Building Through the AI Bubble — An AI Guru Perspective
AI Bubble
Last Tuesday, while sipping coffee across from a portfolio manager at a major fund, the tension in the room was palpable. This was no casual chat. He manages $4 billion in capital, a seasoned investor; I (AI Guru) was brimming with excitement about our AI products.
He leaned in, eyebrows furrowed, and said:
“You’re building the real stuff. But here I am, locked into 35% of my fund in AI stocks. If things slide, my LPs will fire me—yet I can’t exit easily without underperforming while this hype continues.”
Then came the question that’s been haunting me too:
“How do you stay bullish about AI while being cautious about the valuations?”
That moment crystallized the tension at the heart of today’s AI ecosystem: the technology shift is real, but we might already be in a bubble. Differentiating between the two is everything.
When Revolution Meets Expectation: The Bubble Paradox
The Coffee That Explained It All
Three weeks ago, AI Guru wrote about the $600 billion shock rippling through technology markets. Two weeks ago, the focus was how companies are choosing AI over labor. Last week, a warning about Gen Z building careers on shaky foundations. A reader emailed:
“You’re describing symptoms. What’s the disease?”
Here’s the reframed answer from AI Guru: There is no disease. AI is the most profound technological shift of a generation.
Yet—here’s the awkward truth—you can be 100% right about the future, and still lose money in the near term.
The markets are pricing in a future that is real—but might be 3–5 years away. That gap between visionary reality and practical adoption? That’s exactly where bubbles form.
I want to walk you through why I'm the most bullish I’ve ever been on AI—and at the same time the most cautious I’ve ever been about AI stock valuations.
The Dual Narrative: Bullish Signals and Alarming Red Flags
The Revolutionary Side: Why I Believe
Real revenue already flowing. OpenAI has hit $3.4 billion and is chasing $10 billion+ levels.
Every major tech company is pulling AI into their core operations. No longer just experiments.
Productivity gains are visible. Across industries, workers reclaim 5–10% of time.
Infrastructure built today lasts decades. These investments become the backbone of future innovation.
Unlike dot‑com, many AI players already have real, profitable businesses.
The Concerning Side: Why I Worry About the Bubble
Global corporate AI investment hit $252.3 billion in 2024, a 13× leap from a decade earlier.
Google, Microsoft, Amazon, Meta are committing $320 billion in 2025 alone.
OpenAI’s projected burn rate through 2029 is $115 billion.
By 2030, the projected spending on data centers alone is $5.2 trillion.
Critically: For many, the spending-to-revenue ratio is 4× to 8× today.
In short: We’re spending hundreds of billions now to generate tens of billions in revenue. That ratio feels extreme—but is it irrational?
Not necessarily. Just like the spending spree to build electricity grids in the early 1900s or internet backbones in the 1990s, many of those bets only paid off years after the speculative froth subsided.
The question isn’t whether AI will justify these investments—it almost certainly will. The question is: Can the market remain patient long enough for reality to catch up to expectations?
The “Industrial Bubble” — A New Breed of Mistake
Most bubbles are driven by pure speculation: people piling into something irrationally. But what if the error is more honest—a misjudgment of timing?
Economist Noah Smith and others have coined the term industrial bubble (borrowed from Jeff Bezos) to capture this nuance: AI doesn’t need to fail to provoke a collapse—it just needs to mildly underdeliver relative to the most optimistic forecasts.
Let’s look at history:
Railroad Boom (1870s)
✅ Transformed transportation
✅ Created huge value
❌ Yet, railroad stocks crashed dramatically
Why? Overbuilding too early; demand followed slowly
Telecom Fiber Boom (Late 1990s / Early 2000s)
✅ Fiber laid everywhere
✅ Ultimately vital for the internet
❌ But 85–95% of fiber was “dark” (unused) four years post‑crash
Why? Infrastructure outran demand
The pattern is clear:
Build massive infrastructure ahead of demand
Revenue lags infrastructure
Debt strains operations
Crash ensues
Recovery comes only when demand catches up
Noah Smith frames it thus: “AI doesn’t have to fail to cause a crash. It just has to disappoint optimists.”
Right now, the fervent beliefs include:
AI will transform every business by 2027
We’ll see 30–40% productivity gains everywhere
AI will create $10+ trillion in value by 2030
But what if reality is more gradual?
Transformation spans through 2030
Gains reach 10–15% productivity
Value generated: $3 trillion by 2030
Same long-term destination. Different timeline. That mismatch can provoke a collapse—not because AI failed, but because expectations were overly ambitious.
Smart Money Is Cautious (But Still Invested)
Recent signals worry me. The Bank of England warned of rising correction risk. The IMF sees valuations reminiscent of the dot‑com boom. Jamie Dimon has flagged a higher likelihood of a “notable decline” in 6–24 months.
Yet: no one is fleeing AI en masse.
Why? Because they recognize what came after the dot‑com crash. Amazon’s stock fell 95%, then became the dominant player. Google went public during the collapse and became Google.
Bezos once called this an “industrial bubble,” but followed with: AI is real, transformative, and inevitable.
The lesson: Bubbles can be productive. The dot‑com bubble paved the path for infrastructure we now take for granted. The AI bubble might do the same.
Smart capital isn’t abandoning AI. It’s being selective. The real question: Which companies survive the reset?
The Web of Circular Finance
Sometimes the funding itself seems circular:
Nvidia → OpenAI: Up to $100 billion investment over 10 years, in return for OpenAI buying Nvidia systems.
OpenAI → AMD: OpenAI receives warrants for up to 10% of AMD in return for future chip purchase commitments.
One perspective: This is vendor-backed financing wrapped as strategic investment. Another: It’s confidence baked in—these firms are tying their fates together.
Cautionary memory: During the dot‑com boom, Cisco invested in startups that then turned around and bought Cisco gear. When the collapse hit, Cisco’s stock plunged ~89%.
But the difference today: Many of these AI players already produce revenue. Nvidia’s customers are profitable giants. These may not be naïve bets—but they are risky entwined bets.
This doesn’t prove doom. But it does concentrate risk.
Why AI Is Different (And Why That Matters)
I don’t want to be mistaken for another cautious skeptic. The case for AI’s uniqueness is strong:
1. Revenue Today, Real Businesses
OpenAI: $3.4 billion and growing.
Microsoft’s AI verticals contribute strongly to Azure.
These aren’t experiments—they’re profit centers.
2. Mature Backers, Not Fresh Startups
Unlike the dot‑com era of hope‑and‑eyeballs, today’s AI push is led by Amazon, Google, Meta, Microsoft—companies sitting on $50–100 billion annual profits and hundreds of billions in capital.
3. Tangible Productivity Gains
At AI Guru:
Strategy workflows compressed by 60–75%
DubLabs takes weeks to hours for video localization
Across domains: 5‑10% documented time savings
This isn’t vaporware.
4. Infrastructure with Enduring Value
The internet’s backbone lay in dark fiber for years. Today’s AI infrastructure—data centers, clusters, model weights—serves the next generation of AI applications.
Still, you can believe in this future and lose money in the near term. Amazon’s fundamentals improved from 1999 to 2001—yet the stock lost 95% before recovering. Technology can triumph but valuations can collapse first.
The Productivity Discrepancy: Expectations vs. Reality
We see tension in the evidence around AI productivity:
Critics point out:
Workers report ~5.4% time savings
77% say AI increased workload
47% don’t know how to unlock its value
Harvard / Stanford note “workslop” — AI output that looks active but fails to move the needle
MIT finds 95% of companies report zero return on AI investments
If this holds, it's a textbook signal preceding industrial bubble collapse: the tech works in theory, but adoption and ROI lag expectations.
But here’s what critics miss:
Transformative technologies historically follow slow adoption curves
“Amara’s Law”: we overestimate short-term impact and underestimate long-term
AI may require 5–7 years to fully deliver
Goldman Sachs predicts meaningful GDP impacts only by 2027, 50% adoption by 2031–2032
At AI Guru:
Consultants compress strategy cycles from 3 weeks to 3 hours
Content creators publish globally overnight
Professionals draft complex comms in seconds
Yes, implementation is hard. Yes, learning curves exist. Yes, benefits are uneven. And yes, workslop is real.
But the path is clear: The market expects impact in months. Reality demands years. That gap is risk.
Concentration Risk: The Hidden Threat
The “Magnificent Seven” tech companies now account for roughly 30% of the S&P 500.
Many retirement portfolios and index funds are heavily exposed.
Translation: If AI stocks dip 50%, typical accounts could fall 15–20% overnight.
This time, nearly everyone knows the risks—but that doesn’t stop exposure.
Three Possible Futures & How You Win in Each
Based on what I see, three trajectories are plausible:
1. Soft Landing (40% probability)
Productivity gains materialize steadily
AI revenue grows 30–40% annually
Valuations compress 20–30%
No crash, just consolidation
How to Win: Hold quality players, keep building, ride through the shakeout.
2. Reset (45% probability)
One or two disappointing quarters
Realization that timelines were aggressive
AI stocks correct 40–50%, broader market 20–25%
Infrastructure overbuilt, debt strains
How to Win: Builders shine. Talent becomes available. Infrastructure assets discounted. The best firms emerge.
3. Multi‑year Stagnation (15% probability)
Sideways trading for years
Revenue catches up slowly
Stock performance moves sluggishly
How to Win: Stay product- and revenue-focused. Ignore market noise. Be ready for when the run resumes.
Here’s the key: In all three scenarios, the best strategy is the same: build real products, generate real revenue, and prepare for opportunity.
Behind Closed Doors: What Founders and Investors Confess
Conversations I have rarely make it to LinkedIn:
A Series B founder: “Burning $8M/month to make $500K. VCs don’t care—they want the exit. But this is unsustainable.”
VP at a cloud provider: “Our compute utilization is ~40%. We built for 100%. If demand doesn’t rise by 2026, we have stranded assets.”
Hedge fund manager: “We’re 30% short on infrastructure stocks. But we can't scale shorts because the long side still pays. The pain trade is up—for now.”
Calling the top early is career suicide. But being too exposed at the wrong time also is.
The Indicator Nobody’s Watching: Infrastructure / Revenue Ratio
Here’s the metric that keeps me up at night: spending on AI infrastructure relative to AI revenue.
In healthy growth markets, infrastructure might be 15–25% of revenue.
Right now? Infrastructure spending is ~800% of revenue—$8 spent for every $1 earned.
If you assume 10× revenue growth in five years, you'd still get 80¢ per dollar—a wild stretch.
This echoes the railroad overbuild and telecom overcapacity models of the past. History suggests many infrastructure builders won’t survive.
Builder’s Playbook: How AI Guru Navigates Risk and Opportunity
Strategy as a Builder
Separate technology timing from market timing. Build for 2030, not 2025.
Raise runway for 24+ months, never 12.
Hire top talent early—quality matters more than cost.
Prove ROI to real customers, not just acquire users.
Plan for multiple outcomes, not just optimistic ones.
What I’m Doubling Down On
Aggressive hiring of AI product and engineering talent
Investing heavily in R&D
Getting paying customers and proving unit economics
Building for 2027–2028 adoption, not 2025 hype
What I’m Watching and Preparing For
Capital and talent becoming available post‑correction
Infrastructure rentals and costs plummeting
Real business models differentiating themselves
Customers demanding traction, not promise
This is “greedy when others are fearful”—but only if you prepare before fear strikes.
Advice for Different Audiences
For AI Builders / Founders
Raise for 2–3 years of runway, even in boom times
Bet on revenue over user growth
Focus on unit economics that hold even if you scale slower
Always have alternatives (Plan B, Plan C)
For Professionals Using AI
Don’t chase hype; build durable skills
Use AI to amplify existing expertise, not replace core competency
Be the domain expert who uses AI, not the “AI technician”
Diversify income streams
For Tech Employees / Engineers
Go beyond prompt engineering—learn how AI connects to business
Focus on leveraging AI to produce leverage (10× productivity)
Keep 12+ months in reserves; corrections bring layoffs fast
Deep domain + AI fluency is rare and valuable
The Silver Lining: Why I’m Quietly Excited
Yes, expected returns may get shaken. But out of chaos comes opportunity.
1. Talent becomes available
Top AI engineers will no longer feel untouchable; they’ll look for meaningful work.
2. Infrastructure becomes cheap
All that GPU and datacenter capacity? Discounts incoming. The cost of entry will shrink.
3. Only real business models survive
During booms, stories raise capital. In corrections, revenue is king. That’s good.
4. Tech marches on
Model improvements, GPT-5/6/7, smarter architectures—they come regardless of stock prices.
So yes—I build, even in the face of risk. Because infrastructure laid during decline powers the next wave.
Just like Amazon in 1999, Google in 2004, and Airbnb/ Uber in 2008—all built during harsh markets—some of the iconic AI companies of 2028–2030 haven’t even been founded yet.
The Final Word: Build Boldly, Be Mindful
Let me close with what AI Guru truly believes:
Build for 2030, not for next quarter’s hype.
Raise capital cautiously—never overextend.
Build real, measurable value—help users save time or make money.
Prepare for volatility, because it’s inevitable.
Win in all outcomes—by having cash, talent, durable tech.
You can be bullish on AI and cautious on valuations. You can be right about the tech and wrong on timing. That’s not a contradiction—it’s strategy.
The market won’t pause for adoption. But the strongest builders will push through anyway.
So here’s the question I pose to you: Are you repositioning your AI exposure, or double‑downing on growth?
Either way, do it with intention. Don’t let the hype blind you. The people hurt most in bubbles are those who aren’t paying attention.
You’re paying attention. That already puts you ahead of most.
Stay resilient. Build wisely. The future is being constructed today.