What a licensing deal is — buying the technology and the people, leaving the company standing
A licensing deal buys the right to use technology built by another company, and hires the people who built it, without buying the company itself. The structure typically has three parts: a licence fee for the technology, usually non-exclusive; hiring offers to the core engineers; and a minority equity investment. The target company survives as a legal entity. Because ownership does not change hands, merger review is difficult to trigger — which is exactly the objection raised against it
The three lines
- Structure — technology licence, plus hiring the engineers, plus a minority stake. The company stays
- Motive — it is not an acquisition, so it does not clear the merger-review threshold. The target can also claim independence
- Objection — two US senators called the shape a 'reverse acquihire' and asked antitrust regulators to look
Key questions
- What is a licensing deal in this sense?
- **Buying the right to use a company's technology, and hiring the people who built it, without buying the company.** It usually has three parts. ① A **technology licence** — the right to use the target's core software or IP, in exchange for a fee, typically written as **non-exclusive** so the target can still license elsewhere. ② **Hiring offers** to the engineers who built it. ③ A **minority equity investment**. The technology and the people move to the buyer; **the target keeps its name and legal existence.**
- Why not just acquire the company?
- Three reasons, of unequal weight. **First, regulation.** Buying a company means clearing merger review, and a competition authority can block or condition it — especially when the target is a competitor. A licence does not transfer ownership, so it does not cross that threshold. **Second, speed.** Merger review takes months to years; in a field that moves in weeks, that delay can make the purchased technology stale. **Third, the target's own position.** Selling requires resolving existing investors and remaining staff; a licence injects cash while letting the business continue. **The first reason is where the controversy sits** — if the substance is an acquisition and only the form differs, the objection is that review has been routed around.
- Has this actually happened?
- Nvidia has done it twice. **First with Groq**, the AI inference chip company: Nvidia took a **non-exclusive licence** to Groq's inference technology and hired founder-CEO Jonathan Ross, president Sunny Madra and other key staff, while Groq continued independently with GroqCloud. An Nvidia spokesperson framed it as: 'We haven't acquired Groq. We've taken a non-exclusive license to Groq's IP and have hired engineering talent from Groq's team.' **Second with Poolside in August 2026** — **$6 billion** for a licence to its Model Factory software, **$1 billion** in equity at a $12 billion pre-money valuation, and hiring offers to **109 engineers**, with the three co-founders staying. This brief covered that deal in 'Nvidia pays Poolside $6bn.'
In August 2026, Nvidia paid $6 billion to the AI startup Poolside.
Poolside was not sold. The company remains; so do its three co-founders.
Nvidia stated plainly that this was neither an acquisition nor an acquihire.
So what was bought?
1. Three parts
| Part | What it is | What the target keeps |
|---|---|---|
| Technology licence | Right to use the core technology, for a fee | Ownership. It can still license elsewhere |
| Hiring | Offers to the engineers who built it | The company remains; the people move |
| Equity investment | A minority stake | Control |
The essential point: ownership does not transfer.
Technology is a copyable asset. A licence conveys use, not title. The target still holds the technology and can, in principle, sell it again.
The word non-exclusive does a lot of work here. It is the basis for the claim that the target remains an independent business.
2. Three names, three different things
| Form | The company | Technology ownership | People | Review |
|---|---|---|---|---|
| Acquisition | absorbed or made a subsidiary | transfers to buyer | transfer with it | merger review |
| Acquihire | usually wound down | absorbed or abandoned | the people are the point | depends on size |
| Licensing deal | survives | does not transfer | move via hiring offers | hard to trigger |
That last column is the reason the structure exists.
3. Why not just buy the company
① Regulation — the big one
An acquisition must clear merger review. A competition authority can block it, or attach conditions.
This bites hardest when the target is a competitor. A company that sells chips buying a company that makes chips — or one that builds models — draws scrutiny.
A licence does not change who owns the company, so it does not cross that threshold.
② Speed
Merger review runs from months to years. In AI, weeks reshape the field. Technology bought at the start of a review can be stale by the end of it.
③ The target's own position
Selling a company means resolving existing investors and remaining employees. A licence injects cash and lets the business carry on.
In Poolside's case, alongside the licence fee, the remaining company raised a further $1 billion.
4. Two real cases, one buyer
Groq
An AI inference chip company.
| Item | Detail |
|---|---|
| Licence | Non-exclusive licence to its inference technology |
| People | Founder-CEO Jonathan Ross, president Sunny Madra and other key staff moved to Nvidia |
| Company | Survived. Continued with GroqCloud |
The Nvidia spokesperson's description is the tightest available summary of the structure:
"We haven't acquired Groq. We've taken a non-exclusive license to Groq's IP and have hired engineering talent from Groq's team."
Poolside
August 2026.
| Item | Detail |
|---|---|
| Licence | Model Factory, its model-development software — $6 billion |
| Investment | $1 billion equity at a $12 billion pre-money valuation |
| People | Hiring offers to 109 engineers behind the open-source model Laguna |
| Company | Survived. Three co-founders stay |
The same buyer, the same shape, twice. This brief covered the deal itself in "Nvidia pays Poolside $6bn — the company selling the chips will now build the models."
5. Which is why it became contested
Form not being an acquisition does not settle whether substance is.
In the United States, Senators Elizabeth Warren and Richard Blumenthal wrote to the Department of Justice and the Federal Trade Commission about the Groq arrangement. Their phrase for it was "reverse acquihire."
The objection reduces to three pairings.
| The form | The substance |
|---|---|
| The company survives | Once the core engineers leave, the same capability does not |
| The licence is non-exclusive | In practice only one buyer can deploy the technology at scale |
| The stake is a minority one | Whoever controls hardware supply sets the target's operating terms |
The third matters most in AI specifically.
Poolside's own letter to investors makes the point without meaning to: staying in the open-source model race would have required more Nvidia hardware than the company could obtain.
To build models you need GPUs; at scale there is effectively one supplier. Before you can be a competitor, you are a customer.
This brief covered another instance of that leverage in "Nvidia's $500 billion computing finance platform — GPUs became collateral."
6. Four questions to ask of any such deal
| Question | Why it matters |
|---|---|
| Is the licence exclusive? | Only a non-exclusive licence supports the independence claim |
| How many people move? | If the core team goes wholesale, the substance is an acquisition |
| What business actually remains? | If what is left is peripheral, only the form survived |
| Is the buyer the target's supplier? | If so, the asymmetry outlives the transaction |
For Poolside, the first is still unanswered — exclusivity was not specified at announcement.
7. In short
- A licensing deal buys technology use, people and a minority stake, and leaves the company standing.
- It is not an acquisition, so merger review is hard to trigger. That is the main reason it exists.
- Nvidia has used it twice — Groq and Poolside.
- Two US senators called it a "reverse acquihire" and asked antitrust regulators to examine it.
- Form and substance separate at the movement of key people and at the supplier-customer asymmetry.
8. What we could not confirm
- Groq's deal value — $20 billion was cited, but reports conflicted on whether it was a purchase price or a licence fee.
- Whether regulators opened a review — the letter is confirmed; the response is not.
- Poolside licence exclusivity — not specified at announcement.
- Treatment under Korean law — no authoritative reading found on merger-notification obligations.
- The "three parts" framework — a summary of confirmed cases, not a defined legal category.
Sources
- Groq Newsroom — Groq and Nvidia Enter Non-Exclusive Inference Technology Licensing Agreement
- Constellation Research — Nvidia's Groq deal: Acquisition, acquihire or creative licensing deal?
- Data Center Dynamics — Nvidia to license tech from AI inference chip company Groq, hire its leadership
- GovConWire — Groq Licenses AI Inference Tech to NVIDIA in Non-Exclusive Deal
- PYMNTS — Nvidia Pays $6 Billion to License Poolside AI Model-Development Software
- The Next Web — Nvidia pays Poolside $6bn to license its model factory and hire 109 staff
- The Motley Fool — Nvidia's 'Aqui-Hire' of Groq Eliminates a Potential Competitor