TLDR
- Baidu launched M100 inference chip (early 2026) and M300 training/inference chip (early 2027) at its annual tech conference Thursday.
- Tianchi 256 and Tianchi 512 supernode products will connect multiple P800 chips using advanced networking technology.
- The company released Ernie 5.0, an upgraded language model capable of processing text, video, and images.
- BIDU stock gained 1.12% in pre-market hours after closing down 2.28% at $128.94 the previous session.
- New products offer Chinese firms domestic alternatives as US export restrictions limit access to American AI chips.
Baidu rolled out two AI semiconductors Thursday at its Baidu World technology conference. The M100 targets inference tasks with an early 2026 release date.
The M300 handles both training and inference work. It arrives in early 2027.
The chips give Chinese businesses access to domestic computing power. US restrictions block many advanced American processors from reaching Chinese companies.
BIDU shares rose 1.12% in pre-market trading following the announcement. The stock finished the previous day at $128.94, down 2.28%.
Linking Chips Through Supernode Tech
The company introduced two supernode systems at the event. Tianchi 256 packs 256 P800 chips into one unit and ships in the first half of next year.
Tianchi 512 doubles that count with 512 chips. Launch comes in the second half of 2026.
Supernodes use networking to connect multiple processors. This compensates for performance gaps in individual chips.
Huawei deployed similar technology with its CloudMatrix 384 system. That product uses 384 Ascend 910C processors.
Industry watchers consider Huawei’s system stronger than Nvidia’s GB200 NVL72. Baidu’s founder Robin Li pledged continued investment in advanced models.
The company started developing its own chips back in 2011. The chip division now represents a growing part of operations.
Multi-Format AI Model Goes Live
Baidu unveiled Ernie 5.0 during the same presentation. The language model processes three formats: text, images, and video.
Previous versions focused mainly on text. The upgrade lets users submit commands across different media types.
Training creates AI models by analyzing large datasets. Inference applies those models to real-world tasks and user queries.
The M100 specializes in inference operations. The M300 covers the complete AI workflow from training through inference.
Chinese Tech Race Accelerates
Alibaba dropped its own AI news Thursday. The company plans to rebrand its Tongyi app as “Qwen” with shopping features.
Around 100 Alibaba engineers are handling the makeover. An international version is in the works.
Alibaba stock surged 4.4% in pre-market trading. Year-to-date gains hit 88% for the company.
Baidu stock is up 53% this year. Analysts point to chips and robotaxis as key growth areas going forward.
The new Baidu chips provide low-cost options for Chinese businesses. They operate under domestic control without US technology dependencies.
Export controls pushed Chinese companies toward homegrown solutions. Baidu’s M100 and M300 answer that need directly.
The Tianchi products link processors through networking infrastructure. This strategy boosts overall system performance beyond single-chip capabilities.
Baidu positioned Thursday’s releases as ceiling-raisers for AI intelligence. The company aims to compete with international players using domestic technology.
Get 3 Free Stock Ebooks
Discover top-performing stocks in AI, Crypto, and Technology with expert analysis.
- Top 10 AI Stocks - Leading AI companies
- Top 10 Crypto Stocks - Blockchain leaders
- Top 10 Tech Stocks - Tech giants
