3 Before 9: March 17, 2026

Nvidia](https://www.amazon.sg/s?k=nvidia+gpu&tag=aiinasia-22) Unveils Feynman Architecture and Ships Vera Rubin Samples at GTC 2026 Jensen Huang used Nvidia's annual GTC conference on 16 March to pull back the curtain on Feynman, the company's next-generation AI chip architecture slated for 2028. Feynman will feature an eighth-generation NVSwitch, a new CPU codenamed Rosa, and a 204Tbps network fabric designed for massive-scale inference clusters. Huang also confirmed that Vera Rubin, the platform succeeding Blackwell, has begun shipping samples to customers ahead of full production in the second half of this year. Nvidia claims Vera Rubin delivers five times the inference performance and ten times lower cost per token than Blackwell.

Alibaba is expected to launch an enterprise-focused AI agent product built on its Qwen foundation model as early as this week, according to Bloomberg. The tool, developed by the team behind workplace platform DingTalk, is designed to let businesses deploy autonomous assistants that can operate computers, navigate browsers, and execute multi-step workflows. DingTalk already serves roughly 700 million users with its existing Tongyi Qianwen-powered agent, giving Alibaba a ready-made distribution channel. The product includes built-in data protection safeguards and will gradually integrate with Taobao and Alipay.

DayOne Data Centres, the Singapore-headquartered operator spun out of China's GDS Holdings, is close to filing confidentially for a US initial public offering, Bloomberg reported on 16 March. The company is targeting a raise of roughly $5 billion at a valuation of up to $20 billion, with JPMorgan and Morgan Stanley leading the deal. DayOne currently operates 480 megawatts of data centre capacity across Hong Kong, Indonesia, Japan, Malaysia, and Singapore, with another 590 megawatts in development. The company closed a $2 billion Series C round led by Coatue in January.