Jordi Visser: This Isn't a Bubble — It's a $90 Trillion Physical Buildout

Lisa Tamati reporting on Jordi Visser's latest analysis.

For informational purposes only — not financial advice. This contains hypothetical analysis. Consult a licensed financial advisor before making investment decisions.

---

The Reframing That Changes Everything

Here's the most important insight from Jordi Visser this week:

This isn't a bubble.

Not because valuations are justified. Not because AI stocks will never correct. But because the underlying demand is structural, measurable, and denominated in infrastructure — not sentiment.

"This isn't a traditional bubble based on the Kindleberger model, but rather the beginning of a $90 trillion AI infrastructure build-out driven by unlimited demand for intelligence."

Dot-com had hype and no cash. This has cash, free cash flow, balance sheets better than the US government, and 80x annualised growth in token consumption that requires physical infrastructure to support.

The difference: You can't consume 80x more electricity without building the power plants to deliver it.

---

The Benchmark Arbitrage Forcing Buying

This is the hidden mechanism most investors miss.

70% of stock market flows are passive. If you own an S&P 500 index fund, you're automatically long AI names weighted by market cap. If AI names are rising, your passive fund's benchmark is rising — so you're forced to buy more to stay weighted.

This creates a structural bid that transcends normal valuation logic.

"If you are long the S&P 500 through passive investing, by definition, you are underweight Intel. You are underweight every single name that is going higher."

Translation: The passive mechanism is systematically buying AI names and selling everything else, regardless of fundamentals.

This explains why traditional bubble logic doesn't apply. In a bubble, you can short the overvalued name. But if passive flows are mechanically buying, shorting means fighting a structural bid that can't be stopped.

---

The Rotation: Semis to Power

Visser is rotating out of semis into power infrastructure.

Anthropic is consuming 80x more tokens than planned. Each token consumed requires power. Power requires infrastructure — grids, turbines, cooling, transmission.

"I believe we're at the part where the bottom end of the five-layer cake, the energy power thing is going to be a problem."

Goldman Sachs just released a major report on data centre power shortages. The grid isn't built for AI-scale compute demand.

This creates a clear sector rotation:

• Semis: Peaking in euphoria
• Power: Just beginning bottleneck recognition

When bottlenecks shift, the winners shift. Passive flows will eventually chase power infrastructure — utilities, turbine makers, grid operators, commodity producers — just as they're chasing semis now.

---

AI Agents Replacing Human Market Structure

Here's where it gets interesting.

"AI agents have no emotion. They have no memory of drawdown. They have one goal: make money."

As AI agents begin dominating market structure — execution, portfolio management, trading — human psychology stops mattering.

Traditional bubbles rely on human psychology: fear, greed, capitulation. Emotionless AI agents will exit positions at mathematical inflection points, not at emotional extremes.

This creates "speed crashes" — markets move faster both up and down, with shorter correction durations.

"The market will no longer be dominated by human beings who get emotion-anchored devaluation and comfort."

Risk management implication: Position sizing matters more. Volatility increases. Corrections happen fast and recover fast — because AI agents rebalance methodically, not emotionally.

---

Token Consumption: The Unmeasurable Demand

Here's the concrete evidence that this is structural, not speculative.

Anthropic planned 10x token growth for 2026. In Q1 they saw 80x.

Tokens are consumed. Each token consumed requires computation. Computation requires electricity. Electricity requires infrastructure buildout.

Unlike dot-com (where demand was speculative), token consumption is a measurable physical requirement.

"Dario Amodei this week: we plan for a world of 10 times growth per year. In the first quarter of this year we saw 80 times. Token consumption is about to go exponential."

This supports the $90 trillion number. If enterprise AI adoption is starting and will scale 80x faster than planned, the infrastructure required is immense — and built out over decades.

---

Inflation Is Coming Back

Commodity supply bottlenecks are emerging.

Power demand for AI. Fertiliser prices rising. Oil storage running low. Shipping capacity constrained by infrastructure bottlenecks.

"Inflation returning — CPI heading to 3.7%."

This cuts against the AI-as-deflationary narrative. Yes, AI reduces labour costs. But AI infrastructure requires physical commodities with constrained supply.

Net result: Stagflation risk. Software companies benefit from AI efficiency. Hardware, power, and commodity companies benefit from scarcity.

---

The Trading Positioning

• Long: Power infrastructure, utilities, turbine makers, commodities (silver, copper), renewable energy.
• Hedge: Long volatility (AI agents create speed crashes), short duration bonds (if inflation returns), defensive equities.
• Reduce: Semiconductor concentration — peak euphoria, rotation timing.
• Avoid: Traditional software and services getting rotated out of passive portfolios.

---

What This Means

Visser is saying something radical:

This isn't a bubble. It's a structural $90 trillion buildout. Valuations might compress. Corrections might happen. But the underlying demand is real, measured in tokens consumed per second, and requires physical infrastructure that takes decades to build.

The only way this is wrong is if AI adoption plateaus below current expectations. Everything in the data — 80x token growth vs 10x planned, corporate $1.3–1.4T cloud commitments, agentic proliferation — suggests that's not happening.

The structural bid from passive flows means this rides higher than traditional fundamentals would justify. The benchmark arbitrage forces buying of AI infrastructure plays regardless of price.

This is a generational buildout, not a bubble.

---

This is analysis of an investor's stated thesis and positioning, not financial advice. All trade ideas are hypothetical and educational. Valuations can compress even if the long-term thesis is correct. Verify claims from original sources. Consult a licensed financial advisor before making investment decisions.

Lisa Tamati is a professional ultra-endurance athlete, author, and host of the Pushing the Limits Podcast. She runs a longevity health practice and supplement company from Taranaki, New Zealand.