Billion-dollar deal: Nvidia invests in Marvell

The AI industry continues to run hot. US chip giant Nvidia has invested two billion US dollars in Marvell Technology. The two companies are also entering into a strategic partnership. This is intended to make it easier for customers who build on Nvidia architectures to use the custom AI chips developed by Marvell in combination with Nvidia's network components and processors. Both companies announced this on Tuesday in identical statements.

The partnership will provide customers with “more choice and flexibility in developing next-generation infrastructures,” it says. Marvell will contribute custom chips and network solutions compatible with Nvidia's NVLink Fusion, while Nvidia will provide supporting technologies such as CPUs, network cards, and connections.

“The turning point in AI development has been reached. The demand for token generation is increasing rapidly, and AI factories are being developed worldwide with great urgency,” Nvidia CEO Jensen Huang is quoted as saying in the statement. “Together with Marvell, we are enabling our customers to leverage Nvidia's AI infrastructure ecosystem and scale specialized AI computing capabilities.”

The partnership with Nvidia underscores the growing importance of high-speed connections, optical connections, and accelerated infrastructure for scaling AI, explained Marvell CEO Matt Murphy. “By combining Marvell's leading expertise in high-performance analog electronics, optical digital signal processing (DSP), silicon photonics, and custom silicon technology with Nvidia's growing AI ecosystem through NVLink Fusion, we enable our customers to build scalable and efficient AI infrastructures.”

Scalability and energy efficiency increasingly important

Jacob Bourne, an analyst at market research firm eMarketer, points out to the news agency Reuters that scalability, bandwidth, and energy efficiency play a key role in AI systems and data centers. “The acquisition [of Marvell, editor's note] expands Nvidia's ecosystem with more specialized silicon chips, thereby strengthening Nvidia's position as a central access point for increasingly diverse AI workloads,” the expert said.

Both companies will also collaborate to advance advanced network solutions for AI, including optical interconnect solutions and silicon photonics technology that enable fast and energy-efficient data transmission, they announced.

Given the global AI boom, major tech corporations are currently investing billions in building AI infrastructure. Last fall, Nvidia announced together with OpenAI the “largest AI infrastructure project in history” and to invest a total of 100 billion US dollars in the ChatGPT developer. Several data centers were planned, which were to have an energy requirement of ten gigawatts, equivalent to the output of ten nuclear power plants. Meanwhile, Nvidia's OpenAI plans sound less ambitious. Although Nvidia CEO Huang denied media reports that the investment plans were on hold. But the originally mentioned sum of 100 billion US dollars now also appears unrealistic. Nvidia does want to continue to invest in OpenAI – but not 100 billion.

(akn)