Zhou Hongyi on DeepSeek's open source: technical confidence, the AI era needs "sovereign AI"

Download the app

+1 Zhou Hongyi on DeepSeek open source: technological confidence, the AI era needs "sovereign AI" The Paper reporter Fan Jialai 2026-10-23 16:35 Source: The Paper ∙ 10% Company > Listen to full text Font size

Recently, DeepSeek announced that it would fully open source DeepInfer, the AI inference framework used internally, becoming another landmark move in the open source community.

On October 23, 360 Group founder Zhou Hongyi told The Paper that behind this "is a kind of technological confidence, and even more a philosophy of open source inclusiveness"; looking deeper, this matter also has a more important significance—technological sovereignty in the AI era. He also disclosed that 360 has built a sovereign AI operating system, which has already been deployed in some government and enterprise units.

It is understood that DeepInfer, through optimized scheduling, lets the CPU and graphics card work at the same time, reducing waiting and repeated computation, allowing the same machine to handle more tasks. In Zhou Hongyi's view, the value of such open source achievements goes beyond the technology itself: as open source models and inference technology continue to mature, enterprises no longer need to train large models from scratch, and also have the opportunity to deploy AI on their own servers, build agents that truly understand their own business, and keep competitiveness in their own hands.

Zhou Hongyi further mentioned that in the agent era, AI sovereignty is especially important for enterprises. Many enterprises currently use AI that is "rented"—the models, computing power and platforms are all in others' hands, like building on a faucet controlled by someone else: if the other party raises prices, you can only accept it; if the service is adjusted, the business may stall along with it. More critically, once AI starts working for enterprises, who owns the data processed by agents every day and the knowledge and experience accumulated directly determines the boundary of an enterprise's core assets. "The stronger AI becomes, the more enterprises are constrained by others—this is absolutely unacceptable."

In his view, what sovereign AI must solve is who controls the models, computing power and data, and who owns the knowledge and experience. He introduced that the sovereign AI operating system 360 has built covers capabilities including computing resource and scheduling, sovereign large models, enterprise AI knowledge bases, agent factories, Agent Teams (human-machine collaboration platform) and security governance. In response to the potential risk of agents operating beyond their authority, the system builds in security capabilities, controlling and auditing agents' permissions, tool calls and execution processes, with key operations checked by humans, and ultimate control kept in the hands of enterprises.

Sovereign AI is becoming a global issue. McKinsey estimated in a 2026 report that by 2030, 30% to 40% of global AI spending may be affected by sovereignty requirements, corresponding to a market size of about $500 billion to $600 billion. Responsible editor: Zheng Jingxin Image editor: Shen Ke Proofreader: Zhang Liangliang The Paper news tip: 021-962866 The Paper, no reproduction without authorization +1 Favorite I want to report #ZhouHongyi #DeepSeekOpenSource #SovereignAI View more