At the recent Hot Chips 2025 conference, Nvidia detailed its latest GB10 system-on-chip (SoC) architecture, representing a miniaturized application of the Blackwell GPU architecture. Through collaboration with MediaTek, Nvidia integrated 20 Arm v9.2 CPU cores with a high-performance GPU into a single 2.5D package, creating a compact version of the Grace Blackwell platform.
Quantum computing developer Rigetti Computing has projected that the technology will become commercially viable when it is integrated into existing data centers, delivering clear advantages over current CPU and GPU systems. The company expects this milestone to occur within three to five years, signaling a critical phase for quantum computing beyond research and development.
Wei Shaojun, vice chairman of the China Semiconductor Industry Association and professor at Tsinghua University, has called on Asian nations, particularly China, to cut their reliance on Nvidia GPUs for training and deploying artificial intelligence (AI) models. Speaking at a forum in Singapore, he urged countries to accelerate the development of homegrown AI chips.
Generative AI tools have provided much amusement for consumers, yet the same enthusiasm has not translated to the enterprise market. This stems from significant limitations in the large language models (LLM) behind generative AI applications, leading to highly fragmented AI agent applications, according to Taiwan industry observers, who suggest using a quadrant-like framework to define the necessity of AI tasks.
Potential uses for drones are ever-increasing, with Ukrainian President Volodymyr Zelensky once stating that 34% of Russia's bombers were destroyed using just 117 drones. This offers valuable experience for Taiwan, according to Teng-hu Cheng, chief robotics scientist at Tron Future: the 117 drones caused approximately US$7 billion in losses to Russia, while their cost was four orders of magnitude lower, demonstrating a massive leverage effect.
Intel is tightening its PC processor roadmap, confirming it will roll out an Arrow Lake Refresh in 2026 and follow with the next-generation Nova Lake platform by late 2026 or early 2027. The update was disclosed by Intel corporate vice president John Pitzer at the Goldman Sachs Communacopia + Technology Conference.
Taiwan's AI-driven semiconductor and server supply chains are growing rapidly, but experts say the wider economy and public benefit little. There are increasing calls for the government to spearhead AI expansion and explore novel ideas like converting AI tokens into a monetary system to boost inclusion.
Google's cloud computing service, Google Cloud, has become its most important growth engine. CEO Thomas Kurian recently stated at the Goldman Sachs Communacopia + Technology Conference held in San Francisco that the business had generated billions of dollars in revenue through artificial intelligence (AI) services and established diversified business models. Currently, it holds a backlog exceeding US$106 billion, with approximately 55% of these orders expected to convert into revenue within the next two years—about US$58 billion.
As generative AI matures, the traditional division of labor within enterprises is undergoing a profound transformation. At Taiwan Mobile's D.E.E.P. Tech Day 2025 on September 9, 2025, Chief Information Officer Rock Tsai introduced the concept of "reversing specialization," arguing that AI is not merely a productivity upgrade but a force that will redistribute corporate roles, reshape value chains, and accelerate the rise of an "AI agent" ecosystem in external markets.
Oracle is rapidly shaking off its reputation as a laggard in cloud computing, with new AI contracts and a strong outlook pushing its growth plans into overdrive.
Generative AI has moved beyond the "model arms race" into practical deployment, and Taiwan Mobile (TWM) is moving quickly to capitalize on the momentum. Company president Jamie Lin announced that the company will invest NT$930 million (approx. US$30.7 million) in 2025 on R&D to create solutions that support enterprises in AI adoption and deployment.
More coverage
