Nvidia's $40 Billion Question: Circular Financing or the Widest Moat in Tech?

July 10, 2026

Last updated: July 24, 2026

A while back I posted a short that was basically one line: Nvidia doesn't invest fifteen billion dollars for fun. It's the best-performing short I've ever put out.

But the comments all asked the same thing. Okay Mikey — then what is Nvidia doing with all that money?

So I pulled the whole list. And once you see the pattern behind it, you can't unsee it.

The short answer

Nvidia has committed more than $40 billion in equity investments across the AI stack in 2026 alone, according to data compiled by CNBC from public filings and FactSet. It's not random. The money sorts into four clear buckets: the customers who buy its chips, the suppliers who build the data center around them, strategic comeback bets, and moonshots on what comes after the data center boom. Critics call it circular financing. I call it a moat.

The setup nobody explains properly

Nvidia isn't sitting on its cash. It's the most valuable company on the planet, it's printing money selling AI chips, and it's taking a big chunk of that money and investing it right back into the companies that buy those chips.

Think about what that means. Nvidia gives a company a few billion dollars. That company turns around and spends it on Nvidia GPUs. The money goes out the front door and comes right back in the back door.

People have a name for this: circular financing. A lot of folks online are mad about it — they say Nvidia is funding its own demand, propping up its own sales, inflating a bubble.

I'll give you my honest take at the end. But before you judge it, you've got to see what's actually in the portfolio.

Here's the analogy I keep coming back to. In a gold rush, the smart money isn't the miners — it's the guy selling shovels. Nvidia has been the shovel guy this whole time. But now the shovel guy is also funding the miners, buying the railroad out to the mine, and paying to widen the river so more people can get there.

Everywhere you look in this gold rush, the equipment runs on Nvidia.

The four buckets

Bucket	What Nvidia is buying	Key names
1 — Customers	Demand for its own chips	OpenAI, Anthropic, xAI, CoreWeave, Nebius
2 — Suppliers	Removing its own bottlenecks	Coherent, Lumentum, Marvell, Corning, Ayar Labs
3 — Strategic	Ecosystem control + returns	Intel, Nokia, Synopsys, SiFive
4 — Moonshots	The next wave after data centers	Figure AI, Wayve

Bucket one: the customers

This is the loop everybody argues about.

OpenAI is the big one — and the most important lesson in this entire piece. See the update section below.

Anthropic, the company behind the Claude AI assistant, took a commitment of up to $10 billion from Nvidia. Nvidia participated in Anthropic's Series G, a $30 billion round that pushed the company's valuation to $380 billion.

xAI, Elon Musk's AI company, got up to $2 billion — and the structure on this one is wild. Nvidia didn't just hand over cash. A separate vehicle buys the Nvidia chips and then leases them to xAI. So Nvidia's money is literally financing the purchase of Nvidia's own hardware. xAI completed its merger with SpaceX in February 2026.

CoreWeave may be the most telling. These guys rent out Nvidia GPUs to other companies. Nvidia keeps adding to its stake — its latest 13F showed 47.2 million shares worth roughly $3.66 billion, or about 11% of the company. Nvidia also put $2 billion into CoreWeave in a data center buildout deal in January 2026 and another $2 billion into Nebius Group.

Three of the four most advanced AI labs in the world now have Nvidia on their cap table. As Jensen Huang put it in an April 2026 podcast appearance: Nvidia doesn't pick winners — it tries to support all of them.

Bucket two: the suppliers (the bucket everyone skips)

This might be the smartest money Nvidia is spending, and almost nobody covers it.

Here's the problem Nvidia is solving. Inside a massive AI data center you have hundreds of thousands of chips that all have to talk to each other constantly, at insane speeds. Right now most of that data moves over copper wire — electricity through metal. Copper eats a ton of power and it's hitting a physical wall.

So the industry is racing to switch from copper to light. Instead of pushing electricity through metal, you send data as actual light through fiber. It's called photonics. Way faster, way more efficient.

Think of it this way: copper is a packed two-lane road at rush hour. Photonics is a wide-open highway. Same cars, no traffic jam.

Nvidia went on a shopping spree to lock up that highway — roughly $2 billion each into Coherent, Lumentum, and Marvell, a deal with Corning (yes, the Gorilla Glass people, who actually make the fiber), and backing for a startup called Ayar Labs.

Here's why this bucket matters: none of it is circular. Nvidia isn't funding customers to buy chips here. It's funding the supply chain to remove the one thing that could slow its own growth. That's not propping up demand — that's clearing the runway. Big difference.

Bucket three: the strategic plays

Intel is where Nvidia made the move that should quiet anyone calling these investments charity.

Nvidia put $5 billion into Intel — its old rival, a company that had been struggling for years — buying 214,776,632 shares at $23.28 each through a private placement, roughly a 4% stake.

It printed. By May 2026, CNBC reported that $5 billion bet was worth over $25 billion — a historic return in a matter of months.

So when people tell you this is all funny money and bubble nonsense, the Intel bet is the receipt that says otherwise. Nvidia can pick winners when it wants to.

Nokia — yes, the old phone company, now a serious data center networking supplier — got roughly $1 billion, aimed at building next-generation AI-powered mobile networks. Synopsys, the chip-design software firm, got about $2 billion. And SiFive is the sleeper: a chip design company built on RISC-V, a completely different architecture from the one powering Nvidia's empire today. Nvidia backed SiFive's $400 million Series G at a $3.65 billion valuation in April 2026.

That last one is Nvidia buying an insurance policy — hedging against its own dependence on one type of chip.

Bucket four: the moonshots

Figure AI builds humanoid robots. Actual walking, working robots. Wayve does self-driving car technology.

This is Nvidia buying lottery tickets on the next wave. If physical AI is the gold rush after data centers, Nvidia wants to already be the shovel guy there too.

What's changed since I filmed this

I'm updating this rather than quietly editing it, because the update is the lesson.

In the video I said the widely reported $100 billion OpenAI investment was a letter of intent — an "up to" number, with Nvidia's own CFO acknowledging in late 2025 that no definitive agreement had been signed. My point was that the headline number and the money actually committed are two very different things.

That turned out to be exactly right. The deal never got off the ground in its original form; OpenAI pivoted away from developing its own data centers and leaned on partners like Oracle and Microsoft instead. The final Nvidia investment in OpenAI landed at $30 billion, committed in late February 2026 — the largest equity stake a semiconductor company has ever taken in an AI lab, and still less than a third of the headline.

So: $30 billion, not $100 billion. Always read past the headline. That lesson goes way beyond Nvidia.

Genius or bubble? My honest take

I'll be straight with you. Yes, a lot of this is circular. Nvidia invests in these companies, hands them cash, and they turn right back around and pump it into Nvidia chips. Money out the front door, back in the back door. And the critics are right that some headline commitments shrink dramatically between announcement and signature.

But here's where I land.

That loop is exactly what's keeping the AI trade alive right now. This entire build-out — the most expensive infrastructure project in human history — is running on that cycle.

And honestly? I don't blame Nvidia one bit. If you're the most powerful company in the room and you can use your own cash to lock up your customers, fix your supply chain, and own a piece of the next decade — you do it. Every time.

That's not a scam. That's a moat.

And the moat is the whole point. When one company wraps itself around its customers, its suppliers, and the next three waves of technology all at once, it stops being just another stock. It becomes the anchor everything else gets built around.

How I actually own Nvidia (and how you probably do too)

Full transparency, because that's the point of this channel: I own a ton of Nvidia. I just don't own it as a single stock pick.

I own it as the anchor.

Nvidia is the top holding in SCHG, my main growth fund
It's near the top of VOO, my S&P 500 fund
And it's riding inside QQQI, my income fund

So I'm not avoiding Nvidia. I'm holding it through funds that are literally built with Nvidia anchored at the top. And because it's already sitting at the top of all three, I don't need to stack more of it onto one single ticker to bet on this company.

The moat we just spent this whole article breaking down? I already own it, through those funds. And if you hold any broad growth or S&P 500 fund, there's a good chance you do too — whether you realized it or not.

That's the real reason I organize everything in three buckets instead of chasing individual names. Know what you already own before you buy more of it.

Frequently asked questions

What is circular financing in AI?

Circular financing describes an arrangement where a supplier invests capital in its own customers, who then use that capital to buy the supplier's products. Critics argue it can inflate apparent demand; defenders argue it's standard ecosystem investment when the positions are minority stakes in companies with broad customer bases.

How much has Nvidia invested in AI companies?

More than $40 billion in equity commitments in 2026 alone, per CNBC's analysis of public filings and FactSet data, spread across roughly 15 companies covering AI labs, cloud providers, photonics suppliers, and robotics.

Did Nvidia really invest $100 billion in OpenAI?

No. The $100 billion figure was an announced framework, not a signed commitment. The final investment came in at $30 billion, committed in late February 2026.

Is Nvidia's investment strategy a sign of an AI bubble?

Reasonable people disagree. The circular-financing critique has real substance for the customer-side deals. The supplier and strategic investments — photonics, Intel, Synopsys — have independent commercial logic and, in Intel's case, produced a large paper gain.

Do I already own Nvidia?

If you hold a broad U.S. growth fund, an S&P 500 fund, or a Nasdaq-100 income fund, almost certainly yes. Check your top ten holdings before adding a standalone position.