Networking switch
– Thinkstock / kynny

Networking startup Upscale AI raised some $200 million as it seeks to become one of the first firms to launch an Ultra-Accelerator Link (UALink) switch.

Venture capital arms from Intel and Qualcomm were among those backing the California-based startup. The Series A round was led by Tiger Global, Xora Innovation, and Premji Invest.

Upscale’s appropriately named SkyHammer line of switches are designed to supercharge scale-up performance for AI workloads, leveraging innovations from industry protocols like the UALink, Ultra Ethernet (UEC), and the Microsoft-developed Software for Open Networking in the Cloud (SONiC) open source platform.

“Upscale AI has built extraordinary momentum in an exceptionally short time,” Upscale AI Executive Chairman Rajiv Khemani touted in a statement. “The market is demanding open, scalable AI networking solutions, and Upscale AI is uniquely positioned to help customers break through today’s networking constraints.”

The startup was founded last year by Khemani – an industry veteran with time at Intel, NetApp, and Cavium under his belt – alongside Juniper Networks and Palo Alto Networks alum Barun Kar.

Prior to Upscale, the pair founded Auradine, which sells servers specially designed for bitcoin mining.

Their latest endeavor looks to break the bottlenecks in AI infrastructure, providing high-speed, low-latency connections between GPUs, accelerators, memory, and storage.

Upscale AI's SkyHammer switch
– Upscale AI

Upon unveiling its SkyHammer design last October, the startup claims it is designed to power the world’s largest AI clusters, arguing that traditional front-end switches, PCIe fabrics, and proprietary topologies “strain under the demands of explosive AI growth.”

Central to the startup’s approach to switch design is interoperability, aiming to take advantage of cluster operators wanting to avoid vendor lock-in by making its alternative switch offering open as well as scalable.

The startup claims it has seen “strong early traction” for its solutions from both hyperscalers and AI infrastructure operators, though no names were disclosed at the time of writing.

It plans to use the funds to scale up its engineering, sales, and operations teams ahead of plans to launch commercial deployments.

“This investment accelerates our mission to fundamentally re-architect networking for the AI era,” said Kar, who serves as Upscale’s CEO. “With a world-class team and strong customer pull, we have a once-in-a-generation opportunity to build the open AI networking platform the industry has been waiting for.”