Latest update of this week

  1. Google brings Gemini AI deeper into the legal industry

Google launched Gemini Enterprise for Legal on August 25. The platform adds AI agents for legal research, administration and complex professional tasks. It connects with platforms from Thomson Reuters, Harvey and LexisNexis. Google also plans similar industry tools for financial services.

  1. New physics AI model moves beyond language models

Accelerated Understanding unveiled an AI system designed around physics rather than language. The company says its neural-operator architecture handled five trillion pieces of data in one query during testing. Possible applications include chip design, robotics, weather forecasting and energy exploration.

  1. Nvidia faces a major test as Vera Rubin approaches

Nvidia will report second-quarter earnings on August 26. Analysts expect revenue near $92.18 billion, almost double last year. Attention is shifting toward its next-generation Vera Rubin chips and whether AI infrastructure spending can maintain its current pace.

  1. Xiaomi launches its second major in-house smartphone chip

Xiaomi introduced the Xring O3 processor on August 24. Sources told Reuters that TSMC will manufacture it using a 3-nanometre process. The chip is expected to power an upcoming Xiaomi foldable phone as the company reduces dependence on external chip suppliers.

  1. Xpeng raises more than $900 million for humanoid robots

Xpeng's robotics business raised over $900 million in its first funding round. The deal values the unit above $6.3 billion. Xpeng plans to develop physical AI models, expand manufacturing and begin commercial robot deliveries in China and overseas markets during 2027.

  1. Unitree's dramatic share fall raises robotics bubble concerns

Shares of Chinese humanoid robot maker Unitree have fallen around 45 percent from their post-listing peak. The decline followed a fivefold surge during its Shanghai market debut. The volatility has raised questions about whether investor expectations for humanoid robotics are moving faster than commercial fundamentals.

  1. Alibaba plans $10.2 billion share sale for AI investment

Alibaba launched a $10.2 billion share placement to finance AI chips, infrastructure and model development. Its Hong Kong shares initially dropped sharply because investors worried about dilution and execution risk. The offering still attracted about $28 billion in orders, according to Reuters sources.

  1. China AI chipmaker Enflame moves toward major Shanghai IPO

Tencent-backed Enflame Technology plans an IPO worth roughly $892 million on Shanghai's STAR Market. The company intends to use the proceeds for fifth and sixth-generation AI chips and related software. The listing reflects China's continuing effort to build a domestic semiconductor industry.

  1. Britain and Ukraine establish AI defence partnership

Britain and Ukraine signed an agreement to develop AI tools for defence and security. British researchers will gain access to Ukraine's battlefield AI platform containing millions of annotated images. Planned projects include AI sensors, autonomous systems and lower-power chips for drones.

  1. OpenAI slows frontier training after cybersecurity concerns

OpenAI disclosed that it temporarily slowed advanced model development after the OpenAI-Hugging Face incident and signs that an upcoming model could reach a high cybersecurity capability threshold. The company paused a major reinforcement-learning run while strengthening monitoring, alignment and research-environment security.

  1. Google offers eligible students one year of AI access

Google announced one-year AI plan offers for eligible college students. U.S. students can receive Google AI Pro, while eligible students across many international markets can receive Google AI Plus. Google also introduced study notebooks, interactive learning tools and Deep Research inside Gemini Live.

  1. Google's Gemma models pass one billion downloads

Google DeepMind said the Gemma open-model family has passed one billion downloads. Developers have created more than 100,000 model variants. Google highlighted deployments ranging from edge devices and healthcare systems to AI workloads running aboard satellites.