Marvell Technology, Inc. has introduced an advanced multi-die packaging platform designed to optimize custom AI accelerator silicon for data centers. This innovative solution enables larger, more efficient chiplet designs, reducing power consumption and total cost of ownership while offering supply chain flexibility. Already qualified with a major hyperscaler, the platform is now entering production to support tailored AI solutions.
Marvell’s multi-die packaging platform addresses the growing demands of AI infrastructure by enabling chiplet designs 2.8 times larger than conventional single-die solutions. The platform integrates 1390 mm² of silicon with four HBM3/3E memory stacks, utilizing six re-distribution layer (RDL) interposers. This approach enhances compute density, improves signal integrity, and reduces power consumption, making it ideal for high-performance AI clusters and cloud environments.
Unlike traditional silicon interposers, Marvell’s modular RDL interposer is form-fitted to individual computing dies, connected by high-bandwidth paths. “Advanced packaging is one of the primary vehicles for advancing compute density in AI clusters and cloud,” said Will Chu, senior vice president and general manager of Custom Cloud Solutions at Marvell. This design reduces material usage, increases chiplet yields, and allows manufacturers to replace individual dies, lowering costs and enhancing supply chain flexibility.
Marvell’s platform supports complex AI designs through collaboration with industry leaders. “Leading-edge packaging technologies are critical to the adoption of chiplet architectures in current and future generations of AI and accelerated compute devices,” said Dr. Mike Hung, senior vice president at Advanced Semiconductor Engineering (ASE). The platform integrates passive devices to minimize signal noise and supports HBM3, HBM3E, and future HBM4 memory, ensuring scalability for next-generation AI applications.
“2.5D packaging technology continues to modernize heterogeneous IC packaging, enabling high-performance, cost-effective integration of multiple chiplet and memory modules,” said Kevin Engel, chief operating officer at Amkor Technology. By combining this platform with Marvell’s custom HBM and co-packaged optics solutions, the company is building a comprehensive ecosystem for custom XPU designs, positioning it to meet the evolving needs of hyperscalers and AI-driven data centers.
Marvell’s advanced packaging platform marks a significant leap in AI infrastructure, enabling hyperscalers to deploy efficient, scalable, and cost-effective custom accelerators. As the industry shifts toward chiplet-based architectures, this solution sets a new standard for performance and innovation in data center computing.
To deliver the data infrastructure technology that connects the world, we're building solutions on the most powerful foundation: our partnerships with our customers. Trusted by the world's leading technology companies for over 25 years, we move, store, process and secure the world's data with semiconductor solutions designed for our customers' current needs and future ambitions. Through a process of deep collaboration and transparency, we're ultimately changing the way tomorrow's enterprise, cloud, automotive, and carrier architectures transform for the better.
TechIntelPro empowers B2B technology decision-makers with the latest industry trends, deep insights, leadership perspectives and actionable intelligence to drive both organizational and professional milestones—and thrive on the tech pulse.