China’s AI ecosystem is going strong. So much so that it is starting to compete internally, and the much-talked-about DeepSeek is no longer at the top.
Zhipu AI is not a name that typically comes up in casual conversations about AI supremacy. It doesn’t have the fanfare of DeepSeek or the benchmark-breaking headlines of Alibaba’s latest models.
But this week, OpenAI made it clear that Zhipu is a threat worth watching.
In a blog post that reads more like a geopolitical intelligence memo than a developer update, OpenAI called out the Beijing-backed startup as a significant player in China’s AI playbook. “While we hear the most about new models, just as significant is CCP headway in getting other governments around the world to adopt its AI,” the post warned.
At the centre of that strategy is Zhipu AI, which appears to be building the scaffolding for China’s global AI infrastructure.
Despite US sanctions, Zhipu is not starved for capital. According to a Reuters report, it recently raised $69 million in a Series D round led by state-owned Huafa Group. This comes on the heels of two recent funding rounds that the company secured from multiple local government bodies.
Its current valuation is estimated to be $2.74 billion. The company has reportedly begun preliminary steps toward an IPO. Zhipu has also received support from Chinese tech giants Tencent and Alibaba, further blurring the line between state-backed innovation and private-sector speed.
More Than Just Another LLM Startup
Founded in 2019, Zhipu AI has been dubbed one of China’s “AI tigers,” a term used by Chinese state media to describe the handful of LLM unicorns spearheading Beijing’s push to reduce dependence on Western technology.
Unlike DeepSeek, which has become the poster child of Chinese AI ambition with its R1 model, Zhipu is playing a different game — one that involves partnerships with foreign governments, stealthy international expansion, and direct support from the Chinese Communist Party.
State media reports say the startup has secured over $1.4 billion in state-backed investment and frequently engages with top-level Chinese officials, including Premier Li Qiang.
The company also reportedly has working relationships with the Chinese military — a detail that led to its inclusion on the US Commerce Department’s Entity List in January 2025, effectively blacklisting it from buying American components.
But the ban hasn’t slowed it down.
Zhipu AI has established offices in Singapore, Malaysia, the United Kingdom, and the Middle East. It’s also running joint “innovation centers” in Southeast Asia — including in Indonesia and Vietnam — as part of what OpenAI describes as China’s strategy to embed its AI stack globally before Western alternatives can take hold.
What is up with OpenAI?
OpenAI’s blog post seems less concerned with model performance and more alarmed about the architecture of influence. By offering “AI infrastructure solutions to governments around the world,” Zhipu is laying the groundwork for a global Chinese AI ecosystem, one that could prove sticky for decades.
“The goal is to lock Chinese systems and standards into emerging markets before US or European rivals can, while showcasing a ‘responsible, transparent and audit-ready’ Chinese AI alternative,” OpenAI wrote.
The language is telling. This is not about tech supremacy in the traditional sense. It’s about building trust, negotiating procurement deals, and establishing data and infrastructure dependencies.
Just this week, Zhipu AI launched AutoLM Rumination, a new AI agent that can carry out deep research, draft comprehensive reports, and help users plan complex tasks. Powered by the company’s proprietary models, it’s available for free.
According to Zhipu, these models are not only as capable as DeepSeek R1, but also run eight times faster while consuming just one-thirtieth the compute resources. That’s a remarkable claim — one that, if accurate, could give Zhipu serious leverage in low-resource markets.
The release comes shortly after another Chinese player, Manus AI, launched a general-purpose agent that outperformed OpenAI’s deep research tools on the GAIA benchmark. Manus AI is now a commercial product with monthly plans, suggesting that Chinese companies are not just building fast but are also monetising fast.
VC Bill Gurley questioned OpenAI’s motives in spotlighting Zhipu AI, suggesting the move may have unintended consequences. “Very odd decision for OAI to openly promote this Chinese AI company ‘Zhipu AI’ in a blog post,” he wrote on X. He suggests that the action taken simply boosts competitors’ visibility among both funders & customers, effectively putting them on the map.
Gurley’s criticism echoes a broader sentiment in Silicon Valley that Zhipu, once obscure outside China, now finds itself in the global spotlight not because of a model release but because OpenAI chose to write about it.
“The opposite of love is indifference,” Gurley added, implying that attention, even critical, is a kind of endorsement.
OpenAI’s unease isn’t just about competition. It’s about losing narrative control. While the US has been aggressively promoting its AI stack globally through Project Stargate and government deals, trade missions, and strategic partnerships, China’s quiet, infrastructure-first approach may prove more enduring.
This is the US building walls. And unlike DeepSeek, which focuses solely on performance and visibility, Zhipu may be the one digging the trenches for now.
