Nvidia has a new way to acquire an AI company without acquiring it. Pay for the software. Hire the people. Take a stake in whatever's left. Then send a letter to investors insisting, in writing, that none of that counts as a purchase.

Let's get the number on the table first: $6 billion. That's what Nvidia is paying the coding-AI startup Poolside to license a piece of software Poolside calls its "Model Factory" — the internal system it used to train its open-source coding model, Laguna ($6B license ([Bloomberg](https://www.bloomberg.com/news/articles/2026-08-20/nvidia-to-pay-ai-startup-poolside-a-6-billion-license-newcomer-says))). On top of that, Nvidia is putting in another $1 billion at a $12 billion pre-money valuation, and handing job offers to 109 of the people who built the thing ([The Next Web](https://thenextweb.com/news/nvidia-poolside-6bn-model-factory-licence)).

Add it up and Nvidia is wiring roughly $7 billion to a startup that was worth $3 billion less than two years ago — for a license it doesn't get exclusively, and staff it swears it isn't buying.

This is the acqui-hire, reengineered for a regulator that's watching.

🧠 Why This Matters

When a big tech company wants a small AI lab's people and tech, the tidy move is to just buy the company. But an outright acquisition triggers antitrust review, and regulators have spent the past two years squinting hard at every deal Nvidia touches. So the industry invented a workaround: license the technology, hire the humans, invest in the shell, and let the original company keep breathing.

Nvidia has now run this play three times. It reportedly paid around $20 billion to license Groq's tech and bring over its founder, and roughly $900 million for the networking startup Enfabrica ([Yahoo Finance](https://finance.yahoo.com/technology/ai/articles/nvidia-pay-poolside-6-billion-181448803.html)). Poolside is number three. Each time, the structure is nearly identical, and each time the message to Washington is the same: nothing to see here, no company changed hands.

Poolside's own investor letter reportedly says it plainly — this is "not an acquisition and it is not an acquihire" ([The Next Web](https://thenextweb.com/news/nvidia-poolside-6bn-model-factory-licence)). Whether anyone believes that is the whole story.

📊 Deep Dive

Poolside was founded in 2023 by Jason Warner, GitHub's former chief technology officer, and Eiso Kant. Its pitch was to build the most capable AI for software development, aimed squarely at enterprise and government coding work, with offices in the US and Paris. In October 2024 it raised a $500 million Series B at a $3 billion valuation, led by Bain Capital Ventures with DST Global and eBay — and, notably, Nvidia was already on the cap table ([TechCrunch](https://techcrunch.com/2024/10/02/ai-coding-startup-poolside-raises-500m-from-ebay-nvidia-and-others/)).

Twenty-two months later, that $3 billion valuation is $12 billion — a 4x jump — and Nvidia is paying more for a license than the entire company was worth in 2024. The reason Poolside gives is brutally practical: to keep competing in open-source models, it would have needed more Nvidia chips than it could get its hands on ([PYMNTS](https://www.pymnts.com/news/artificial-intelligence/2026/nvidia-pays-6-billion-to-license-poolside-ai-model-development-software/)). When your bottleneck is the supplier, sometimes the supplier just buys the bottleneck.

Here's how the three Nvidia "license-and-hire" deals stack up:

  • Poolside (2026): ~$6B license + $1B investment at a $12B valuation; 109 staff hired; non-exclusive license; three co-founders stay.
  • Groq (2026): ~$20B to license the tech and bring over the founder — the largest of the three.
  • Enfabrica (2026): ~$900M for the networking startup's technology and team — the smallest.

The kicker is what Poolside plans to do with the cash: distribute the $6 billion to its investors by the end of next year ([The Next Web](https://thenextweb.com/news/nvidia-poolside-6bn-model-factory-licence)). A license fee that functions like an exit — without an exit ever formally happening.

"Less than 70 people built this model. Less than 115 between engineering and researchers, like, together did this effort."
— Eiso Kant, Poolside co-founder

⚠️ The Catch

The catch is right there in the choreography. Lawmakers have already criticized this exact deal structure as a way to sidestep merger reviews — get the assets and the talent of an acquisition while avoiding the scrutiny of one ([Yahoo Finance](https://finance.yahoo.com/technology/ai/articles/nvidia-pay-poolside-6-billion-181448803.html)). Call something "a non-exclusive license" often enough and regulators may eventually decide the label doesn't match the substance.

There's also a talent question hiding in the numbers. If Nvidia hires the 109 people who built Laguna, and Kant himself says fewer than 115 built it, then Poolside is licensing away its factory and waving goodbye to nearly everyone who knew how to run it. The company keeps its independence on paper. What it keeps in practice is thinner.

And Wall Street noticed the price tag. Nvidia's stock ended that week down about 5% as investors added up how much cash it's spraying across the AI ecosystem ([Yahoo Finance](https://finance.yahoo.com/technology/ai/articles/nvidia-reportedly-strikes-7b-licensing-205155931.html)).

🎯 What Happens Next

Poolside says it will keep operating independently — it still runs Poolside Infrastructure, which is building a 1.2-gigawatt data center in Texas, and it can license the Model Factory to other customers since Nvidia's deal isn't exclusive ([The Next Web](https://thenextweb.com/news/nvidia-poolside-6bn-model-factory-licence)). On paper, it's a well-funded company with a shiny valuation and a fat check to pass to its backers.

The bigger thing to watch is Washington. Three of these deals in one year is a pattern, and patterns invite rules. If regulators decide that "license plus hire plus invest" is just an acquisition wearing a trench coat, the entire mechanism the AI industry has been using to consolidate quietly could get a lot more expensive.

🧩 Bigger Picture

Zoom out and the Poolside deal is a snapshot of who actually holds power in AI right now. It isn't the model labs, however clever their software — it's the company that makes the chips they all need. When your growth ceiling is set by how many GPUs your supplier will allocate you, the supplier can rewrite the terms of your existence with a single wire transfer.

Nvidia gets the coding tech, the engineers, and a slice of the upside, all without a name change or a merger filing. Poolside's investors get liquidity. And the rest of the industry gets a template: if you can't buy the company, you can buy everything the company is.

Six billion dollars for software, a billion for a stake, and 109 new hires — and everyone agrees, on the record, that nothing was purchased. In AI's arms race, the most valuable thing Nvidia licensed this week might just be the word "no."

Sources