One big selling point of Rubin is dramatically lower AI inference costs. Compared to Nvidia's last-gen Blackwell platform, inference workloads on Rubin can be run at a 90% lower cost per token. Tokens are units of data processed by AI models, and it's how customers of those models are generally charged for use.

When it comes to artificial intelligence (AI) infrastructure, there has been no company as dominant as Nvidia (NASDAQ: NVDA). Its graphics processing units (GPUs) have become the primary chips used to power AI workloads, and the ecosystem the company has built around its chips has created a nice moat.

As the company explains it, the new Rubin line of chips are built to treat the AI data center as the "unit of compute," not just a single GPU server. The platform includes six chips that are built to work as one, with GPUs, CPUs, and other components co-designed to share data faster.

The Chinese government this week told some tech companies it would only approve their purchases of Nvidia's H200 AI chips under special circumstances, such as for university research, the Information reported on Tuesday,

AI chip startup Etched raised about $500 million in a new funding round, according to people familiar with the matter, part of an effort to compete with Nvidia Corp. in the booming market for artificial intelligence processors.

Nvidia stock was languishing Monday despite news of a partnership with pharmaceutical company Eli Lilly. The chip company’s investors remain focused progress on sales in China and the health of the wider artificial-intelligence sector. The chip maker’s shares were flat $184.94 on Monday. The stock fell 0.1% in Friday’s trading session.

Nvidia still dominates the AI training landscape, but it's now going toe-to-toe with AMD and even some of its own customers in chip design. Most importantly, Nvidia's evolution from chip supplier to end-to-end platform provider will prove a crucial long-term growth engine.