China's leading search engine and AI developer, has activated a massive 30,000-chip computing cluster using its self-developed Kunlun P800 processors, CEO Robin Li announced Friday at the company's annual developer conference.

Next-Gen AI Infrastructure Goes Live

The newly "illuminated" (activated) cluster represents Baidu's third-generation Kunlun AI chips, capable of:

Training massive AI models like DeepSeek (with hundreds of billions of parameters)

Simultaneously handling 1,000+ customers fine-tuning smaller billion-parameter models

Supporting financial institutions and internet companies already adopting the P800 chips

New AI Models Unveiled

Baidu launched two upgraded systems:

Ernie 4.5 Turbo – Matches top global benchmarks in coding, reasoning, and language tasks

Ernie X1 Turbo – A specialized reasoning model for complex problem-solving

The company plans to integrate these AI capabilities across its ecosystem, including Baidu Cloud Drive and its content platform Baidu Wenku.

Shift From Models to Applications

"Models are plentiful, but apps rule the world. Without applications, models and chips are worthless," Li emphasized, signaling Baidu's pivot toward practical AI implementations beyond chatbots.

Competitive Pressures Mount

As China's AI race intensifies:

Baidu faces rivals like Alibaba, Tencent, and startups (e.g., DeepSeek) in monetizing AI

Its Ernie Bot chatbot has struggled for market share despite early ChatGPT-inspired investments

The Kunlun cluster rollout strengthens Baidu's position as one of China's few companies with full-stack AI capabilities – from chips to models to apps – amid tightening U.S. restrictions on advanced AI semiconductors.

What’s Next?
Analysts will watch whether Baidu can convert its technical investments into profitable services as China’s AI market shifts from hype to revenue-generating applications.