Bubble vs. Build-Out: Understanding the AI Moment

I’ve spent months arguing that AI isn’t a bubble—it’s infrastructure.

Then Michael Burry bet against Nvidia and the S&P 500. Smart money is calling this the biggest financial bubble since 2008.

So which is it? Bubble or build-out?

The uncomfortable answer: Both.

And understanding the difference between these two things—the financial speculation and the technological transformation—is what separates people who capture value from people who get caught in the crash.

Let me explain.

Two Different Things Are Happening

Here’s what I got wrong in my original “AI as Infrastructure” thesis:

I conflated two separate phenomena. The financial markets (stock prices, venture capital, corporate valuations) and the technological capability (what AI can actually do, what it enables).

These are not the same thing.

The Financial Bubble: Nvidia’s stock price. The circular investment flows. VC money going into AI companies who buy Nvidia chips with money Nvidia invested in them. This is speculative. This might be overheated. Michael Burry might be absolutely right that this corrects hard.

The Technological Build-Out: The actual infrastructure being created. The models. The compute capacity. The applications being built. The 48-64x productivity multipliers I measured. The fundamental transformation of how knowledge work gets done.

These two things can move in opposite directions.

In fact, history proves they do.

The Dot-Com Lesson

The best precedent is 1999-2002.

The Bubble (What Crashed): Pets.com, Webvan, eToys. Companies with “$300M valuations and zero revenue.” Stocks disconnected from reality. Speculative capital chasing “.com” in a company name.

When the crash came in 2000, the NASDAQ dropped 78%. Trillions of dollars evaporated. Pets.com went from IPO to liquidation in 268 days.

That bubble was real. And it popped.

The Build-Out (What Remained): The fiber optic cables. The server farms. The TCP/IP protocol. The fundamental infrastructure of the internet.

Here’s the key insight: The bubble funded the build-out.

All that “stupid” speculative money—the circular VC flows, the overvalued IPOs, the irrational exuberance—paid to lay thousands of miles of fiber optic cable. Cable that sat “dark” (unused) for years after the crash.

But that infrastructure became the foundation for Google (founded 1998, survived the crash), Amazon (nearly died in the crash, survived on the infrastructure it built during the bubble), and the entire modern cloud computing economy.

The financial bubble burst. The technological transformation continued.

Michael Burry was right about the bubble. Marc Andreessen was right about the internet.

Both. At the same time.

We’re In That Moment Again

Look at the pattern:

The Financial Indicators (Bubble Territory):

• Nvidia market cap: $3+ trillion
• Circular investment flows: Nvidia invests in AI companies → AI companies buy Nvidia chips → repeat
• Oracle spending $300B on data centers powered by Nvidia
• Microsoft, Google, Amazon in a capex arms race ($150B+ combined annual spending)
• Every company adding “AI” to their pitch deck

This looks speculative. The smart money (Burry) is betting it corrects.

The Technological Indicators (Build-Out Territory):

• I built a $2.5M platform in 80 hours (48-64x productivity multiplier)
• JPMorgan Chase: $2B in direct benefits from AI right now
• Developers are 10-50% more productive with AI co-pilots today
• Legal teams processing contracts in hours instead of days currently
• Researchers finding drug candidates faster measurably

This isn’t speculation. This is present-day ROI.

Both are happening. At the same time.

The Problem Isn’t the Spending. It’s the Timeline.

Here’s what triggers a correction—even when the underlying technology is transformative.

Investors expect massive returns in 2-3 years. The real transformation takes 10+ years.

That mismatch is the bubble.

An AI applied scientist put it perfectly: “The long-term returns from AI are certain. The problem is investors expecting those returns in the short term.”

This is the exact dot-com pattern:

1999: Wrong Timeline Expectations

• Investors expected immediate consumer revenue from internet companies
• Reality: The internet’s value took a decade to materialize (Google scaled slowly, Amazon nearly died, cloud computing came 10+ years later)
• Result: When short-term returns didn’t match expectations → panic → NASDAQ drops 78%

But: The infrastructure kept working. The fiber optic cables didn’t disappear. The TCP/IP protocol didn’t break. Google and Amazon survived because they were building for the 10-year transformation, not the 2-year expectation.

2025: Same Pattern, New Technology

• Investors expect AI capex ($150B+/year) to generate proportional revenue in 2-3 years
• Reality: The real value is a 10+ year infrastructure transformation
• Likely result: When short-term returns don’t match expectations → panic → correction

But: The productivity gains don’t disappear. The 48-64x multipliers remain. The models keep working. Claude Code doesn’t stop functioning because Nvidia’s stock drops.

The Economist’s View:

A staff software engineer who focuses on bubbles put it in formal terms: “Bubbles exist when valuations aren’t tied to NPV [Net Present Value] of future earnings.”

His take: Current AI valuations ARE tied to enormous future earnings. In economic terms, this isn’t a bubble—it’s rational pricing of a General-Purpose Technology.

But he’s pricing the 10-year NPV. Most investors are pricing the 2-year NPV.

That gap is where the panic happens.

Your Hedge:

Build for the 10-year transformation, not the 2-year expectation.

If you’re profitable from day one, delivering real ROI, and not dependent on continued cheap capital—you survive the timeline correction.

If you’re burning VC money expecting a 2025 IPO based on AI hype—you’re Pets.com.

Why This Split Matters

Here’s why understanding this distinction is critical:

If you only see the bubble: You sit on the sidelines. You wait for the crash. You miss the transformation. You’re the person in 2002 saying “I told you the internet was a scam” while Google is building the future.

If you only see the build-out: You ignore the financial risk. You assume straight-line growth. You get caught when the correction comes. You’re Pets.com, not Amazon.

If you see both: You can make intelligent decisions about timing, risk, and opportunity.

The question isn’t “bubble or build-out?”

The question is: How do I capture the value of the build-out while managing the risk of the bubble?

That’s what I’m trying to figure out.

Decoupling Your Bets

Here’s the framework I’m working with—for now:

What Survives a Correction:

• Real productivity gains (50x multipliers don’t disappear because Nvidia’s stock drops)
• Actual business value delivered (if you cut someone’s costs by 40%, they don’t care what the S&P does)
• Domain expertise + AI fluency (skills compound, stock prices don’t)
• Infrastructure you’ve built (the fiber optic cables stay in the ground)

What Doesn’t Survive a Correction:

• Stock valuations (Nvidia might be overpriced at $3T)
• Venture capital flows (funding winters are real)
• Hype-driven demand (“AI washing” dies in a downturn)
• Companies built on speculation rather than revenue

My strategy—and I’m still testing this—is: Bet on the first list. Hedge against the second.

Not All Spending Is the Same

This is where I think the distinction gets really important.

There are two completely different types of AI spending happening right now.

Speculative Spending (Vulnerable to Correction):

• VC-funded AI startups with no revenue, burning cash on customer acquisition
• Companies adding “AI” to pitch decks for valuation bumps
• Consumer-focused AI products expecting immediate mass adoption
• “AI strategy consultants” selling roadmaps and frameworks

This spending depends on continued cheap capital and inflated expectations. When the timeline correction comes, this dies first.

Strategic Infrastructure Spending (Survives Correction):

• Microsoft, Google, Amazon, Oracle building data centers
• Enterprise AI-assisted development (my 80-hour platform)
• Internal productivity tools delivering measurable ROI
• Brownfield modernization projects with clear cost savings

This spending isn’t for quarterly returns. It’s existential.

A co-founder put it bluntly: “Circular bets have been propping up capital markets for a very long time. So long as we have a Navy, you can keep buying.”

His point: The big players (Microsoft, Google, Amazon) aren’t playing the consumer market game. They’re playing the strategic infrastructure game. They’re building what they need to exist in 10 years.

This is “Navy spending.” It doesn’t stop in a recession. It accelerates.

Why This Matters:

When the correction comes:

• The speculative spending crashes (funding winter, startup failures, “AI washing” exposed)
• The strategic spending continues (Microsoft doesn’t stop building Azure AI because the stock market corrected)

The question I’m asking myself: Am I delivering real ROI right now? Or am I promising returns in a future that depends on the bubble staying inflated?

That’s the difference between Amazon (survived dot-com) and Pets.com (liquidated in 268 days).

---

This is Part 1 of a 3-part series. Part 2 covers tactical guidance by role. Part 3 explores real patterns I’m seeing and what I’m actually doing about it.

If you’re working through these questions yourself, I’d be interested to hear what you’re seeing. These are just the patterns I’m noticing—for now.