Ex-Alibaba AI Chief's Startup Hits $5 Billion Valuation in Three Months
Pragmatik Labs, the Shanghai-based AI startup founded by former Alibaba research leader Lin Junyang, has raised a new funding round at a $5 billion valuation, more than doubling its worth since its August debut, according to Bloomberg reporting cited by multiple outlets on Friday [1][2][3].
The new round brought in tens of millions of dollars — a modest sum compared to the several hundred million dollars raised in August at a $2 billion valuation. New investors include HyT Capital, CDH Investments and automaker Geely Automobile Holdings [2]. Zaobao reports the same investors but transliterates the first as "Hua Capital" [3]. The company, also known as p7k, did not respond to requests for comment from Bloomberg or PYMNTS [1][2].
From Qwen to a "neolab"
Lin was a central figure in Alibaba's pivot to AI, credited with leading development of the Qwen family of large language models that underpin the company's main AI applications and services [1][2][3]. He announced his departure on X on March 3, writing "bye my beloved qwen." and revealed the new venture on August 11, saying Gaorong Ventures and HSG (also known as HongShan) co-led the initial round with participation from Tencent and the Shanghai Engine Fund (Shanghai Future Industry Fund) [1]. Pragmatik Labs' website went live at the same time [3].
The speed of the valuation jump — from $2 billion to $5 billion in roughly three months, on a comparatively small capital injection — reflects intense investor appetite for Chinese "neolabs": research-oriented startups founded by veterans of big-tech AI labs, aiming to challenge established players such as DeepSeek and Moonshot [2]. The pattern mirrors the U.S. trajectory where researchers like Ilya Sutskever and Yann LeCun left large companies to found ventures of their own [2].
Agents for digital and physical worlds
Pragmatik Labs says it is focused on "the research of next-generation agents across digital and physical worlds." Its website describes digital agents for knowledge work, operations and industry-scale workflows, and physical agents — embodied intelligence that "adapts, acts and solves long-horizon tasks in the world" [1]. The company frames its work as building systems that "reason, use tools, learn from feedback and coordinate action" [1]. The company's public materials outline research directions [1]. Bloomberg reports that the neolabs want to emulate what Anthropic and OpenAI have achieved but within narrower specialties such as equipping AI with an understanding of the physical world or creating coding systems that can improve on their own [2].
Founder's own caution
Lin has publicly tempered expectations for China's AI trajectory. He told Bloomberg that China probably will not overtake Anthropic and OpenAI within the coming three to five years, citing limited computing resources — a constraint expected to worsen after Washington introduced new restrictions aimed at preventing Chinese access to Nvidia's high-end AI accelerators through data centers in Southeast Asia [2][3].
That assessment is notable coming from a founder whose startup just commanded a $5 billion valuation on a small raise. It underscores the gap between private-market enthusiasm and the hardware realities that govern large-model development.
Alibaba presses on
Meanwhile, Alibaba continues to pour capital into AI infrastructure. On Thursday, the company reported that capital expenditures rose 60% year over year to $7.5 billion. Its AI Cloud and Compute Services segment revenue increased 53% to $9.1 billion, and its AI Labs and Applications segment revenue grew 51% to $661 million [1].
Lin's departure and rapid fundraising illustrate how the talent that built Alibaba's core AI stack is now being backed to pursue more ambitious, agent-oriented architectures — but whether those architectures can deliver on their promise without the compute scale of U.S. hyperscalers remains an open question.