A few months ago, Chinese LLM DeepSeek-R1 made headlines. The world was introduced to an open-source AI model that matched OpenAI o1’s reasoning abilities at a fraction of the cost. A relatively obscure startup topped Apple’s App Store rankings and sent shockwaves through global stock markets.
DeepSeek’s breakthrough proved AI could be built without massive capital investment or Nvidia’s highest-grade chips. More importantly, it showed that an open-source strategy could challenge America’s AI hegemony. China is betting that widespread, open-source use offers more value than the West’s paywalled models, such as ChatGPT, Anthropic’s Claude, and Perplexity.
But if DeepSeek was a minnow, a Chinese whale is now having its own open-source moment: Baidu, China’s answer to Google, with 704 million monthly active app users and a market cap of 31.6 billion USD, has announced it will open-source its powerful generative AI model, Ernie. This release represents a dramatic policy reversal for Baidu, whose CEO Robin Li had previously advocated for proprietary, closed-source models as the only viable path for AI development.
If DeepSeek demonstrated that China could compete with the West, Baidu’s open-source pivot makes Chinese AI seem almost unstoppable. The commoditization of AI is accelerating, and China’s tech giants are redrawing the battle lines with the West from a performance race into a price war.
Here’s why Baidu’s open-source play matters more than DeepSeek’s January breakthrough.
CCO at NetMind.
Deep Pockets
While DeepSeek was, at least in appearance, a proof of concept from a scrappy startup, Baidu brings the institutional weight, capital firepower, and, crucially, the distribution channels, to ensure widespread adoption.
In the age of AI, it pays to be a big company. Their sheer scale enables larger investments, improves resilience against market shocks and trade sanctions, and offers hyperscalers the advantage of applying this technology to their existing products.
Just look at Google. The American search engine giant is leveraging its existing broad and loyal customer base to attract traffic to its Gemini models, which are being integrated into its search function. At the May Google I/O conference, Google announced that its ‘AI Overviews’ (the AI summaries displayed next to search results) are used by more than 1.5 billion people each month. Baidu will doubtless do the same, leveraging its economies of scale to make Ernie a winner.
Meanwhile, DeepSeek’s momentum has sputtered. The smaller tech outfit has been forced to delay the release of its next-generation R2 model after struggling to procure enough of Nvidia’s high-end graphics processing units to complete training, because of new U.S. sanctions on chip exports to China, according to The Information.
As it stands, Baidu’s Ernie API has an 18% market share, still some way behind DeepSeek’s 34% share; however, Baidu’s size means it can rapidly make up ground.
From Performance Race to Price War
China is commoditizing AI faster than the West can monetize it. When developers can access high-performance AI without Big Tech’s pricing gates, it fundamentally rewrites who can afford to innovate.
DeepSeek-R1 debuted at $0.55 input/$2.19 output, undercutting the then SOTA model o1 by 90%+ on output token pricing. Since then, reasoning model prices have cratered, with OpenAI recently cutting its flagship model price by 80%, according to SemiAnalysis. Baidu already said in March that its Ernie X1 model delivers performance on par with DeepSeek-R1 “at only half the price.”
“If open-source AI becomes just as powerful as proprietary US models,” wrote June Yoon, Asia Lex Editor of the Financial Times in an article earlier this year, “the ability to monetize AI as an exclusive product collapses. Why pay for closed models if a free, equally capable alternative exists?”
The West may be forced to rethink its pricing strategies and business models when China can release equally good AI at virtually no cost.
Developer Ecosystem
China could never compete when the U.S. is making deals the size of Stargate, which is set to rise to $500 billion, and imposing tariffs on the tools used to build AI. That’s why China was forced to become self-sufficient.
By open-sourcing AI, they can bypass U.S. sanctions, decentralize development, and access global talent to improve models. Restrictions on Nvidia’s chips matter less when the rest of the world can refine China’s models on alternative hardware.
This approach may render restrictions moot, as much development occurs on Nvidia infrastructure outside China. By publishing the code of its flagship LLM, Baidu aims to foster broader adoption and a developer community around the technology.
CEO Li said as much to Chinese developers in April this year: “Our releases aim to empower developers to build the best applications — without having to worry about model capability, costs, or development tools.”
More developers using Ernie’s code will help it scale to new heights.
Game on
As more Chinese tech giants embrace open-source AI, they’re creating an alternative technology stack that bypasses U.S. control. Every developer who builds on Ernie instead of GPT-4 is one less customer for Silicon Valley and one more node in China’s AI ecosystem.
While DeepSeek faces delays and hardware shortages, established players like Baidu can take on the baton and sustain the open-source assault indefinitely. They have the capital to subsidize free AI and the scale to support millions of users.
For Western tech companies clinging to closed models and subscription fees, the message is clear: the game has changed. Adapt or become irrelevant.
We list the best IT Automation software.
This article was produced as part of TechRadarPro’s Expert Insights channel where we feature the best and brightest minds in the technology industry today. The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/news/submit-your-story-to-techradar-pro