AI Industry Development & Policy Dynamics

According to Caixin, on July 22, Nvidia CEO Jensen Huang said the U.S. should not restrict Chinese AI models, stating that "excellent open-source models should be used." Huang added that policymakers have previously misread the impact of both DeepSeek and

Moonshot AI's flagship open-source model Kimi K3 is now available on B.AI Web Chat as of today (July 23). The move lowers the barrier to entry following B.AI's earlier API integration. Users can now experience the first open-source 3-trillion-parameter model's capabilities in long-context

Alphabet reported Q2 revenue of $119.8 billion on local time May 22, a 24% increase from the prior year, and raised its fiscal year 2026 capital expenditure guidance to $195-205 billion to support AI infrastructure expansion. The announcement triggered a 4.40% surge in the KOSPI index on May 23,

According to Beating monitor, Anthropic has made Claude Security's deep scanning capability available as a Claude Code plugin in beta, accessible to all Claude Code paying users. The hosted version of Claude Security remains limited to Enterprise customers. Users can scan entire code

OpenAI launched OpenAI Presence, an enterprise platform designed to deploy AI agents for mission-critical operational tasks. The platform enables organizations to build intelligent systems that answer customer questions, resolve technical issues, interact with proprietary software, execute

Axe Compute (NASDAQ: AGPU) has signed new customer contracts totaling more than $1.3 billion in the U.S. and Europe, surpassing its full-year 2026 target of $1 billion. Under the Axe Compute Build program, the company will design, deploy, own, and operate dedicated AI infrastructure for customers, w

According to CCTV Finance, China's research team unveiled a new brain-computer interface device yesterday afternoon (July 23), marking the world's first cross-region simultaneous brain signal collection from over 1,000 people. The breakthrough enables neural foundation model training and brain-compu

OpenAI launched Presence, an enterprise-grade platform for deploying AI agents on mission-critical tasks, today (July 23). The platform enables organizations to build intelligent systems for handling customer inquiries, resolving technical issues, and executing approved actions while maintaining

According to BlockBeats, Tencent Cloud released CodeBuddy NPC, its first cloud-based AI agent for research and development workflows, on July 23. The agent, an upgrade to Tencent Cloud's CodeBuddy code assistant built around Harness (runtime governance layer for cloud-based AI agents), is designed f

According to Amap, the Chinese mapping platform upgraded its ABot software stack for embodied AI robots on July 22, adding five components for navigation, manipulation, reasoning, memory, and motion control. The latest release includes navigation model ABot-N1, manipulation model ABot-M0.5, reasonin

According to Business Insider, Amazon is overhauling Alexa+ infrastructure by routing more requests to proprietary AI models and reducing reliance on Anthropic's Claude through caching mechanisms to optimize GPU utilization and cut operational expenses. Internal documents show Alexa+ AWS cloud

SK Telecom established a new subsidiary, SK Hyper, following a board decision on the 23rd to accelerate its AI data center business. The South Korean telecom operator approved an investment of 750 billion won through 2030 to build business foundations for the new entity. SK Hyper, wholly owned by SK

According to Yonhapinfomax, SK Telecom established a new subsidiary, SK Hyper, on July 23 to accelerate its AI data center business. The company's board approved a plan to invest 750 billion Korean won through 2030 to develop the subsidiary. SK Hyper will focus on securing land, building and

According to DeepSeek founder Liang Wenifeng in an investor call on July 23, the company has partnered with Huawei and received 16,000 Huawei 950 chips for ecosystem validation. Liang stated that Huawei's 950 supernode can fully replace Nvidia's GB200 and GB300 in both performance and

According to Beating, Tencent Hunyuan multimodal understanding lead Hu Han has resigned to start a venture. Former OpenAI researcher Tian Yonglong will take over visual language model (VLM) research, continuing to report to Yao Shunyu. Hu Han joined Tencent from Microsoft Research Asia in

According to Guru Club, Beijing's development authorities announced policy measures on July 23 to accelerate smart agent-led innovation and foster a new intelligent economy. The city will encourage Token economy development and cultivate new business models including Token-as-a-Service (TaaS),

According to Guru Club, economist Liu Yuhui warned on July 23 that the AI industry faces a structural imbalance: semiconductor producers enjoy soaring margins while downstream cloud giants are burning through cash. China's competitive advantage lies in superior industrial capacity and engineering

According to Beating, developers have discovered that OpenAI's ChatGPT Work agent—designed to handle files and execute tasks over several hours—runs a Linux cloud environment that can be remotely accessed. Pro users gain access to a cloud setup with 9 CPU cores, approximately 21GB of memory, and 63G

Alphabet raised its full-year 2026 capital expenditure guidance to $195-205 billion in its second-quarter earnings, driven primarily by AI infrastructure investments. The company's second-quarter capital expenditure reached $44.9 billion. Google Cloud added over $50 billion in new backlog orders,

According to K25.ai's official announcement on July 23, the AI-native prediction market platform K25.ai completed its Series A funding round with strategic support from Amber Group, achieving a post-investment valuation of $200 million. The valuation has doubled in less than 60 days. Amber Group