TL;DR
- Palo Alto Networks (NASDAQ: PANW) agreed on November 18, 2026 to acquire Applied Compute, a San Francisco AI startup founded in 2025 by former OpenAI researchers, in an all-stock deal valued at $4.5 billion.
- Applied Compute's AC2 platform post-trains and serves open-weight AI models on company data; the company reported a $100 million annualized revenue run rate on November 2, 2026.
- The $4.5 billion price sits about 38% above the $3.25 billion valuation at which Applied Compute raised $350 million earlier this month.
- The deal is announced, not closed: it is expected to close in Palo Alto's fiscal third quarter of 2027, subject to regulatory clearance, and Palo Alto expects it to add $1 billion of ARR by fiscal 2029.
Watch the breakdown in 40 seconds: Episode 30, "Palo Alto Bought the AI Factory," is live on Instagram — https://www.instagram.com/reel/Df6KoWIvfXB/
The most expensive AI purchase of the year is not a chatbot company, a chip company, or an app. It is a factory.
Palo Alto Networks, the largest cybersecurity company in the world, announced on Wednesday that it has entered a definitive agreement to acquire Applied Compute, a San Francisco startup that builds the platform enterprises use to post-train and serve open-weight AI models. The deal is all stock, valued at $4.5 billion, subject to adjustments. It is expected to close in Palo Alto's fiscal third quarter of 2027, subject to customary closing conditions including regulatory clearance.
Applied Compute was founded in 2025 by Yash Patil, Rhythm Garg, and Linden Li, researchers who previously worked at OpenAI on products including Codex and the o1 reasoning model, and on reinforcement-learning infrastructure. Its platform, AC2, introduced in August, lets a company fine-tune open-weight models on its own data, run inference, and keep improving those models from production feedback — in public clouds or in sovereign clouds. The company has about 70 employees, all U.S.-based and all expected to join Palo Alto Networks. Patil, its co-founder and CEO, becomes a Co-Chief AI Officer at Palo Alto after the close.
The numbers describe the fastest repricing of the AI cycle so far. On November 2, Applied Compute said it had reached a $100 million annualized revenue run rate, less than eighteen months after founding. Earlier this month it raised $350 million at a $3.25 billion valuation, in a round led by Elad Gil with participation from Nvidia, Blackstone, Kleiner Perkins, and Lux Capital, among others. Two weeks later, Palo Alto agreed to pay $4.5 billion — roughly 38% above that mark, for a company that has not yet had a second birthday.
"Every enterprise is going to run a mix of open and closed models, and every one of those models needs to be secured," Palo Alto chairman and CEO Nikesh Arora said in the announcement. "Applied Compute has built the best post-training and inference platform we've seen, and its research team is exceptional. We saw that firsthand as a prospective customer, and our Unit 42 Frontier AI Assessment confirmed their strong security foundations." Speaking to the Wall Street Journal, Arora described the deal as Palo Alto venturing outside of cybersecurity.
The plan attached to the price is an infrastructure build, not a feature launch. Palo Alto intends to build more than 100 megawatts of inference capacity around the platform, with projected capital spending of under $3 billion across fiscal 2027 through 2030, and expects the acquisition to contribute $1 billion of annual recurring revenue by fiscal 2029. That is the tell. The world's biggest security vendor is not buying a product to resell. It is buying the capacity to train, refine, and run the models its customers' agents will live on — and then sell the security layer on top of its own factory floor.
Read it as the latest move in a pattern this publication has tracked all year: the money in AI is migrating from the models everyone can see to the infrastructure underneath them. Open-weight models only become enterprise products once someone owns the post-training, the inference, and the feedback loop. Until today, that layer belonged to startups. Now the largest incumbent in enterprise security owns a leading piece of it — and every other platform vendor has to answer for why it does not.
The announcement landed on a strong day for the buyer. Palo Alto also reported fiscal first-quarter results on Wednesday: revenue of $3.44 billion, up 39% year over year, ahead of expectations, with the full-year outlook raised. The shares rose on the combined news.
What to watch: the regulatory path to a fiscal Q3 2027 close; whether the roughly 70-person team stays through it; whether other security and cloud incumbents answer with platform acquisitions of their own; and whether the $1 billion ARR-by-2029 target survives contact with an inference market whose prices fall every quarter. The factory has been bought. What it produces is the next story.
Sources: Palo Alto Networks press release via PR Newswire, "Palo Alto Networks to Acquire Applied Compute, Bringing Secure Open-Weight AI to the Enterprise," November 18, 2026 (also filed as Exhibit 99.2 to Form 8-K); Wall Street Journal; SiliconANGLE; Security Boulevard; Unite.AI.
The White Report — Where Super Intelligence Meets Money.