Arrcus taps Nvidia as both AI partner and investor

　　Virtualized routing and switching startup Arrcus has brought in $30 million in funding this week plus a new investor: chipmaker giant Nvidia. The companies are also teaming up to integrate Nvidia's AI infrastructure into Arrcus' ACE Platform.

　　Prosperity7 Ventures, Nvidia, Lightspeed, Hitachi Ventures, Liberty Global, Clear Ventures and General Catalyst participated in this funding round. Arrcus plans to use the new financial backing to expand platform support and increase sales and marketing.

　　Arrcus brought in $15 million in funding in July 2023, and with this latest addition the company's total funding to date is about $157 million.

　　"The goal of this investment is to essentially extend the Arrcus platform to become a common network operating system layer underneath the fabric of AI, for extended and distributed network infrastructure," Arrcus CEO Shekar Ayyar told Light Reading.

　　The AI effect

　　Arrcus's Connected Edge (ACE) platform is cloud-native and supports use cases such as spine-leaf and TOR switching, multicloud networking, 5G edge and cell-site routing, and core virtual distributed routing. Arrcus offers additional capabilities such as routing security, the ArcIQ analytics service and an Egress Cost Control (ECC) to automate the visibility and management of cloud data transfer fees.

　　"What is increasingly becoming evident is that as AI becomes more prevalent, the compute infrastructure and the processor infrastructure for AI needs to get a lot more distributed," said Ayyar.

　　Many enterprises likely use multiple racks, data centers, edge computing and inferencing infrastructure to support their AI applications – all of which can make the goal of efficiently managing AI workloads challenging, he explained.

　　"When you have that form of AI, you need efficient networking that can extend your core GPU infrastructure to the center or to the distributed architecture," said Ayyar.

　　Arrcus is utilizing Nvidia's BlueField DPUs within the Arrcus ACE platform to address enterprise interest in more efficient networking and AI infrastructure, and management of networking applications such as security and traffic engineering.

　　In today's AI infrastructures, many compute or processor cycles could be offloaded to an adjacent processor where they could perform more efficiently, said Ayyar. The Nvidia BlueField DPUs become a "triaging" point to support the offloading of these cycles for Arrcus' customers.

　　Value in DPUs

　　Nvidia's investment in Arrcus "shows that there is a lot of value to unlock in these DPUs and ASICs to offload functions on the traditional server cores and add additional functionality previously unavailable," Alan Weckel, founder and technology analyst for 650 Group, told Light Reading in an email. "Arrcus highlights security and traffic engineering, but many use cases are being discovered where you can change or add how services are consumed in the network."

　　In addition, the technological collaboration between the two companies could contribute to better AI performance for customers in the near term, explained Weckel. In the future, "we will see additional telemetry pre/post-processing and additional security to improve the performance of AI and lower TCO," he said.

　　Arrcus' service provider customers and investors, such as Liberty Global, can also benefit from "the lower TCO of a modern, more agile network" and the opportunity to add new revenue streams to improve the customer experience for their own enterprise customers, said Weckel.