With artificial intelligence rapidly transforming industries, collaborations have become an integral support by creating viable solution ecosystems.

From its custom accelerators to open ecosystems, Broadcom Ltd. is extending its networking chops to enable next-gen AI technologies in partnership with Dell Technologies Inc.

Discussing the future of AI networking with Broadcom’s Jas Tremblay.

“There are two parts of our AI strategy,” said Jas Tremblay (pictured), general manager of the Data Center Solutions Group at Broadcom Ltd. “One part is we do custom accelerators for hyperscalers. The other part is we do connectivity for merchant silicon — so, scale-up fabrics, scale-out fabrics, front end, back end, the whole interconnect piece from switching silicon all the way to optics. What customers are looking for is the complete cluster. They turn to Dell for that, and we hand-help Dell with all the connectivity pieces.”

Tremblay spoke with theCUBE Research’s Dave Vellante and Savannah Peterson at Dell Technologies World, during an exclusive broadcast on theCUBE, SiliconANGLE Media’s livestreaming studio. They discussed Broadcom’s approach to ensuring that the infrastructure needed for the AI era is robust, efficient and ready to support the next wave of technological breakthroughs. (* Disclosure below.)

Networking’s critical role in AI

Broadcom’s networking solutions are essential for handling the massive AI workloads that span multiple data centers. These workloads, often referred to as “elephant workloads,” require sophisticated interconnectivity across thousands of GPUs, according to Tremblay.

“You need to have all that interconnected,” he said. “The industry is standardizing on a way to interconnect that. It starts with your AI server that has a dozen GPUs, a dozen NICs and dozens of NVMe drives. And you need to build an internal fabric for that, which is based on PCIe, and we participate there.”

Broadcom’s role is to provide the necessary networking technology, from internal fabrics based on PCIe to ethernet-based scale-out networks. This connectivity is crucial for the performance and efficiency of AI data centers.

“It sounds simple when I explain it, but there’s a lot of networking technology that you need to build this up,” Tremblay said. “Going from a compute server to an AI data center world, the complexity and scope of connectivity goes up exponentially.”

Here’s the complete video interview, part of SiliconANGLE’s and theCUBE Research’s coverage of Dell Technologies World: 

(* Disclosure: Broadcom Ltd. sponsored this segment of theCUBE. Neither Broadcom nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.)

Photo: SiliconANGLE