The Math That Doesn't Work

We've talked about what bubbles are and what happens when they pop.

Now let's look at the actual numbers in AI. Because the math is worse than most people realize.

The Gap Is Massive

$560 billion invested in AI. $35 billion in revenue.

That's a 16:1 ratio of spending to revenue. No business survives that long term.

Tech companies are spending roughly $400 billion per year just to train and operate AI models. By 2030, the industry will need $2 trillion in annual revenue just to cover computing costs.

The projected actual revenue? Around $200 billion.

That's a $1.8 trillion gap. Just to break even.

The Money Is Circular

Here's where it gets weird.

You see a headline: "NVIDIA invests $100 billion in OpenAI."

Sounds massive. Like NVIDIA is really bullish on AI.

But where does OpenAI spend most of its money? Buying NVIDIA chips.

So NVIDIA gives money to OpenAI. OpenAI turns around and gives that same money right back to NVIDIA for GPUs.

Both companies announce huge numbers. NVIDIA reports record sales. OpenAI reports record investment.

But the money just went in a circle.

It Gets Even More Complex

OpenAI struck a deal with AMD. They'll purchase tens of billions in chips. In exchange, OpenAI becomes one of AMD's largest shareholders.

OpenAI buys AMD chips. AMD's stock goes up. OpenAI's shares in AMD increase in value. That value offsets the chip costs.

Money goes around. Everyone's numbers look great. But is anyone actually profitable?

This is the "if we fail, you fail" strategy.

Strategic Interdependence

OpenAI is deliberately tying itself to Microsoft, NVIDIA, AMD, and Oracle.

Not just as customers. As partners whose survival depends on OpenAI's success.

If OpenAI struggles, these deep-pocketed companies have massive incentive to prop them up. They can't afford to let OpenAI fail. Too much of their own value is tied to it.

Make yourself too interconnected to fail. It's a built-in survival strategy.

Smart? Maybe. Sustainable? That's the question.

Meanwhile, In The Real World

A recent MIT study found that 95% of AI pilot projects fail to yield meaningful results.

That's with $40 billion in corporate investment trying to integrate AI into actual business workflows.

Companies are trying to figure out where AI fits. Most can't make it work versus the cost. And that's at today's artificially low prices.

The Pricing Problem

Think about what you pay now.

ChatGPT Plus is $20 a month. Claude Pro is $20 a month.

To close the investment gap, those prices would need to be 15-20x higher. Around $300 to $400 a month.

That's not sustainable for most users.

Every query you make costs real money. They're subsidizing your usage, hoping to outlast the competition.

Sound Familiar?

It's the streaming playbook. It's the dot-com playbook.

Burn cash. Outlast competitors. Hope you're still standing when the bill comes due.

The difference is AI has way higher operational costs than streaming or websites ever did.

OpenAI's Own Projections

OpenAI is projecting losses until 2029.

Five more years of burning money. Hoping for a breakthrough that justifies all of it.

The headlines look better than reality. Circular money or not, the math still doesn't work.

You can't spend $16 for every $1 you make. Not forever.

What This Means For Us

I use these tools daily. They're incredibly valuable. They save me hours on client projects.

But I'm building on platforms that are fundamentally unprofitable.

At some point, that catches up. Either prices go way up, or companies go under, or both.

If you're building business systems that depend on AI APIs, you need to understand this isn't sustainable at current pricing.

Plan accordingly.

Next up in Part 4: The AGI gamble. OpenAI is betting everything on achieving general intelligence. And it's probably not coming.