Brain-Inspired AI Model 🧠 5G Remote Surgery 🏥 $10B Sinopec-Aramco Deal 🛢️

2025-09-15 | subscribe | homepage

Welcome back to China Insights Weekly. Here are some of the key highlights for this week’s edition:

• Momenta and other Chinese robotaxis expand overseas through Uber, Lyft, and Grab partnerships

• Europeans now prefer Chinese cars over American ones, marking a major shift in trust

• Anti-aging boom drives 30% sales surge in China’s beauty market

• Chinese display makers surpass 50% global share, overtaking Korea and Japan

Remember to follow us on LinkedIn, X or Facebook.

🚀 Headlines

Uber and China’s Momenta to launch robotaxis in Munich in 2026, Chinese robotaxis are expanding globally through Uber, Lyft, and Grab (link, link)

Momenta, a Chinese autonomous driving company, will begin testing Level 4 robotaxis in partnership with Uber in Munich, Germany, in 2026. This collaboration aims to integrate Momenta's AI-driven autonomous driving technology into Uber's global platform, with plans to expand to other European cities. Momenta has already partnered with leading German brands like Mercedes-Benz and BMW, solidifying its position in the German automotive industry. The company has deployed ADAS systems in over 400,000 vehicles and operates a robotaxi service in Shanghai, with plans to achieve commercial deployment without onboard safety operators by year-end.

Chinese robotaxi firms are rapidly expanding overseas through strategic partnerships with global ride-hailing platforms like Uber, Lyft, and Grab. Baidu plans to launch Apollo Go in the UK and Germany in 2026 via Lyft, with further European expansion. WeRide secured investments from Grab and Uber, aiming to roll out Level 4 robotaxis in Southeast Asia and 15 new cities globally. Pony.ai partnered with Uber and ComfortDelGro for Middle East and global expansions. These collaborations leverage the platforms' vast user bases and operating systems to accelerate commercialization. Despite industry-wide losses, these partnerships aim to overcome challenges in cost reduction, regulation, and user adoption, positioning Chinese firms for international scalability.

Chinese researchers create brain-inspired AI model, Alibaba launches 1-trillion-parameter Qwen-3-Max AI model (link, link)

Chinese researchers have developed SpikingBrain, a brain-inspired large language model (LLM) that offers significant energy savings compared to mainstream models like OpenAI’s ChatGPT and Google’s BERT. Created by a team at the Chinese Academy of Sciences, SpikingBrain uses event-driven spiking neurons, mimicking the brain’s energy-efficient signaling. This approach requires only about 2% of the pre-training data used by traditional models, achieving comparable performance in language understanding and reasoning tasks. SpikingBrain’s efficiency is due to spike-based thresholds, enabling fast and sparse computation. For instance, it processed a one-million-token input 27 times faster than a comparable transformer model. The model’s ability to handle ultra-long sequences makes it suitable for applications in legal and medical document analysis, high-energy physics, and DNA sequence modeling.

Alibaba unveiled Qwen-3-Max-Preview, a 1-trillion-parameter large language model (LLM) on September 5, 2025, positioning itself as a top-tier contender in the global AI market. This model, designed for fast retrieval and ultra-long context windows, supports a 262,144-token context window and performs well across 100+ languages, particularly excelling in mixed Chinese-English use cases. Internal benchmarks show it outperforming Alibaba's previous models and competitors like Kimi K2 and Claude Opus 4 in technical tasks such as mathematical reasoning and code generation. Qwen-3-Max is optimized for enterprise use, with a focus on speed and scale, making it ideal for RAG pipelines, AI agents, and technical assistants. The model is part of Alibaba's RMB 380 billion (USD 52 billion) investment in AI over three years, aiming to lead the global AI race.