Wall Street Loves AI. The Math Doesn’t (Yet)

Everyone’s heard about how AI now makes up nearly 40% of the U.S. stock market. Microsoft, Nvidia, OpenAI, Oracle, AMD all pouring mind bending amounts of money into GPUs, data centers, and AI models.

But here’s the catch: the money going in and the money coming out still don’t match. That mismatch is what I call the “gap share.”

1. The Money Flood

Let’s look at what’s being spent:

• Microsoft: over $250 billion worth of Azure credits and data-center spend for OpenAI.
• Nvidia: investing tens of billions in GPU supply and data-center partnerships.
• Oracle: spending $40 billion on Nvidia chips to power OpenAI servers.
• AMD: promising up to 6 GW of GPUs for OpenAI — equal to hundreds of thousands of processors.

Together, that’s roughly $600–700 billion already committed just to keep AI running.

2. The Money Coming Back

Now the returns so far: (If, for a minute we ignore the chart below )

The AI Capital Loop: How Big Tech’s $Trillion Partnerships Feed the Boom

Company Annual AI-related revenue (2025 est.)~

OpenAI $13 B,

Nvidia~$150 B,

Microsoft~$10–15 B directly (more indirectly via Azure)

Oracle~$3–4 B,

AMD~$4–5 B,

If you total that up, about $180 B per year is being earned against $600 B+ spent.

That’s like investing in a house for $600 k that currently rents for $18 k a year — decent growth prospects, but you’re not breaking even soon.

3. The Gigahertz Race

AI models keep getting bigger, which means they need more compute power — more GPUs, more gigahertz, more energy.

OpenAI alone plans to scale from 2 GW today to 30 GW of compute power in the coming years. At about $50 B per GW, that’s a potential $1.5 trillion infrastructure bill — just for one company.

So the “gap” keeps widening: the more powerful the AI becomes, the more power (and money) it eats.

4. What Needs to Happen to Close the Gap

To make all this investment make sense, these companies would need roughly a 10–15% yearly return. That means:

• OpenAI would need to reach $50–100 B in annual revenue within a few years.
• Oracle must grow AI-cloud income from $4 B → $20 B+.
• Microsoft has to squeeze $40 B+ extra out of Azure AI.
• Nvidia needs to keep selling every chip it makes for the next decade.

If they pull it off, the AI boom sustains itself. If not, the bubble bursts — not because AI failed, but because revenue couldn’t catch up fast enough.

5. The Bottom Line

AI today is like building a fleet of Ferraris when there are only a few drivers with licenses. The cars are real, fast, and valuable — but until everyone can drive (i.e., until AI is used everywhere), the returns will lag the hype.

That gap share — the distance between today’s cost and tomorrow’s payoff — is where fortunes will be made or lost.