Is a new Dot-Com (AI) crash possible?

EDITORIAL: The Great Money Bonfire of 2026

By the AI-Radar Chief Editor February 14, 2026

If you have looked at the stock market lately, or perhaps just glanced at the sheer amount of concrete being poured in Northern Virginia, you might be asking yourself a trillion-dollar question: Is this real?

Everyone being professionally involved in AI,Tech in general and finance spends its days sifting through earnings calls that sound more like science fiction novels and balance sheets that look like the GDP of small European nations. We are standing in February 2026, and the air is thick with the smell of burning cash—specifically, Capital Expenditure, or "Capex" for those of us who enjoy financial jargon.

The question on everyone’s mind—from the retail investor in their pajamas to the pension fund manager sweating in a suit—is simple: Are we in a bubble that is about to burst with the force of a supernova, or is this actually the industrial revolution 2.0? (or both?)

Let’s grab a scalpel and dissect this financial beast. And don’t worry, I’ll keep the math painless.

--------------------------------------------------------------------------------

Part 1: The $690 Billion Poker Game

First, let’s look at the chips on the table. The numbers for 2026 are not just big; they are historically absurd. The "Hyperscalers"—our affectionate name for the tech giants Amazon, Microsoft, Google, Meta, and Oracle—have decided that money is apparently an infinite resource.

According to the latest data, these five companies alone are projected to spend between $660 billion and $690 billion on infrastructure this year. To put that in perspective, that is roughly the GDP of Switzerland. They are spending an entire Switzerland just on computers and buildings in a single year.

GRAPHIC 1: The "Big 5" Spending Spree (2026 Projection)

I must point out the irony here. These companies used to be beloved by Wall Street because they were "asset-light." They wrote software code once and sold it a billion times. Now? They look like 19th-century steel barons. They are pouring concrete, buying copper, and hunting for nuclear power plants.

Capital intensity (the percentage of revenue spent on building stuff) has hit 45-57% for some of these firms. Imagine if a lemonade stand spent 57 cents of every dollar sold just buying new lemons and stands. That is the new reality.

Part 2: The "Hamster Wheel" of Depreciation

Here is where the "Bubble" argument gets some sharp teeth. In the old days, when you built a factory, it stood there for 40 years producing cars or widgets.

AI infrastructure is different. It rots.

Well, not literally. But an AI data center is filled with GPUs (Graphics Processing Units). These chips run hot, run hard, and become obsolete faster than a viral TikTok trend. The lifespan of these assets is roughly 3 to 4 years.

MATRIX 1: The Maintenance Capex Trap

(Source: MarketWise)

This means that for every $10 billion Microsoft or Meta spends today, they have to spend another $6-7 billion in three years just to stay in the same place. They aren't just buying assets; they are buying liabilities that eat cash.

I find this fascinatingly precarious. If AI revenue doesn't skyrocket to cover these replacement costs, these tech giants aren't building money-printing machines; they are building money incinerators.

Part 3: The Circular Economy (Or: "I'll Pay You to Pay Me")

We cannot talk about bubbles without talking about "Reflexivity" or, as I like to call it, the "Scooby-Doo Mask Reveal."

There is a growing concern about the circularity of AI revenue. Here is how it works:

1. Microsoft invests billions into OpenAI (mostly in the form of cloud credits).
2. OpenAI uses those credits to pay Microsoft for cloud computing.
3. Microsoft reports this as "Cloud Revenue Growth."

It’s a brilliant perpetual motion machine of finance. Reports indicate that the entire cohort of pure-play AI vendors (OpenAI, Anthropic, etc.) will generate less than 35 billion in combined revenue in 2026.Yet, the infrastructure spend is 690 billion.

Do you see the gap? We are spending nearly $700 billion to generate $35 billion in direct AI model revenue.

Now, to be fair, the Bull case argues that the real value isn't in selling ChatGPT subscriptions, but in embedding AI into everything else—banking, coding, pharma, and manufacturing. And the data supports this somewhat.

GRAPHIC 2: The ROI Reality Check

Part 4: The Debt Bomb

Here is the part that keeps risk managers awake at night. The Big Tech firms have run out of their own cash. They are now hitting the credit card.

In 2025, the Big Five raised 108 billion in debt. Projections suggest the tech sector needs to raise 1.5 trillion in debt over the next few years to finish this buildout.

What happens if interest rates don't come down? What if the "soft landing" becomes a "hard thud"? These companies are leveraging up to build assets (GPUs) that depreciate in 3 years. That is a risky financial profile usually reserved for airlines, not software monopolies.

Part 5: Is It 1999 All Over Again?

This is the comparison everyone loves. Are we partying like it's 1999, right before the Dot-Com crash?

Let’s look at the AI-Radar Bubble Matrix.

MATRIX 2: Dot-Com vs. AI Boom (2026 Comparison)

(Source: Elston Consulting, Allianz)

The Verdict: This is not a Dot-Com bubble. In 2000, companies had no revenue. In 2026, Microsoft, Amazon, and Google are generating massive operating cash flow. They can afford to be wrong for a while.

However, just because it isn't a "pop" doesn't mean it won't leak.

The risk isn't a total market collapse; the risk is a "Capex Air Pocket". This happens when the Hyperscalers look at their empty data centers, realize they bought too many GPUs from Nvidia, and suddenly stop buying.

If that happens, the entire supply chain—Nvidia, TSMC, SK Hynix, the memory makers—could see their stock prices correct by 20-30% overnight. It wouldn't be the end of the world, but it would be a nasty hangover.

Part 6: The "Soft Landing" and The Hardware Wars

Despite the doom-mongering, the broader economy seems to be holding up. The consensus for 2026 is a "soft landing". Inflation is sticky but manageable.

Meanwhile, the hardware war is providing a safety net. Demand for chips is still "sky high" according to Nvidia's Jensen Huang. Supply is the constraint, not demand. Samsung is rushing to ship HBM4 memory (46% faster than previous tech) just to keep up.

When you have customers begging for product they can't get, it's hard to call the top of a bubble. The bubble usually bursts when supply exceeds demand. Right now, we are still supply-constrained.

Editorial Conclusion: The "Fragile Equilibrium"

So, as your AI-Radar Chief Editor, what is my final ruling?

We are in a state of Fragile Equilibrium.

1. The spending is real, but it is dangerously reliant on debt and circular revenue models.
2. The technology works, but the revenue from using AI hasn't caught up to the cost of building AI yet.
3. The crash risk is low, but the "correction" risk is high. If Hyperscalers tap the brakes on spending in late 2026 or 2027, watch out below.

Advice for the brave: If you are investing, look for the "Pick and Shovel" plays that have moats. Power generation (nuclear/utility) is the new gold because AI needs electricity. Look at the memory makers (SK Hynix, Samsung) entering the "Memory Supercycle".

Advice for the cautious: Beware of the "Pure Play" AI wrappers that have no moat. And keep an eye on that debt. If Big Tech starts struggling to pay the interest on its $100 billion credit card bill, it's time to cash out.

I would not compare the current situation to the Dot-Com crash, however i would put a lot of control on the different fragile capex situations (OpenAI anyone?)

Welcome to 2026. The future is expensive, slightly circular, and running very, very hot.

A vaguely preoccupied AI-radar.

--------------------------------------------------------------------------------

Cited Sources:

• SK Hynix Newsroom: "2026 Market Outlook"

• Futurum: "AI Capex 2026: The $690B Infrastructure Sprint"

• Futurum: Pure-play AI vendor revenue analysis.

• TTMS: "AI Solutions for Business in 2026"

• Whalesbook: "AI's Finance Reckoning"

• OneReach.ai: Manufacturing productivity stats.

• Allianz: "From unipolar to uneven"

• Introl: "Hyperscaler CapEx Hits $600B"

• MarketWise: "Hyperscaler Investment to Surge 71%"

• Timberlake Consultants: "Is the World Heading for a Soft Landing?"

• ROIC.ai: "NVIDIA CEO Jensen Huang Declares AI Demand 'Sky High'"

• The Tech Buzz: "Samsung Ships First Commercial HBM4"

• Microsoft Industry Blogs: "Scale and grow with AI"

• Elston Consulting: "The AI Boom vs. The Dot-Com Bubble"

• Costar / MarketWise (Cross-reference on asset lifespan).

• Deloitte: "The State of AI in the Enterprise 2026"