C.W.K.
← EssaysMarket

Why “NVIDIA to $10T (and $20T or $30T)” Is Harder Than You Might Wish — A Simple Math Intuition Test

Why “NVIDIA to $10T (and $20T/$30T)” Is Harder Than You Might Wish — A Simple Math Intuition Test

This isn’t TA. This isn’t “who has the best model.” This isn’t politics.

It’s capitalism + arithmetic.

When people casually treat $10T NVDA as a base case—and float $20T/$30T as “not crazy”—my brain doesn’t argue narratives. It runs a simple intuition test:

If a single “lens” (one company) is priced at $10T, what must the “population” behind it (the global AI economy) look like for that price to make physical sense?

A caveat even frontier models miss:

If you ask a frontier model “what earnings justify $10T?”, many answers will implicitly double today’s net income—treating the jump from $5T to $10T as linear growth.

That’s wrong.

The current ~$5T already prices in accelerating growth (the second derivative). It’s not a snapshot of today’s earnings; it’s a bet on a steepening curve.

So $10T can’t just be “twice the earnings.” It must price in an even steeper future curve—higher growth, sustained longer, with more certainty (or an unusually resilient multiple at massive scale).

The math isn’t linear. It’s compounding assumptions.

Read on for the clearer picture.

1) The Population/Sample Friction: the $100T shadow

NVIDIA is one company—a single sample lens into the AI stack.

If that lens is priced at $10T, the implied population behind it (the AI economy that ultimately funds it) cannot be “a bit bigger.” It must be an order of magnitude bigger.

Why? Because NVDA can’t eat 100% of the ecosystem.

Even if NVDA is the dominant “pickaxe seller,” the world still needs:

Once you accept the ecosystem must exist and share profits, the $10T lens throws a long shadow: something like a $100T+ population is what it starts to imply if you want the pricing to be more than faith.

That’s the first friction point: sample size vs. implied population size.

2) The Reality Gap: global GDP is not a cheat code

Now sanity-check that “population” scale against reality.

Global GDP is roughly $115T–$120T.

A $10T valuation for one company—especially one selling hardware into the AI stack—implicitly assumes that the world can sustain an AI spending/profit regime so large that it starts to resemble a meaningful fraction of total human output.

Put differently: for NVDA to be worth $10T without breaking the physics of the market, the world would need to devote something like high single digits of total economic output to one company’s hardware/stack over time (directly or indirectly), while still leaving enough economic profit for everyone else in the chain.

This isn’t “it can’t happen.”

It’s: the implied scale is civilization-level.

3) The 2nd Derivative Trap: verticality is not sustainability

Here’s the part that gets people.

At small and mid caps, you can justify massive upside by saying “earnings will grow.”

At multi-trillion scale, you can’t just grow earnings. You have to grow them and sustain a belief in accelerating growth—because the market cap already embeds a steep future curve.

That’s what I mean by the “second derivative trap”:

To go from $5T → $10T, the market isn’t just saying “bigger.”
It’s saying “bigger faster for longer,” in a world where the Law of Large Numbers exists.

And the higher you climb, the more the market demands a curve that doesn’t merely slope upward—it threatens to become near-vertical.

In the bulls’ world, the curve never bends.
In the physical universe, every curve eventually bends.

4) The $300B Profit Problem (and why “$300B fixes it” is still too casual)

A simple valuation identity:

Market cap = multiple × profit

If NVDA is worth $10T at a “reasonable” 30× profit multiple, that implies roughly $330B in annual profit.

That’s an enormous fraction of the profit pool of major equity indices—i.e., not “big,” but structurally dominant.

And the key isn’t that $330B is arithmetically impossible. The key is what the market is implicitly demanding on top of that:

At $10T, the market usually won’t pay a rich multiple unless it believes:

So $330B isn’t a finish line.
It’s a checkpoint—followed by the expectation that the curve stays steep.

That’s why “just get to $300B profit” is not the mic drop people think it is.

5) Capitalism’s iron law: buyers eventually attack toll booths

The AI stack is not powered by passive customers. It’s powered by hyperscalers with:

If one supplier captures too much rent for too long, the response is predictable:

That’s not politics. That’s capitalism.

So the “$10T NVDA” story implicitly requires one of two extreme assumptions:

  1. hyperscalers fail to materially compress NVDA’s take-rate for a very long time, or

  2. the AI “population” (total profit pool) explodes so violently that even a shrinking take-rate still yields near-vertical profit acceleration.

Both are heavy lifts. The second starts to look like bubble arithmetic.

6) The Ouroboros: circular flows are not the same thing as receipts

The AI ecosystem increasingly resembles an Ouroboros—a snake eating its own tail:

chips fund models → models justify capex → capex funds more chips → repeat.

Circular flows can produce huge revenue numbers, huge capex numbers, and huge stories.

But a bubble doesn’t pop because people “stop believing.”
It pops because receipts fail to match the curve—because durable economic profit doesn’t scale at the speed the market cap curve demands.

The bull case loves to talk about the curve.
The bear case asks one question: where is the non-circular cash?

7) Final intuition test

If you want a one-line “math brain” test, it’s this:

A $10T NVDA requires an AI profit pool so large that the implied population starts to cast a $100T shadow, while simultaneously assuming capitalism won’t compress the toll booth and the growth curve won’t bend.

And when people casually extend that to $20T or $30T, the assumptions don’t just scale. They compound:

At some point, it stops being “optimism” and becomes “I assume the Law of Large Numbers has been repealed.”

Conclusion

I don’t chase vertical lines in a finite world.

If NVDA ever truly earns $10T, the receipts will be obvious: a broad, durable profit pool across the ecosystem, not just a curve drawn by imagination.

Until then, treating $10T/$20T/$30T as “inevitable” is not analysis.

It’s numerology with a ticker symbol.

And honestly? My take is that even sustaining the current ~$4T–$5T depends on those "$10T... $20T... $30T..." narratives staying alive—not as achievable destinations, but as load-bearing mythology.

The moment the market stops believing in the next doubling, the current price loses its floor.

There's a word for that. Starts with B.