Marvell Technology, Inc., a leader in data infrastructure semiconductor solutions, today announced the launch of its custom Ultra Accelerator Link (UALink) scale-up solution, designed to meet the growing performance and scalability demands of next-generation AI systems.
Built on open standards, Marvell’s custom UALink offering delivers low-latency, high-efficiency interconnects for hundreds to thousands of AI accelerators—providing a critical scale-up infrastructure layer for hyperscale data centers.
“We are pleased to introduce our new custom UALink offering to enable the next generation of AI scale-up systems,” said Nick Kucharewski, SVP and GM, Cloud Platform Business Unit at Marvell. “This addition to our custom portfolio gives customers the flexibility to optimize AI infrastructure with standards-based interconnect technology.”
Purpose-Built for Scale-Up AI
The Marvell UALink solution incorporates a comprehensive suite of interoperable IPs, including:
- 224G SerDes and UALink Physical Layer IP
- Configurable UALink Controller IP
- Scalable, low-latency Switch Core and Fabric IP
- Advanced packaging options like co-packaged copper and co-packaged optics
Together, these technologies allow custom accelerators and switches to be tightly integrated with Marvell’s proven custom silicon capabilities, optimizing performance, latency, and efficiency at rack-scale.
Supporting the AI Era
As AI workloads become increasingly data-intensive and distributed, hyperscalers face mounting pressure to scale up efficiently while maintaining high throughput and low power consumption. Marvell’s custom UALink offering solves these challenges with:
- Direct, low-latency communication between accelerators
- Flexible switch topologies for seamless scaling
- Standards-based interoperability to reduce vendor lock-in and increase ecosystem agility
“We are excited to see UALink custom solutions from Marvell, which are essential to the future of AI,” said Forrest Norrod, EVP and GM, Data Center Solutions Group at AMD. “Open standards and ecosystem collaboration are the foundation for scalable, efficient AI infrastructure, and Marvell’s work helps push this forward.”
A Founding Member of the UALink Consortium
Marvell is a founding member of the UALink Consortium, an open industry group focused on defining specifications for high-performance interconnects between accelerators and related devices. The consortium’s mission is to accelerate interoperability, boost innovation in AI infrastructure, and drive transformative breakthroughs in computing by unifying industry leaders under open standards.
