"1 Person + 10,000 GPUs" - Can This Formula Create a Unicorn? OpenAI's Prophecy Is Being Verified in Shanghai

OpenAI CEO Sam Altman said that in the future, one person plus ten thousand GPUs could create a billion-dollar unicorn.

Sam Altman.

Whether Altman’s prediction will come true is being tested in the AI innovation towns of North Yangpu in Xuhui and Zhangjiang in Pudong. These areas are rapidly gathering a group of OPCs (one-person companies), with many founders even eager to graduate early.

Currently Founding

Peng Zimin, a PhD student at Shanghai Chuangzhi College, registered Suiyue Intelligence a few months ago.

Peng Zimin chose to start a business rather than seek employment.

Vibe coding has made him restless. This AI programming buzzword for 2025 enables users to create websites, mini-programs, and even games without coding—simply by describing ideas to AI. With Vibe coding, Austrian tech enthusiast Peter Sternberg developed the open-source personal AI agent project OpenClaw in just ten days, which became a global hit, making him a “super individual” in the AI era.

Peter Sternberg developed the globally explosive OpenClaw.

Peng Zimin aims to build a technical platform behind the super individual. He believes that open-source supply is booming, and super individuals need an Agent (intelligent agent) to help lower the barriers to using the entire open-source chain.

Currently, Peng Zimin has secured funding from Qiji Chuangtan, and his product is undergoing paid validation.

Qiji Chuangtan, founded by former Microsoft and Baidu executive Dr. Lu Qi, is a benchmark for angel investment in AI hardware technology. It incubates about 120 projects annually, providing each with a $300,000 seed fund, but less than 1% of applicants are accepted.

Qiji Chuangtan, founded by former Microsoft and Baidu executive Dr. Lu Qi.

Similarly, doctoral student Zhang Kexin is also racing against time, founding a drug design platform called “Lüsheng Wanyou.”

According to Zhang Kexin, some specialized models in biomedicine are “powerful but expensive,” making them hard for scientists and R&D personnel in pharmaceutical companies to accept. Zhang Kexin aims to integrate AI resources to offer a more affordable and efficient drug design platform. On his Lüsheng Wanyou platform, the time to connect a molecule has been reduced from dozens of seconds via traditional methods to just a few tenths of a second.

Zhang Kexin shared a real-life story with reporters—his ideal scenario—four years ago, Facebook founder Mark Zuckerberg laid off a protein team of over ten scientists. Unwilling to give up, they started an AI protein company. Two years later, Zuckerberg had to buy it back at several times the original cost.

Combining his previous internship at ByteDance and involvement in developing a large scientific model, Zhang Kexin believes that major domestic and international tech companies will continue to make more refined moves in “AI + scientific research,” which is a key reason for his quick entry into the field.

Fear of Missing Out

Ding Juntao also fears missing out.

As the founder of Gusu Technology, he clearly sees the demand—platforms like Xiaohongshu need “daily updated” videos to attract traffic, but video production is time-consuming. They urgently need an Agent to help screen clips, craft stories, write scripts, and edit videos. Currently, Gusu Technology’s Agent product is inviting initial users for in-depth testing.

Ding Juntao said, “The market is eager, and leading companies haven’t acted yet. If I miss this wave, I will regret it!”

His company, including himself and two co-founders, has six people in total, with three interns. Ding Juntao remains calm because, under AI empowerment, “a single person can now do what previously required a team—becoming a full-stack engineer, naturally leading to OPCs.”

OPC, literally “one-person company,” is commonly used to describe lightweight startups operated by very small teams. According to the AI department of the Shanghai Municipal Commission of Economy and Information Technology, OPC emphasizes the efficiency change of “people + AI,” and most AI companies with fewer than 20 employees can be considered OPCs.

OPC allows a self-described “pure liberal arts student” screenwriter Zhang Chenchen to “form a team alone,” founding Liu Meng Culture.

Zhang Chenchen explained that her two dramas took six and eight years respectively to finally air on Hunan TV and Tencent platforms. It wasn’t until someone suggested she try large models. Last year, SenseTime launched the intelligent agent Seko. After experimenting with prompts and storyboarding, she decided to establish her own company and enter the AI short drama track within a week.

She said AI gave her courage and amplified her personal abilities and creative stories exponentially.

SenseTime’s Seko 2.0 makes “one-person crews” in the short and manga drama industries possible.

Yang Yang also practices lightweight operation.

She has a bachelor’s and master’s degree in coding, then pursued a PhD in psychology at Southwest University in the US. Now she founded Pandors, focusing on non-invasive brain-computer interfaces. Pandors developed its own chips, creating a 64-dimensional emotion model—the first of its kind globally—and developed intervention schemes covering audio, video, and gaming, offering services in cognitive rehabilitation, emotion analysis, and health alerts. One of Pandors’ consumer brain-machine headbands quickly became the top seller on JD.com within three months of launch.

A consumer brain-machine headband from Pandors.

Yang Yang explained that the minimal team for running a hard tech product is just four people: herself, one software engineer, one hardware engineer, and an artist/developer.

With such a lean team, the secret is AI. For example, developing audio, video, and gaming for therapy or intervention usually takes one to two years, but with AI assistance, a demo version of a new game can be created in a day. Additionally, Pandors can process over 20 user feedbacks daily, reducing product development to about two weeks from start to delivery. Yang Yang said, “Academic background is no longer as important; computing power and creativity are key.”

And “Comrades”

Shanghai, keenly aware of OPC trends, is also quick to act.

Since launching the “Super Individual 288 Action” in Lingang last August, and with Pudong and Xuhui introducing targeted support policies for OPCs by December, over 600 OPCs have settled in the two key innovation towns—Xuhui Beiyang and Zhangjiang in Pudong.

The service capacity of a “dedicated government” behind OPCs will determine their quantity and quality.

In mid-2022, Kong Xiangyu quit his job in the US to focus on “AI + Education,” founding an OPC. He was once torn over which city to settle in. While researching policies to attract tech talent, he was touched by the details in the Shanghai Pudong International Talent Port mini-program—each step of the application process had a contact person and phone number, emphasizing reliability.

Now, Kong Xiangyu’s Lingohow has moved into the East Tower of Zhangjiang Science City’s Twin Towers, enjoying a six-month rent-free period for four workstations. What reassures him even more is the support from Zhangjiang AI Innovation Town—offering subsidies for computing power, models, and corpora—eliminating the biggest initial cost concerns and enabling more aggressive moves.

Zhangjiang Science City Twin Towers.

In Zhangjiang and Beiyang, whether it’s Shengxi Fasuo working on “AI + Law” or Gusu Technology creating AI video editing tools, the government helps “connect” with investors. Lin Le, deputy director of the Xuhui AI special team, said that Xuhui has established a Youth Entrepreneurship Fund, known as the “Early Bird Fund.” In less than three months, Beiyang AI Innovation Town has completed 50 seed investments.

In the eyes of many OPC founders, innovation towns are not just “landlords,” but “comrades.” Yang Yang said that when companies focus on “hammers” (technology), the government always reminds and recommends “nails” (scenarios). Her Pandors company, less than a week after landing in Shanghai, struck a major health cooperation deal with a large enterprise with over 50 chain stores. Considering brain-machine interfaces controlling embodied intelligent robots, Zhangjiang also enthusiastically recommended local humanoid robot resources to Pandors.

The reporter saw in the policy packages of the two innovation towns that the “comrades” provide comprehensive services for OPCs—from talent apartments, financing, and application scenario support to offering display windows, supporting overseas orders, and providing agency bookkeeping and legal services.

OPCs, though highly efficient, are not fighting alone. AI lowers the barrier to entrepreneurship, but a city’s overall attitude and infrastructure raise the ceiling of what can be achieved.