AI News Feed
Market watch
Revolut alumni's Nevis raises $70m at $500m valuation as Revolut launches UK personal loansAnthropic Files for Nasdaq IPO Under Ticker ANT, Disclosing $16.3 Billion in Half-Year RevenueApple unveils OLED iPad Mini with A20 Pro chip as ZDNET report cites conflicting pricesSpaceX Secures $40 Billion GPU Financing From ApolloSony's WH-1000XM4C revives a beloved headphone design with older ANC techBuilding an Xbox Series X-equivalent PC costs about $1,200 with new parts
Revolut alumni's Nevis raises $70m at $500m valuation as Revolut launches UK personal loansAnthropic Files for Nasdaq IPO Under Ticker ANT, Disclosing $16.3 Billion in Half-Year RevenueApple unveils OLED iPad Mini with A20 Pro chip as ZDNET report cites conflicting pricesSpaceX Secures $40 Billion GPU Financing From ApolloSony's WH-1000XM4C revives a beloved headphone design with older ANC techBuilding an Xbox Series X-equivalent PC costs about $1,200 with new parts
AI Chips & Compute

Marvell Says Early Process-Node Bet Paid Off as AI Demand Grows

Marvell says its early leap to advanced process nodes helped it gain AI data-infrastructure business, as it rolls out optical interconnect and memory products.

AI News Feed · 2026-10-27 02:43

Will Chu, executive vice president and general manager of Marvell’s Custom Cloud Solutions Business Unit, said Marvell had products at 12 and 16 nanometers and the next logical step would have been 7 nanometers, but the company jumped to 5 and then continued to 3 nanometer, 2 nanometer and 14 angstrom. “We skipped 7 nanometer,” Chu told theCUBE. “We had products like 12 and 16 nanometer and the next logical one would have been 7. We jumped all the way to 5…we kept going – 3 nanometer, 2 nanometer, 14 angstrom.” He said the company has been running that playbook and that it served as the foundation for its growth in the cloud.

A cornerstone of Marvell’s platform is serializer/deserializer technology, which provides high-speed, multi-gigabit data transmission for AI clusters, cloud data centers and enterprise networking. In September, the company unveiled 2-nanometer optical interconnect technologies designed to support AI scale-up architectures, which tightly connect multiple computing and memory resources to operate as a larger system.

David Lazovsky, executive vice president and general manager of Marvell’s Data Center Networking Business Group, said processor-to-processor data transmission is roughly 85% inside scale-up networks and only 15% across the scale-out network. “If you can positively impact the energy efficiency, you can move the needle significantly,” Lazovsky told theCUBE. He said Marvell is developing technologies that leverage optics and serializer/deserializer technology to reduce total power consumption.

In August, Marvell announced a Photonic Fabric Memory Module designed to enable a shared-memory tier across AI infrastructure. The technology combines DDR5 memory, high-bandwidth memory caching and optical connectivity to allow memory resources to be shared across AI systems. Lazovsky said Marvell has developed optically interconnected high-capacity, high-bandwidth memory, with two stacks of high-bandwidth memory, and that the module supports up to eight stacks of DDR5. He said it provides the ability to decouple memory capacity from compute for the first time.

Nvidia Corp. announced a $2 billion investment in Marvell earlier this year, and Marvell recently announced an agreement to design custom AI chips for Google LLC, according to SiliconANGLE. Chu said the buildout is “the biggest buildout in CapEx in human history” and that Marvell’s connectivity, networking and custom businesses are attached to the fastest-growing parts of those markets. He cited leading technology, customer credibility and a track record of execution.