Palo Alto Networks is acquiring Applied Compute, a platform for post-training and inference of open-weight AI models. The platform enables companies to securely train and deploy open models, even in sovereign environments.
The $4.5 billion figure is subject to adjustment. Two weeks earlier, Applied Compute raised $350 million at a $3.25 billion valuation. Elad Gil led that round, with investors including Kleiner Perkins, Lux Capital, Nvidia, and Blackstone.
Open models with built-in security
According to Palo Alto, AI within companies is becoming hybrid. Organizations are combining closed-source frontier models with open-weight models tailored to their own data, costs, and control. However, the company states that no major provider currently delivers the second category securely. It aims to change that with Applied Compute.
The startup builds on open models such as Qwen, Kimi, GLM, and DeepSeek. It further trains these models on customer data and company-specific workflows and then handles deployment, evaluation, and continuous improvement. Customers will soon be able to do this in the public cloud or in sovereign environments. Meanwhile, Applied Compute continues to serve AI-native companies, digital natives, and large enterprises.
“Every enterprise is going to run a mix of open and closed models, and every one of those models needs to be secured,” says CEO Nikesh Arora. Palo Alto initially evaluated the platform as a potential customer. Its Unit 42 research division then assessed the security fundamentals through a Frontier AI Assessment.
Applied Compute was founded in 2025 by three former OpenAI researchers. CEO Yash Patil worked there on post-training infrastructure and Codex, while CTO Rhythm Garg contributed to the o1 reasoning model. After the acquisition closes, Patil will become Co-Chief AI Officer at Palo Alto. His team will integrate AI across the entire product portfolio and train the security specialist’s large and small models.
“Joining Palo Alto Networks will put security at the center of all our products ,” said Patil.