Chinese AI agent startup Manus is returning to independent operations, marking a concrete break from Meta months after Chinese regulators forced the US tech group to unwind its acquisition of the company.

Nvidia is intensifying its open-source strategy through its upcoming trillion-parameter model in the Nemotron 4 family, which is meant to compete with the world's leading open-source models, according to The Information. While this may put it in the awkward position of vying against its customers, the chipmaker's advocacy for open-source AI could see the company benefit from more chip sales across the board.

At the 2026 OCP APAC Summit, debate over networking architectures in cloud AI data centers sharpened as industry players said the main bottleneck is no longer compute; beyond memory, networking has become a key constraint on how far AI infrastructure can scale.

SpaceXAI has launched Grok Bot, an AI agent designed to take on workplace tasks and operate digital tools with limited human intervention, marking the company's latest move to push Grok beyond a conventional chatbot.

Amazon is moving ahead with a dedicated gas-fired power plant for a new data center campus in Pecos County, Texas, as part of an effort to speed AI infrastructure deployment outside the state grid. The project, called GW Ranch, is designed to support behind-the-meter self-generation and avoid the interconnection delays that have slowed other power-heavy facilities.

Siliconware Precision Industries (SPIL), a subsidiary of ASE Technology Holding (ASEH), broke ground on a new plant in Douliu, Yunlin County, Taiwan, amid surging global demand for artificial intelligence (AI) and high-performance computing (HPC). The plant will cover an area of approximately six hectares and is expected to introduce chip-on-wafer-on-substrate (CoWoS) advanced packaging technology, with a total investment of about NT$100 billion (approx. US$3.1 billion). The first phase is scheduled to begin operations in 2028.

IBM and Together AI have announced a collaboration to deploy a dedicated Nvidia-powered inference cluster on IBM Cloud, a move that could reshape how enterprises access open-source AI services worldwide. The planned system is designed to improve performance, lower token costs, and expand global access to scalable AI infrastructure as demand for real-time applications grows.

Taiwan's National Science and Technology Council (NSTC) announced the winners of its annual Future Tech Award on August 10, with the aim of encouraging the licensing and commercialization of breakthrough research projects. Among the selected technologies were Academia Sinica's 20-qubit superconducting quantum computer and quantum chip process development project, headed by distinguished research fellow Chii-dong Chen, and research into complementary field-effect transistors and other advanced stacked transistors, led by Chee Wee Liu of the Department of Electrical Engineering at National Taiwan University (NTU).

Surging demand for AI computing is accelerating server and data center infrastructure deployment. Speaking at Delta Electronics' Sustainable AI Summit on August 7, Wistron CTO Shen Ching-Yao said GPU lead times have shortened significantly. Orders now generally require around six months of advance planning.

AI-driven power demand is surging, and Delta Electronics Chairman Ping Cheng said microgrids and energy resilience are becoming the next key battleground for AI data centers and the broader electrification push. Speaking at the Delta Sustainable AI Summit, he said that, for the first time in 2026, global electrification and net-zero emissions will be discussed at the same level during UN climate talks, making Delta's business transformation path clearer.

As companies accelerate the adoption of AI agents to reshape business processes, attackers are using the same technology to increase the speed and scale of cyberattacks.