Marvell's Process Node Gamble Positions It as AI Infrastructure Leader
By skipping generations of chip technology nearly a decade ago, Marvell Technology has built a commanding position in AI data center infrastructure, backed by major investments from Nvidia and custom chip deals with Google.
Nearly a decade ago, Marvell Technology Inc. made an unconventional decision about semiconductor manufacturing that has since shaped its trajectory in the booming AI infrastructure market. The company chose to abandon the traditional progression of process nodes—the generational steps chipmakers typically follow as they shrink transistor sizes—and instead leap ahead to cutting-edge technology.
In semiconductor manufacturing, nanometer designations refer to generations of chipmaking technology rather than literal transistor dimensions. Advancing to smaller nanometer nodes generally allows for higher transistor density and improvements in performance and power consumption. Historically, chipmakers have moved methodically through each successive generation, making the decision to skip levels a substantial gamble.
Will Chu, executive vice president and general manager of the Custom Cloud Solutions Business Unit at Marvell, described the strategic shift: "Back around 2018, 2019, we made a decision that if we wanted to do data infrastructure, we needed to be a leader and not a follower, and that meant jumping to the latest process node with the best technology. So, we skipped 7 nanometer. 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. We've been running that playbook, and it's really served as the foundation for our growth to be a leader in the cloud."
Optical solutions for AI architectures
Marvell has built its AI infrastructure strategy around serializer/deserializer (SerDes) technology, which enables high-speed, multi-gigabit data transmission across AI clusters, cloud data centers and enterprise networks. The company has leveraged this capability to address a fundamental challenge in scaling modern data centers: energy efficiency.
In September, Marvell introduced 2-nanometer optical interconnect technologies intended for AI scale-up architectures, which bind multiple computing and memory resources into unified systems. David Lazovsky, executive vice president and general manager of the Data Center Networking Business Group at Marvell, explained the significance of this shift: "Why interconnectivity with scale-up moving from copper to optics is so important is because processor-to-processor data transmission, roughly 85% of total data transmission, is happening inside of these 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. We're developing technologies that leverage optics and SerDes … that significantly reduce the total power consumption."
Marvell's efficiency-focused approach has also driven innovation in memory systems. During August, the company unveiled a Photonic Fabric Memory Module designed to create a shared-memory tier across AI infrastructure. This technology merges DDR5 memory, high-bandwidth memory caching and optical connectivity, allowing memory resources to be distributed across multiple AI systems.
Lazovsky noted the breakthrough: "For the first time ever, Marvell has developed optically interconnected high capacity, high bandwidth memory. There are two stacks of high bandwidth memory, and this supports a module with up to eight stacks of Double Data Rate 5. It provides you the ability to decouple memory capacity from compute for the first time."
Marvell's advances in memory and optical networking have solidified its standing in the AI infrastructure sector. Nvidia Corp. made a $2 billion investment in the company earlier this year, and Marvell secured an agreement to design custom AI chips for Google LLC.
https://www.youtube.com/embed/NnvOJN1rg6Y?feature=oembed
Chu characterized the market opportunity: "This is the biggest buildout in CapEx in human history. In terms of our connectivity business, our networking business, and our custom business, we're attached to the fastest growing part of those markets. We have the leading technology, we have the customer credibility, a track record of execution."