Nvidia’s Jensen Huang Says There’s No AI Bubble Because AI Is ‘Incredibly Profitable’ and ‘He’s Happy to Pay Hundreds of Millions of Dollars’ for AI

Nvidia’s Jensen Huang Says There’s No AI Bubble Because AI Is ‘Incredibly Profitable’ and ‘He’s Happy to Pay Hundreds of Millions of Dollars’ for AI Caleb Naysmith Wed, August 5, 2026 at 3:53 PM GMT+2 4 min read NVDA Image of Jensen Huang by jamesonwu1972 via Shutterstock The most-repeated objection to the AI buildout is that nobody can show the money coming back. It's something billionaire Mark Cuban explicitly said on the Bihe Big Tech Podcast, that "they'll never get it," with 'it' referring to the trillions of dollars spent on the AI buildout. He further classified it as "they're just sh*tting away that money.

" The billionaire has exclaimed repeatedly about the trillions being spent on the AI buildout. Possibly the largest benefactor in the world of this trend, however, has a completely polar take on the subject. Nvidia (NVDA) CEO Jensen Huang was asked a version of this by Axios ' Mike Allen: "When does the party stop?

When does your finance team start to worry? " Instead of answering with a forecast, he answered with an invoice. More News from Barchart Jeff Bezos Says He's Selling $1 Billion In Amazon Stock Every Year to Fund Blue Origin — 'It's The Most Important Work I'm Doing' Apple's New CEO Is Bringing a Familiar Face Back From Retirement.

The Shift Is Happening. Nasdaq Futures Climb as Tech Rally Continues on Palantir Boost, U. S.

JOLTS Report and SpaceX Earnings on Tap Get exclusive insights with the FREE Barchart Brief newsletter. Subscribe now for quick, incisive midday market analysis you won't find anywhere else. "These coding agents are incredibly profitable," Huang said.

"They're doing useful work for very high-paying jobs. And so many companies like ourselves are happy to pay hundreds of millions of dollars a year so that we can go use these AI services to help us augment our coding capability. " Read that construction carefully, because it is doing two things at once.

Huang is describing a class of buyers—"many companies like ourselves"—and he is placing Nvidia inside it. He is not disclosing an Nvidia line item, and Nvidia does not break out AI-tooling spend in its filings. What he is doing is putting a public number on the order of magnitude that a large technology company now pays to rent coding intelligence: not millions, not billions, but hundreds of millions annually.

Nvidia Is Also a Customer That is the part worth sitting with. Nvidia sells the hardware these coding agents run on. It is the single largest beneficiary of the labs' compute purchases.

And it is also writing them checks. Huang was explicit about being a customer earlier in the same interview and unusually specific about it. "I advise all of my customers to use OpenAI and Anthropic," he said, "because it's easy, it works incredibly well, and I use it.

I use Perplexity and Anthropic and OpenAI and use these services as much as you can. " The company at the center of the AI trade is, by its CEO's own account, a paying subscriber to three of its customers. Story Continues Now the denominator, which is where numbers like this usually get abused.

Hundreds of millions of dollars a year is a large amount of money in absolute terms and a small amount for this particular company. Against an income tax expense of roughly $21. 4 billion in the fiscal year ended Jan.

25, 2026, a few hundred million dollars is on the order of one to two percent. Set against total operating expenses, it would be smaller still. This is not a company straining to afford its software.

Which is precisely why Huang raised it. His argument is not that the spend is painful; it is that it is voluntary, repeatable, and rational—a real cost that a rational buyer has decided is worth paying, rather than an experimental budget that evaporates when a CFO gets nervous. "So that flywheel has now started," he said.

"And so I think the industry has now arrived at a very critical juncture and inflection point, if you will, where we now have profitable AI because we have useful AI, and because of that, we want to build a lot more AI. " AI Spending Is Still a Loop, but the Demand, and Profits, Are Real The circularity is real, and it is fair to be uneasy about it. Nvidia sells chips to labs; the labs sell coding agents back to Nvidia and to everyone else; the revenue from those agents funds more chips.

Every participant can point to genuine demand at their own layer, and the whole loop still tightens on itself. Huang's answer to that is that the demand is not the loop—it is the software engineering getting done at the end of it, on behalf of "very high-paying jobs. " That claim is testable in a way most AI-bubble arguments are not.

What makes the figure useful to an investor is that it is a floor, disclosed voluntarily, by a party with no incentive to understate it. If the largest hardware vendor in the world considers hundreds of millions a year a sensible price for coding assistance, the addressable market for that category is not speculative. It is being paid, now, by companies that can read a P&L.

It also reframes the bubble question slightly. The interesting number in AI is no longer what the labs raise or what the data centers cost. It is what sophisticated buyers who are not required to spend anything have decided to spend anyway—and Huang has now put that number, for at least one class of buyer, on the record.

On the date of publication, Caleb Naysmith did not have (either directly or indirectly) positions in any of the securities mentioned in this article. All information and data in this article is solely for informational purposes. This article was originally published on Barchart.