China’s open-source embrace upends conventional wisdom around artificial intelligence

China has embraced open source AI models for trend market watchers and insiders, boosting the adoption and innovation of AI in the country, with some people suggesting it is a “android moment” for the sector.

Open Source Shift AI startup Deepseek was on the lead, and earlier this year the R1 model challenged America’s technological advantage, raising questions about Big Tech’s large language model and large spending on data centers.

R1 created a splash in the sector for its performance and low-cost claims, but some analysts say Deepseek’s most important impact is driving the adoption of open source AI models.

“Deepseek’s success proves that open source strategies can lead to faster innovation and wider adoption,” said Wei Sun, a leading artificial intelligence analyst at Counterpoint Research.

“Now, R1 is actively reshaping the Chinese AI landscape, and we see that large companies like Baidu will move their own LLMS to open source with a strategic response,” she added.

On March 16, Baidu released the latest version of the AI ​​model, Ernie 4.5, and a new inference model, Ernie X1, which was free for individual users. Baidu is also planning How to open source the Arnie 4.5 model series from the end of June.

Experts say Baidu’s open source plan represents a broader change in China, moving away from its own licensing-focused business strategy.

“Baidu has always been a huge supporter of its own business model and has been speaking out against open source, but disruptors like Deepseek have proven that the open source model is just as competitive and reliable as its own.”

Open source generally refers to software in which the source code is freely available on the web, and may be modified and redistributed.

The AI ​​model that called itself open source existed before the advent of Deepseek. Meta‘s llama and GoogleHowever, Gemma is a major example of the United States. Some experts These models argue that they are not actually open source because the license limits certain uses and changes and the training dataset is not public.

Deepseek’s R1 is distributed under the “MIT License.” This is one of the most acceptable and widely adopted open source licenses for Counterpoint Sun, which promotes unlimited use, modifications and distribution, including commercial purposes.

The Deepseek team said,Open Source Week“Last month we released more technical details on the development of the R1 model.

The DeepSeek model itself is free, but the launch fee for the application programming interface allows the AI ​​model and its features to be integrated into applications from other companies. However, API fees are advertised much cheaper than the latest products from Openai and Anthropic.

Openai and humanity generate revenue by accessing parts of the model to individual users and businesses. These models are considered “closed sources” because they are datasets, and the algorithm is not open for public access.

In addition to Baidu, Alibaba Group and Tencent We offer AI free of charge, and more models are open sourced.

For example, Alibaba Cloud said last month it was open source. AI Models for Video Generationreportedly released by Tencent Five new open source models Earlier this month, there is the ability to convert text and images into 3D visuals.

Smaller players are also driving trends. Manusai, a Chinese AI company that recently launched an AI agent that claims to be superior to Openai’s deep research, says it will shift to open source.

“This is not possible without an incredible open source community, so we promise to give back.” Demo Video. “Manusai operates as a multi-agent system with several different models, so we are looking to open source some of these models later this year,” he added.

Zhipu AI, one of the nation’s leading AI startups, announced this month on WeChat that 2025 will be the “year of open source.”

Ray Wang, lead analyst and founder of Constellation Research, told CNBC that companies are being forced to make these moves following the emergence of Deepseek.

“With Deepseek Free, it’s impossible for other Chinese competitors to charge the same thing. They need to move to an open source business model to compete,” says Wang.

AI scholar and entrepreneur Kai-Fu Lee believes this dynamic will affect Openai. Social Media Posts When competition is “free and formidable,” it is difficult for a company to justify pricing.

“The biggest revelation from Deepseek is that open source won,” says Lee, who has built an LLM platform for businesses that want to use Deepseek.

Openai launched AI Frenzy when it released its ChatGpt bot in November 2022, but does not show that it is planning to shift away from its own business model. The company that began as a nonprofit in 2015 is Commercial structure.

Sun says Openai and Deepseek both represent very different ends of AI space. She adds it The sector continues to see the division between open source players innovating with one another and closed companies under pressure to maintain high-cost, cutting-edge models.

The open source trend has raised questions about the large funds raised by companies such as Openai. Microsoft invested $13 billion in the company. CNBC is currently in discussions to raise up to $40 billion in the funding round that it confirmed at the end of January that it will raise its valuation to $340 billion.

In September, CNBC confirmed that the company expects a loss of around $5 billion, earning revenue of $3.7 billion. Openai CFO Sarah Friar said $11 billion in revenue for the company this year is “undetectedly in the realm of possibilities.”

Meanwhile, Chinese companies are choosing the open source route to compete with the more unique approaches of US companies, said Wang of Constellation Research. “They want to adopt faster than the US closure model,” he added.

Speaking to CNBC’s “Street Sign Asia” on Wednesday, Tim Wang, managing partner of tech-focused hedge fund monolith management, said the model from companies such as Deepseek is “China’s excellent enablers and multipliers,” indicating how to do things with more limited resources.

According to Wang, the open source model has pushed down costs and opened the door for product innovation. He says that Chinese companies have been very good historically.

He calls development “Android moments.” Operating system source code is freely availablefacilitating innovation and development of the non-apple app ecosystem.

“We thought China was 12-24 months behind. [the U.S.] With AI, I think it’s probably three to six months,” says Wang.

However, other experts have downplayed the idea that open source AI should be seen through the lens of China-US competition. In fact, some US companies have consolidated and benefited from Deepseek’s R1.

“I don’t think the so-called Deepseek moment is about whether China is better than the US or vice versa. It’s really about the power of open source,” Alibaba Group chairman Joe Tsai told CNBC’s Converge Conference in Singapore earlier this month.

Tsai added that the open source model empowers AI to everyone, from small entrepreneurs to large companies, leading to more development, innovation and spread of AI applications.