Nvidia sells the chips. Increasingly, it's also deciding who gets to buy the buildings to put them in.
Two sources told CNBC that Nvidia has been introducing companies looking for its GPUs to data center operators in the Nordics with capacity available. One source said Nvidia had reached out to a data center company directly, sounding out potential customers.
Nvidia often doesn't name who it's representing when it makes these calls, the source said.
Nvidia's Own Words on Matchmaking
This isn't entirely new territory for Nvidia. CFO Colette Kress described the practice herself back in June.
"How can we help them obtain land, power, shell?" Kress said. "How do we help them in terms of standing up the compute as fast as possible for what they need to do?"
A second source described the arrangement more plainly: Nvidia is "helping make sure people who have money and demand for GPUs also have data center space."
Nvidia didn't respond to CNBC's request for comment.
The matchmaking isn't charity. Every successful introduction likely means another data center filling up with Nvidia-powered racks rather than a competitor's chips.
Helping a customer find land, power and shell space faster also means that customer starts buying GPUs sooner. Speed sells chips just as effectively as price does.
Why the Nordics, Specifically
The region has quietly become one of the most attractive spots on earth for AI infrastructure.
Power is cheap and abundant. Land is plentiful. The cold climate helps keep chips from overheating without burning through as much energy on cooling.
Real estate firm Savills ranked Oslo, Stockholm and Helsinki among the top six locations worldwide for future data center development potential.
The deals are already stacking up. Pure DC is investing 1.5 billion euros in a 110 megawatt Finland campus that could eventually scale past 550 megawatts.
Nebius is building one of Europe's largest AI facilities there. Microsoft added capacity at an Nscale site in Norway in April.
Norway's grid operator, Statnett, says 2.3 gigawatts of data center capacity is currently queued for future grid connections. That's a striking figure for a region with a fraction of the population of a single major US metro.
"For AI-driven data centre growth, the Nordics offer one of the clearest delivery propositions globally," said Rupert Duckworth, an associate director at Savills.
A Familiar Pattern, One Step Removed
Nvidia typically sells chips to equipment manufacturers, who build them into servers before selling the finished product on. That's the official supply chain.
The matchmaking role sits outside that structure entirely. It's not a sale.
It's Nvidia using its position as the industry's most important supplier to shape which data centers get filled and which GPU buyers get capacity. It does this without ever appearing on the transaction itself.
That's a subtler version of the same dynamic showing up in Nvidia's direct equity deals with companies like OpenAI and Groq. Influence over who gets matched with whom is worth almost as much as owning a stake outright.
Regulators have started paying closer attention to Nvidia's web of investments and relationships across the AI supply chain. Matchmaking that never touches a balance sheet is much harder to track, and even harder to regulate, than an equity stake with a dollar figure attached to it.
What This Means for Miami
Miami's data center developers and investors are competing for the same GPU-hungry tenants the Nordics are now attracting, just with a very different cost and climate profile.
South Florida can't match Nordic power prices or a cooling climate. What it can offer is proximity to Latin American markets and lower latency for US and regional customers, an advantage worth emphasizing directly to Nvidia and its allocation partners rather than assuming raw infrastructure numbers alone will win the pitch.
The bigger takeaway is about leverage. If Nvidia is informally steering GPU allocation toward capacity it helps identify, Miami-area developers courting AI tenants may find it useful to build a direct relationship with Nvidia's infrastructure team, not just with the hyperscalers and neoclouds shopping for space.
