AI Industry Development & Policy Dynamics

According to Anthropic's IPO roadshow disclosure, Claude Mythos completed GPU kernel migration for AMD in 3 days, enabling a Claude model to run key computational workloads on AMD MI355X. The GPU kernels were almost entirely migrated or written by Claude itself, with minimal human intervention. Ant

According to PANews on November 11, Baidu released ERNIE 5.5 Flash Preview, a post-training version based on its previously released Base model. The model features 280B total parameters and 14B active parameters, supporting 1M context window. It demonstrates strong performance in programming, 3D mod

Microsoft has been pressuring the Trump administration to abandon plans to ban imports of new models of optical transceivers from China, according to three sources cited by Reuters in reports covered by Hong Kong media. The optical transceivers are critical components for AI data centers that

According to Reuters, citing three sources, Microsoft has been pressuring the Trump administration on November 11 to abandon plans for banning imports of new optical transceivers from China. Optical transceivers are critical components for AI data centers that transmit data at light speed through

Teco Electric & Machinery, GMI Cloud, and Kaishin Data Infrastructure Corporation (KSDI) are collaborating to construct a 6.4MW AI modular data center (MDC) in Satsumasendai City, Kagoshima Prefecture, Japan, with a 9-month construction timeline. Teco will serve as the MDC contractor, responsible

On Tuesday, SoftBank reported a Q2 net loss of 1.12 trillion yen ($7.1 billion), primarily due to non-refundable prepayments for U.S. data center power equipment, along with a $370 million impairment charge. The stock plunged 11% at the open but recovered, rising 0.8% by afternoon as investor

According to ChainCatcher, Huawei's Kirin 9050 series chip features a new-generation Davinci architecture NPU that achieves a 140% performance improvement and 66% power consumption reduction. The NPU supports running 30-billion-parameter large models directly on-device without requiring internet con

According to Google DeepMind, the company rolled out Gemini Robotics 2.5, a vision-language-action model for robotics, after testing it on over 200 tasks. In a proof-of-concept deployment, Apollo 2 humanoid robots from Apptronik sorted automotive parts on Mercedes-Benz production lines in Germany an

New York University and University of Calgary researchers released public code for Deletion Tuning, a post-training method, in support of their NeurIPS 2026 paper. The method learns which pretrained LLM weights to delete, and the paper reports that zeroing fewer than 0.05% of a model's weights, with

According to Dassault Systèmes, on November 11 the French 3D design and engineering software maker signed a memorandum of understanding with Singapore's Economic Development Board to establish an ASEAN Industrial AI Virtual Twin Lab in Singapore. The lab will serve as a regional hub enabling businesses to simulate data center cooling, energy use, and industrial operations changes before deployment. Dassault Systèmes plans to employ more than 30 AI professionals at the lab within its first five y

According to PANews, Gonka, a decentralized AI compute network, processed 3.5 trillion AI tokens in October 2026, marking a significant increase from over 1 trillion in September, with a daily average of approximately 113 billion tokens. The Gonka Foundation has completed legal registration and

According to CNBC, U.S. Treasury Secretary Scott Bessent announced Tuesday that Washington and Beijing have established a new dialogue mechanism on artificial intelligence, or "super intelligence" (SI), aimed at managing shared risks and preventing misunderstandings between the two nations. The mec

Taiwan's Cathay Philadelphia Semiconductor ETF (00830), the only domestic fund tracking the Philadelphia Semiconductor Index (SOX), attracted strong buying interest as the SOX fell below 10,000 points. The ETF's stock price, which surged past 100 TWD earlier this year amid AI momentum, drew

Alibaba Cloud and the Dubai Chamber of Digital Economy signed a memorandum of understanding on AI agent adoption for Dubai's private businesses during the week of November 1. The agreement was signed at the Dubai Business Forum in Shenzhen, according to GCC Business News, and covers AI agents, which

According to IT之家, Alibaba released Qwen Code 1.0 on November 11, merging Qoder and Qwen Code as open source. The new version integrates Harness capabilities from Qoder into Qwen Code, available as a Preview release, providing developers with an embeddable Agent runtime foundation. Qwen Code 1.0

Meta announced Muse Ember, an efficiency-focused AI model, on November 9, 2026 (US time). The model is available through Muse Code, the Meta Model API, Google Cloud, and Oracle Cloud. Meta designed it for tasks that AI agents run repeatedly, including tool calling, coding subtasks, routing, extracti

Meta announced on Monday that its Muse AI agent is now available on the Apple Watch and Android's Wear OS, allowing users to interact with the assistant through voice calls, text messages, or dictation directly on their smartwatch. The smartwatch launch follows Meta's introduction of a dedicated iPa

Microsoft has been named a Leader in the 2026 Gartner Magic Quadrant for CRM Customer Engagement Center. The recognition, referenced in a Gartner report dated November 2, 2026, reflects the results customers are achieving with Dynamics 365 Customer Service, which Microsoft is strengthening alongside

Alibaba's Qwen team announced on Nov 11, 2026 that Qwen4-Max is now in preview on Alibaba's Token Plan. The post describes Qwen4 as offering deeper knowledge, richer world understanding, and stronger coding and agentic capabilities, built for real-world tasks. Early access to the new model is also a

Google Cloud has launched the Gemini factory, a co-innovation initiative for enterprise software partners. The unified Gemini experience lets enterprises bring third-party partner capabilities directly into Gemini through standardized connectors, including Model Context Protocol (MCP) connectors, Ag