Running Hot Into Scarcity Is the New Market Risk
- The market is not simply replaying an old tech bubble. It is moving through a regime shift where AI demand, infrastructure scarcity, and policy choices are colliding.
- The immediate risk is not that the AI buildout disappears. The risk is that too much capital, too much momentum, and too much forced positioning have run into bottlenecks at the same time.
- Inflation pressure matters again because scarcity in compute, power, materials, cooling, labor, and logistics can keep nominal activity hot while limiting what the Federal Reserve can safely do.
- Parabolic winners can remain structurally important and still become bad near-term risk/reward when leverage, ETF flows, and weak breadth force everyone into the same trade.
- The better framework is rolling bubbles and speed crashes: capital will keep rotating toward scarce assets, but the path will be violent because market structure is faster than human judgment.
The headline is scarcity, not euphoria
Jordi Visser’s latest market note is a warning against using the wrong mental model. The easy story says the AI trade is a bubble. The equally lazy counter-story says the buildout is so important that price no longer matters. Neither frame is good enough.
The better version is that markets are now running hot into scarcity.
That phrase matters because it connects the pieces investors usually separate. AI demand has moved from a software story into a physical buildout. Compute, chips, cooling, power, optical fiber, chemicals, data centers, and the skilled work needed to assemble them are no longer background inputs. They are the cycle. At the same time, fiscal policy and industrial policy are pushing money through a system that was not built for this speed of demand.
That does not make every AI-adjacent stock cheap. It does not make every parabolic chart safe. It means the old business-cycle map is increasingly useless. The constraint is not whether people can imagine use cases for AI. The constraint is whether the physical economy can deliver the capacity that the new demand curve is already trying to pull forward.
This is a regime shift, not a simple replay
The important part of Jordi’s framework is that he is not treating this as a normal late-cycle melt-up. He is looking at the interaction between inflation, yields, forced positioning, and a new capital-expenditure cycle.
In the prior cycle, the market rewarded asset-light software, consumer growth, and long-duration narratives. In this cycle, the key question is who owns the scarce rails: compute, power, infrastructure, materials, and the balance-sheet capacity to finance the buildout.
That is why a simple “AI bubble” label misses the point. A bubble can form inside a real regime shift. The existence of speculative excess does not disprove the underlying transition. Railroads, electricity, the internet, cloud computing, and shale all produced versions of this pattern: the capital cycle is real, the social and economic consequences are real, and the market still finds ways to overfinance the wrong parts at the wrong time.
Jordi’s argument is closer to this: AI durability is real, but durability is not the same thing as a straight line. Scarcity can create the profit pool. Scarcity can also create the inflation problem, the political problem, and the crash risk.
The Fed is boxed in by the wrong kind of heat
The market’s biggest mistake may be assuming that any wobble in growth automatically gives the Federal Reserve room to rescue risk assets. That assumption worked better when disinflation was the dominant background condition.
It is much harder when the pressure points are coming from scarcity.
If AI compute demand is pulling forward power needs, chips, cooling systems, construction capacity, and industrial inputs, then the economy can feel hot even where the consumer is not booming. That is the awkward policy mix: nominal activity can remain firm while the real economy becomes more constrained. Inflation expectations can rise even when the old consumer-cycle indicators are uneven.
That is where yields matter. If inflation pressure returns while positioning is already crowded, the Fed does not get to behave like the market’s volatility suppressor. Higher yields do not just change discount rates. They expose leverage, punish duration, and test the financing assumptions behind the buildout.
The point is not that a 1970s replay is guaranteed. It is that the policy constraint has changed. Investors positioned for automatic easing may be using a playbook built for the wrong regime.
Parabolas can be real and still need to be respected
Jordi’s tone on parabolic winners is important. He is not arguing that the leading names are fake. He is arguing that market structure can turn a good thesis into a bad entry point.
That distinction is the difference between analysis and slogan.
Gold, silver, Palantir, Micron, and other high-momentum examples show how quickly capital now concentrates around a scarcity story. Once the narrative, flows, and positioning line up, the move can become self-reinforcing. Leverage ETFs, systematic strategies, retail momentum, benchmark pressure, and underweight institutions all compress time.
But the same structure that accelerates the upside also accelerates the reversal. When breadth breaks down and everyone is forced into the same narrow set of winners, the market becomes more fragile, not less. A stock or asset can be central to the next decade and still be vulnerable to a speed crash over the next month.
That is the uncomfortable part of this cycle. The better the long-term story gets, the easier it becomes for near-term risk to be mispriced.
AI demand has become physical demand
The most important second-order effect is that AI is no longer just a model-performance story. Every improvement in capability increases the pressure on the physical stack.
Agentic AI expands demand because it changes usage patterns. More autonomous workflows mean more inference. More inference means more compute. More compute means more power, cooling, chips, fiber, memory, data-center capacity, and financing. The result is a chain reaction across sectors that the traditional tech index does not fully capture.
That is why the “software versus scarcity” split matters. If the market keeps valuing the old winners as if the old cycle is intact, while the new cycle is being built through physical bottlenecks, there is an enormous benchmark-arbitrage problem. Passive allocations may still be shaped by the last regime even as the next regime is already consuming capital.
The practical investor question becomes less “Is AI real?” and more “Where is the bottleneck, who has pricing power, and who is funding the capacity?”
Private credit and leverage are the hidden transmission lines
The buildout is not just an equity story. It is also a credit story.
When companies, hyperscalers, suppliers, infrastructure partners, and financing vehicles all rush to secure capacity, the leverage does not always sit in the obvious place. Some of it appears in private credit. Some appears in structured financing. Some appears in supply-chain commitments, backlogs, prepayments, and off-balance-sheet arrangements that look harmless until liquidity tightens.
That is where market structure and credit structure meet. If yields rise and crowded equities wobble, the issue is not only whether a few momentum names fall. The issue is whether the financing assumptions behind the capacity race get repriced.
Scarcity creates opportunity, but it also creates hoarding. Hoarding creates double orders, overcommitments, and poor visibility. Poor visibility is tolerable when money is cheap and volatility is low. It becomes dangerous when capital costs rise and investors start asking which orders are real, which backlogs are finance-driven, and which balance sheets are quietly carrying the risk.
The positioning lesson: do not confuse structural bullishness with all-in exposure
A serious market view has to hold two ideas at once.
First, AI is likely a durable regime shift. The physical upgrade is still early. Power, compute, data centers, industrial supply chains, automation, and the edge economy are not one-quarter themes.
Second, the trade can become over-owned. When everyone discovers the same scarcity assets at once, the right response is not religious conviction. It is position discipline.
That is the Jordi frame at its strongest: respect the regime, but respect the tape. Scaling out of a parabolic winner is not a betrayal of the thesis. It is an acknowledgment that the next decade’s winners can still suffer violent resets when flows, leverage, and breadth turn.
The market is probably entering a period where the winning assets rotate through rolling bubbles rather than one clean leadership cycle. Scarcity assets can lead. Then they can correct. Then another bottleneck can become the next obsession. The investor who survives that environment will be the one who separates structural direction from tactical exposure.
What to watch next
The next phase will be defined by a few questions.
Watch inflation expectations. If the market begins to believe scarcity is feeding a stickier nominal backdrop, the Fed put weakens.
Watch yields. Higher yields are the simplest way to expose crowded duration, stretched valuation, and weak financing structures.
Watch breadth. A narrow market can keep rising longer than skeptics expect, but deteriorating breadth tells you the advance is becoming more dependent on forced flows.
Watch credit. Private credit, financing vehicles, supplier commitments, and hyperscaler-adjacent capital structures are where the second-order stress may appear before the headline narrative changes.
Watch the bottlenecks. The most valuable information may come from power, cooling, chips, memory, fiber, construction, and data-center capacity rather than from the most obvious AI software names.
Bottom line
The market is not choosing between “AI is real” and “AI is a bubble.” Both can be true in different time frames.
AI can be the defining capital cycle of the decade, and the current trade can still be vulnerable because the system has run too far, too fast, into scarcity. The regime shift is durable. The path is unstable. That is the market Jordi is describing: not the end of the AI story, but the end of pretending the old market structure can safely absorb it.
The next winners will not simply be the companies with the best story. They will be the owners of scarce capacity, the financers who survive the repricing, and the investors disciplined enough to buy the regime without getting trapped in the parabola.
Source anchor: Jordi Visser Labs, “Running Hot Into Scarcity: Why Bottlenecks Are the Risk to the AI ‘Bubble,’” YouTube, latest channel video as of May 17, 2026.