Market Overview

OpenAI announced on October 24, 2026, that it will no longer use live internet data for training its frontier models, citing "significant risks." This decision represents a major shift in OpenAI's data strategy, potentially impacting the models' real-time capabilities and knowledge boundaries.

Hong Kong Exchanges and Clearing Limited (HKEX) announced on October 23, 2026, the launch of the HKEX Artificial Intelligence Index and the HKEX Semiconductor Index. These new indices provide benchmarks for two key technology themes, expanding HKEX's proprietary index series to capture investment opportunities in the AI and semiconductor sectors.

The 2026 Global Developer Pioneer Conference and International Embodied Intelligence Skills Competition opened in Shanghai from October 23-24, showcasing thousands of robots and embodied AI technologies. At the event, Mifeng Technology launched the world's first full-category high-quality physical AI data crowdsourcing platform, "Mifeng Pie," while LinYan Medical presented its surgical world model product, aimed at providing intelligent support for surgical robots.

AI startup Flapping Airplanes announced on October 23, 2026, that it has completed a $1 billion funding round, valuing the company at $6 billion. The company focuses on researching how to train AI models with less data and computational power to improve chip efficiency, addressing industry concerns about the financial strain of AI compute expansion.

The U.S. National Aeronautics and Space Administration (NASA) announced new arrangements for its Artemis III mission on October 23, 2026. The mission will now conduct a low-Earth orbit rendezvous and docking test in 2027, while the original crewed moon landing objective has been postponed and re-designated to the Artemis IV mission in early 2028.

Moonshot AI released its 3-trillion-parameter open-source large language model, Kimi K3.1 Pro, on October 24, 2026, supporting 1 million context tokens and 3D creation. Concurrently, MiniMax open-sourced its M3.1 Flash model on October 23, which processed over 100 trillion tokens within 72 hours, demonstrating significant improvements in inference efficiency.