So, let’s get straight to the point, because Nvidia appears to be changing forever.
Nvidia announced that it is putting up to $105 billion toward an OpenAI data center in Ohio.
Yes, you read that correctly. $105 billion.
But the most interesting part isn’t the number. It’s what this move tells us. The world’s most valuable company appears to be slowly changing. It is no longer simply selling chips, but financing the entire industry that buys them.
The project is called PORTS-Pike Technology Campus and is located in Pike County, Ohio. SB Energy, a subsidiary of SoftBank, is building and operating it, and will then lease it to OpenAI under a 20-year lease.
The size? It could reach 8 gigawatts. To put that into perspective, a single gigawatt is enough to power up to 750,000 American homes. The first 800 megawatts are expected to come online in 2028.
Nvidia is guaranteeing the first 4.25 GW and has the right to cover the remaining 3.75 GW. It is also putting $1.5 billion into SB Energy.
And pay attention to this. Initially, discussions were around a $250 billion guarantee. Then it fell below $120 billion and eventually settled at $105 billion. So even as the deal gets smaller, we are still talking about numbers we have never seen before.
OpenAI president Greg Brockman put it clearly: “Compute is becoming the new oil. The new scarce resource of the AI era.”
And this is where things get even more interesting.
“Why would a company that sells chips become a guarantor for data centers?” you might be wondering.
Because its technological lead is no longer quite as unique. AMD announced more than 100% growth in its data center business. Google has started selling TPU systems, and its cloud business grew 82%. So Nvidia is looking at its other weapon.
Money.
And we are talking about enormous numbers. Last quarter, it generated $48.5 billion in free cash flow, 18 times more than three years ago. It announced an $80 billion buyback and increased its dividend from 1 cent to 25 cents.
Jensen Huang himself explained the logic: frontier AI labs have enormous demand for compute, but they are growing faster than their balance sheets can support. They do not have the credit profile to finance tens of gigawatts on their own.
And this is exactly where Nvidia comes into play.
In fact, just a week earlier, it did something even bigger. It signed partnerships with Goldman Sachs, BlackRock, Blackstone, Apollo, Brookfield and KKR for platforms that will mobilize more than $500 billion in third-party capital.
The goal? To turn GPUs into a new investment class, similar to real estate, with Nvidia having the right to guarantee 25% of each loan.
In other words, the model has changed.
We went from: “You have $10 billion? Buy GPUs.”
To: “You want $10 billion worth of infrastructure? We’ll find the financing for you.”
And in the end, the infrastructure will be… Nvidia.
And now comes the most impressive part. Where has all this money gone?
$30 billion into OpenAI. Up to $10 billion into Anthropic. Around $5 billion into Ilya Sutskever’s Safe Superintelligence. $5 billion into Intel, which is now worth nearly $30 billion. It also has a roughly $21 billion position in SpaceX.
And it doesn’t stop with AI labs.
$2 billion into CoreWeave, $2 billion into Nebius, and up to $2.1 billion into IREN.
Across the supply chain, $2 billion into Coherent, $2 billion into Lumentum, up to $3.2 billion into Corning, $2 billion into Synopsys, $2 billion into Marvell, and $1 billion into Nokia.
In other words, more than $73 billion across deals where we know the exact amount.
And the Financial Times estimates that it has committed around $90 billion over 16 months, across more than 145 companies.
And to understand how unusual this is: these deals consumed 40% of its operating cash flow. Alphabet, traditionally the biggest investor among Big Tech companies, spends around 6%.
One entrepreneur who received money from Nvidia described it perfectly: “They have no interest in moving the pawn one square. They want to see the pawn become a queen.”
Of course, there is a big “but.”
And it is called circular financing.
What does that mean?
It means that when you finance the same people who are buying your products, it becomes very difficult to distinguish how much of the demand is genuine and how much is being created by your own money.
Huang strongly denies this. “OpenAI will pay the lease,” he wrote.
Cantor analysts agree, saying the deal points to an investment cycle that could continue for a long time.
Others point to something simple. Anthropic reported $65 billion in annualized revenue in July, seven times higher than a year earlier. OpenAI reached $40 billion.
So the demand is there.
“We may have oversupply one day. But that day is not today.”
And if the economics of AI turn out to be worse than expected, the downside will no longer simply be “fewer GPUs will be sold.”
It would simultaneously hit revenue, investments, guarantees, and the residual values of data centers.
And let’s not forget that regulators in the United States, the European Union and the United Kingdom have already requested information about all these deals.
