---

PyUncut Market Breakdown: Is the AI Boom Becoming the Next Dot-Com Bubble?

Welcome back to PyUncut, where we cut through the noise of finance, investing, and technology to get to what really matters.

Today’s episode is about a voice from experience — Rajiv Jain, chairman and chief investment officer of GQG Partners. He’s a veteran stockpicker who made a fortune riding the tech wave with names like Nvidia, Microsoft, and Amazon.

But now, he’s hitting the brakes.

Jain believes that the AI gold rush unfolding today might echo — and possibly exceed in danger — the dot-com bust of the late 1990s.

Let’s unpack why someone who has profited from AI’s rise is now sounding the alarm — and what that means for your portfolio.

---

1. Déjà Vu: Echoes of 1999

When Rajiv Jain says “this is dot-com all over again,” he isn’t exaggerating.

In the late 1990s, companies like Cisco Systems, Lucent, and Global Crossing engaged in what looked like impressive growth. But much of that growth was fueled by vendor financing — where one company funded another just to buy its products.

That circular money flow created artificial revenues, inflated profits, and ultimately led to a spectacular collapse when the cash stopped circulating.

Fast forward to 2025, and Jain sees eerily similar patterns in the AI ecosystem.

He points to the circular investment structure between big AI players:

• Nvidia invests in CoreWeave,
• CoreWeave buys Nvidia’s GPUs,
• Nvidia guarantees future purchases from CoreWeave,
• OpenAI receives investment from Nvidia and then buys compute power from Oracle,
• Oracle, in turn, buys more GPUs from Nvidia.

It’s a beautiful loop — until the music stops.

According to Jain, “This is turning financing cash flows into revenue and artificially creating profits — identical to Cisco and Lucent in 1999. Just much bigger in size and scale.”

---

2. The AI Mirage: Growth Without Profits

The central issue, Jain argues, is that AI doesn’t yet scale like past technologies.

Less than 3 percent of OpenAI’s users pay for ChatGPT, and most of those who do are from price-sensitive emerging markets.

Unlike software businesses — where every new user adds near-zero incremental cost — large-language models are compute-intensive. Every query burns expensive GPU cycles.

That means high usage actually drives higher losses.

In his words:

“Every query is compute-intensive. The customers who use AI the most end up being low-margin revenue. That’s why we think their cash losses are increasing.”

AI companies love to talk about scale, but Jain says there are no moats, no network effects, and no switching costs yet.

If you’re unhappy with one AI model, you can switch to another tomorrow — Anthropic, Mistral, Meta’s Llama 3, or any open-source alternative.

That makes today’s AI models look commoditized — and commodity markets rarely sustain high profit margins.

---

3. The Trillion-Dollar Spending Spree

Jain’s biggest concern isn’t about AI’s potential. It’s about the scale and speed of the spending.

In just two years, over $1 trillion has been allocated to AI infrastructure, mostly in the form of data-center capex and GPU purchases.

But here’s the kicker: most of those chips will be obsolete within three years.

Compare that with cloud computing, which took nearly two decades to mature into a stable, high-margin business. Yet investors are assuming AI will reach that stage in months, not years.

To Jain, that kind of expectation is dangerous optimism.

He cites Nvidia’s recent “$100 billion investment letter” to OpenAI as a red flag. The announcement coincided perfectly with OpenAI’s fundraising — a clever way to boost sentiment, not necessarily backed by cash flow.

As Jain puts it, even Nvidia doesn’t have $100 billion in free cash flow — last year it was just $60 billion.

This kind of “fund-each-other” dynamic feels worryingly like the late 1990s — but with far more zeros attached.

---

4. Accounting Illusions and Free Cash Flow Games

One of Jain’s more technical warnings targets the use of stock-based compensation (SBC).

Big tech companies love to report “adjusted free cash flow,” which conveniently adds back SBC as if it’s not a real expense.

That makes margins look healthier than they truly are.

He points out that even though these firms have grown massively, SBC as a percent of revenue hasn’t declined.

If stock prices stall — as they did after 2000 — these companies will have to pay employees cash instead of stock. That means free cash flow could shrink dramatically, just as they’re ramping up AI capex.

The irony?
AI’s promise of efficiency might actually reduce the very profitability it’s supposed to enhance — at least in the short term.

---

5. The Cloud Slowdown and the Coming Price War

Cloud computing — once the crown jewel of Big Tech — is also showing cracks.

Roughly 80 percent of large U.S. enterprises now use two or more cloud providers, allowing them to shift workloads for better pricing.

Result? A brewing price war.

GPU rental prices fell 20 percent in Q3 alone. Oracle is undercutting others by 40 percent or more to gain market share. CoreWeave’s CEO even said their mission is simple: “Cut prices and gain share.”

If that sounds like the early telecom price collapse of the 2000s, you’re not wrong.

Margins are already compressing:

• AWS saw a 700-basis-point margin decline last quarter.
• Oracle’s cloud revenue grew just 12 percent, which Jain notes is “hardly blistering — and this is the best of times.”

As capacity floods the market, AI compute becomes cheaper, but investor enthusiasm depends on rising capex. If giants like Google or AWS pull back, the entire “AI-as-growth-story” could lose its oxygen.

---

6. The Myth of Infinite Demand

Google CEO Sundar Pichai famously called AI “more profound than fire or electricity.”

Jain’s reply? Not so fast.

He argues that AI’s real economic footprint today is minuscule — around $25 billion in total revenue, compared with the $300 billion-plus Internet economy in 1999.

Even the companies building and selling the infrastructure aren’t seeing sustainable returns yet.

He adds:

“AI isn’t usable for mission-critical purposes. We talk to dozens of large companies every month and almost none are planning to spend meaningful amounts on AI.”

In other words, AI may be cool — but not yet useful enough to justify trillion-dollar spending.

Yes, it’s disruptive for things like coding assistance or insurance underwriting, but remove AI tomorrow, and life goes on.

That’s not exactly the hallmark of a technology revolution — at least, not yet.

---

7. Beyond the Numbers: The Societal Cost

Jain’s warning goes beyond balance sheets.

He worries that pouring a trillion dollars into AI — much of it into short-lived chips and data centers — means starving other innovations that could drive real productivity or sustainability.

It’s a classic misallocation of capital.

Imagine if even a fraction of that money went into renewable energy, semiconductor design innovation, or biotech research. Instead, much of it is tied up in speculative arms races over who can train the next-largest model — models that may never turn a profit.

For investors, that’s not just a market risk; it’s a societal opportunity cost.

---

8. Why This Could Be Worse Than 2000

At the heart of Jain’s thesis is scale.

The dot-com bubble destroyed about $5 trillion in market value globally.

But today’s Big Tech firms are worth over $12 trillion combined, with AI-linked valuations built into nearly every index fund, ETF, and pension portfolio.

That means there’s no escape hatch.

In 2000, you could avoid Pets.com and still sleep well. In 2025, you might hold Nvidia, Microsoft, or Amazon through your S&P 500 fund and not even realize you’re exposed to the same speculative dynamics.

And because these firms are profitable and systemically important, a sharp correction could ripple through the entire market — not just the tech sector.

To Jain, that makes this setup “bigger, broader, and potentially more painful” than the dot-com unwind.

---

9. What Defensive Investors Can Do

Jain has been rotating into more defensive names, trimming exposure to mega-cap tech and focusing on companies with steady cash flows and sustainable dividends.

That doesn’t mean abandoning technology altogether — just being selective and realistic.

His playbook offers useful lessons for retail investors:

• Avoid herd mentality.
  If every CEO, analyst, and influencer is saying “AI will change everything,” that’s often your cue to question it.
• Watch the fundamentals.
  Sustainable revenue growth, cash flow discipline, and capital efficiency matter more than buzzwords.
• Diversify beyond tech.
  Industrials, healthcare, and select financials could benefit if AI hype fades and capital rotates elsewhere.
• Be wary of circularity.
  If you see companies investing in each other to inflate metrics, it’s a red flag — not a growth story.
• Think long-term utility.
  True revolutions — like the Internet, smartphones, or electrification — built clear value ecosystems. AI will get there, but maybe not at today’s valuations.

---

10. The Human Element

In every market mania, there’s a psychological loop.

During the dot-com era, it was “eyeballs over earnings.”
Today, it’s “parameters over profits.”

AI enthusiasts boast about trillion-parameter models as if size alone equals value. But Jain’s message is sobering: complexity doesn’t guarantee durability.

Technologies that last — like Google Search, the iPhone, or AWS — solved clear problems profitably.

Most AI models today are still solutions in search of a business case.

That doesn’t mean AI will fail — just that timing matters.

Investors who mistake early promise for mature profitability risk learning the oldest lesson in markets: valuation matters, even for the future.

---

11. What History Teaches Us

Let’s rewind to 1999.

Everyone “knew” the Internet would change everything — and it did. But most dot-com companies went bankrupt before the real winners emerged.

Fast-forward to today. Everyone “knows” AI will change everything — and it likely will. But between here and there lies a valley of disillusionment.

Remember: Amazon lost 90 percent of its value after 2000 before becoming the behemoth we know.

The next phase of AI will likely see the same — a brutal shake-out followed by real consolidation.

If you’re patient, disciplined, and focused on fundamentals, that’s where fortunes will be made.

---

12. PyUncut Takeaways

1. The parallels to 1999 are real.
   Circular financing, vendor dependencies, and hype-driven spending are back — just bigger.
2. AI economics don’t yet work.
   High compute costs and limited paying users make scaling difficult.
3. Free cash flow illusions matter.
   Stock-based compensation and aggressive adjustments mask real costs.
4. The cloud backbone is weakening.
   Price wars and margin compression will challenge Big Tech’s earnings power.
5. AI’s real revenue is small.
   $25 billion vs $300 billion for the Internet in 1999 — a tiny fraction of what’s priced in.
6. Diversify and stay skeptical.
   Defensive sectors could outperform if AI spending slows.
7. History rhymes.
   The Internet didn’t die after the dot-com crash — it evolved. The same may happen with AI, but not without pain first.

---

Closing Thoughts

Rajiv Jain isn’t saying “AI is useless.” He’s saying “the market is irrationally exuberant.”

The danger isn’t in believing in AI’s long-term potential; it’s in assuming the profits will come tomorrow.

When everyone’s “all in,” even good technologies can become bad investments — at least until reality catches up.

So the question every investor should ask today isn’t “How big can AI get?” but “What if it doesn’t pay off as fast as we think?”

That’s the kind of disciplined skepticism that separates the survivors from the speculators.

---

This has been PyUncut Market Breakdown.
Stay curious. Stay diversified. And remember — sometimes, the smartest move in a gold rush is to sell the shovels, not chase the gold.

---