Gangsta AI

Nvidia Just Paid $6 Billion Not to Buy a Company — and Hired 109 of Its People Anyway

Deal Desk

By The Gangsta AI News Desk · 2026-08-22 · 4 min read

Nvidia has found a new way to swallow a startup without buying it. According to reporting first surfaced by Newcomer and confirmed by Bloomberg and The Information, the chip giant has agreed to pay AI coding company Poolside a $6 billion licensing fee — plus a $1 billion investment at a $12 billion valuation — and will extend job offers to 109 of Poolside's engineers, the exact team that built its models.

Read that again. Nvidia isn't acquiring Poolside. Poolside keeps its name, keeps operating, and its three founders are staying put. Nvidia is simply licensing the crown jewels — Poolside's "Model Factory," the pipeline it uses to build its Laguna family of open-weight coding models — and quietly walking 109 of the humans who run it across the street.

“Nvidia gets the models, the method, and most of the minds. Poolside gets $6 billion for its investors — and a stock ticker's worth of goodwill it can sell all over again.”

Here's the part that should make every rival squint: the license is non-exclusive. Poolside can turn around and offer the same Model Factory rights to Google, Amazon, or anyone else with a checkbook. The reported $6 billion is slated to reach Poolside's investors by the end of 2027. Wall Street wasn't thrilled — NVDA closed the week down roughly 5%.

Why a chip company wants a code factory

Poolside's Laguna models are built to automate software engineering, with a heavy tilt toward government and defense code. For Nvidia, owning that capability — without the antitrust headache and integration mess of a full acquisition — does two things at once. It advances Nvidia's push to seed open-weight AI everywhere, and it guarantees long-term demand for the one thing Nvidia actually sells: the GPUs those models run on. Fund the software, lock in the hardware. It's the same circular logic powering half the deals in AI right now, just dressed up as a licensing agreement.

This is also the third flavor of the same 2026 playbook. Meta bought its way into talent, Microsoft rented it, and now Nvidia has invented the reverse-acquihire: license the tech, hire the team, leave the shell standing. When a $6 billion "license" comes with 109 job offers attached, the word is doing a lot of heavy lifting.

The real lesson: no single coding model is safe at the top

Step back and the deal tells you something bigger about where AI is headed. The reason Nvidia paid a fortune for a code factory — rather than a finished model — is that no single coding model stays best for long. Laguna is open-weight. DeepSeek just priced frontier coding at pennies. Alibaba's Qwen is giving away trillion-parameter weights. GPT-5.6, Claude Opus 5, and Gemini 3.7 trade the lead every few weeks. The value isn't in any one model; it's in the ability to keep making better ones.

That's exactly the trap that catches everyday users. Pick one AI, trust it blindly, and you're betting your answer on whichever model happened to win last Tuesday. The smarter move mirrors what Nvidia just did at billion-dollar scale: don't marry a single model — cross-check them.

That's the whole idea behind Gangsta AI. Ask one question, fire it at ChatGPT, Claude, Gemini, Grok and 30+ more at once, and get a single, cross-checked verdict instead of gambling on one vendor's best guess this week. When the biggest company in AI won't bet on one model, why should you?

Want to know which model actually leads for coding, reasoning, or writing right now? See our always-updated breakdown of the best AI models.

Try Gangsta AI free →

More: Best AI models · Compare all AI · Frontier Models · All articles