3 Before 9: March 17, 2026

##1. Nvidia 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.

Why it matters: Feynman will be manufactured on TSMC's 1.6nm A16 process, marking Nvidia's first use of a sub-2nm node and deepening its dependence on Taiwan's chip fabrication ecosystem. Samsung Electronics and SK hynix are confirmed as key HBM4 memory suppliers for the Vera Rubin platform, cementing South Korea's role in the AI silicon supply chain. For enterprise buyers across Asia, Vera Rubin's inference economics could materially lower the cost of deploying large language models at scale.

##2. Alibaba Prepares Enterprise AI Agent Powered by Qwen

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.

Why it matters: China's enterprise AI agent market is heating up quickly, and Alibaba's aggressive pricing strategy, with model costs cut by up to 97 per cent compared to US rivals, could set a new floor for agentic AI in the region. For Southeast Asian businesses already embedded in Alibaba's cloud and commerce ecosystem, this lowers the barrier to deploying autonomous AI workflows. The move also signals that the commercial battleground in Chinese AI has decisively shifted from model benchmarks to practical enterprise tooling.

##3. Singapore's DayOne Data Centres Nears $5bn US IPO at $20bn Valuation

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.

Why it matters: If the IPO proceeds at the reported valuation, DayOne would become one of the largest Asia-focused data centre listings to date, validating AI infrastructure as a standalone asset class in the region. The company's footprint spans the key markets where hyperscalers are racing to build out inference and training capacity. For enterprise buyers and cloud customers in Southeast Asia and North Asia, DayOne's expansion pipeline signals that regional compute supply is scaling to meet surging AI workload demand.