron is important. He is not abandoning the AI memory thesis. He is saying that after a violent move, capital can be rotated toward other bottlenecks that have lagged the first wave.
That is a serious distinction. The question is not whether AI demand exists. The question is where the market has already pulled forward years of optimism, where the physical economy is still short supply, and where the next marginal dollar of capex has to go.
The first phase of the trade rewarded the obvious compute stack. The next phase is less clean. It moves into power, electrical equipment, grid hardware, cooling, copper, gas, transportation, capital goods, memory, optical networking, and eventually the application layer. A market trained to think in software multiples is being forced to value factories, substations, transformers, cables, chemicals, and uptime.
AI demand is not slowing to human planning speed
One of Jordi’s strongest recurring frames is time compression. The old economy moved through cycles that humans could narrate: orders, inventories, hiring, margins, rate cuts, recoveries. The AI economy is moving through capacity constraints faster than institutions planned for.
In the video, he points to the gap between what major AI companies were planning for and what actually arrived. The important signal is not the exact promotional headline around any single company. It is the repeated pattern: demand arrives faster than annual plans, compute becomes scarce, hyperscalers buy capacity from one another, and the entire stack has to be rebuilt around workloads that did not exist at this scale a few years ago.
That is why the old “is this expensive?” debate is incomplete. Some AI-adjacent names are expensive. Some are not. Some of the most important companies in the physical buildout still screen like industrials or capital-goods businesses even though their end-market exposure is being rewritten by AI infrastructure. Meanwhile, parts of the mega-cap benchmark still carry high PEG ratios because investors default to the last cycle’s winners.
This is what Jordi has been calling benchmark arbitrage. The index is not neutral. It is a historical artifact. It owns yesterday’s winners in size, and it often underweights the companies that become necessary when the next cycle is physical, electrical, and energy-intensive.
The market will not be dominated by human hesitation forever
The most provocative line in Jordi’s video is that the market will no longer be dominated by human beings anchored to valuation, emotion, and old crash memories. That does not mean humans disappear from markets. It means the marginal behavior changes.
Agents will scan data faster. Retail will continue to seek risk where the story is obvious and the reward is immediate. Passive vehicles will keep allocating mechanically. Quant and systematic capital will respond to price, volatility, momentum, and liquidity. Benchmarks will force underweight managers to explain why they missed the move. None of that guarantees a straight line. It does make the old discretionary reflex — “this chart reminds me of a bubble, therefore I will fade it” — less reliable.
The irony is that many professional investors are still trying to win the next cycle with the scars of the last one. They want the perfect recessionary entry point. They want the collapse that lets cash become heroic. But if the system keeps producing parabolas and speed crashes instead of clean resets, the opportunity set changes. Waiting for the old movie to restart can become its own form of risk.
Inflation is the constraint that can turn a clean story messy
The bullish AI-capex thesis does not remove macro risk. It changes where the macro risk lives.
If the buildout is physical, then it touches energy, metals, power equipment, transport, construction, and labor. If it touches those channels, it touches inflation expectations. Oil matters. Food matters. Copper matters. Utilities matter. Long-term yields matter. The second half of the year can get harder not because the AI story fails, but because the AI story runs into the same physical bottlenecks that make disinflation less comfortable.
That is where policy becomes trapped. If growth and employment stay firm while inflation expectations rise, the Fed has less room to give the market what it wants. Rate cuts are easier when the economy is visibly rolling over and inflation is cooling. They are much harder when the market is already rewarding a massive capex boom, commodities are firm, and households can see gasoline, food, insurance, and power costs in real time.
This is why Jordi keeps refusing the simple bear/bull labels. The AI regime can be durable, and the tape can still get dicey. The leaders can remain the right leaders, and the entry points can still become dangerous. A structural shortage can be a tailwind for revenues and a headwind for policy at the same time.
The better question is where scarcity moves next
The useful way to read this market is not to ask whether “AI” is a bubble as one giant category. The better question is: where is the next scarcity, and has the market already paid for it?
Memory had its turn. Semis had their turn. Optical and networking have had their own bursts. Power is becoming more central. Electrical equipment and grid capacity are increasingly difficult to ignore. Gas, copper, cooling, substations, transformers, and industrial suppliers are no longer side stories; they are the physical preconditions for the agent economy everyone claims to be analyzing.
That also means the investor’s job is more complicated than buying the most famous ticker. The trade is rotating through the stack. It is rewarding bottlenecks. It is punishing complacency. And it is exposing portfolios that were built for a disinflationary software-led world but now have to compete in an inflation-sensitive infrastructure-led one.
Bottom line
This week’s Jordi report is a warning against intellectual laziness on both sides. The AI bulls can be right about the regime and still wrong about timing, crowding, and tactical risk. The AI skeptics can be right that parabolas crash and still miss the deeper point that the market’s structure, participants, and physical demand curve have changed.
The strongest reading is this: we are in a regime where AI demand is compressing time, capital is being forced into physical bottlenecks, and markets are increasingly shaped by participants that do not behave like the old discretionary investor. That creates speed crashes, not clean textbook cycles. It creates inflation constraints, not frictionless productivity magic. And it creates benchmark pressure on anyone still waiting for the last cycle’s entry point.
The market may be stretched. Parts of it may be crowded. But calling the whole thing a bubble is not analysis. The real work is tracking where the scarcity moves next — and whether your portfolio is still organized around a world that is already disappearing.