Upscale AI is a Santa Clara-based networking company building the switches, fabric, and open-standards networking stack required to scale modern GPU clusters for AI training and inference. Rather than adapting general-purpose datacenter networking, the company is rebuilding the network from first principles for the traffic patterns that large AI workloads generate.
The company's thesis is that AI clusters are bottlenecked as much by the network as by compute, and that an open, high-performance fabric can displace incumbents like Cisco and Broadcom for AI-specific deployments. Its product line targets the interconnect layer that links thousands of GPUs into coherent training and inference systems.
Upscale AI emerged from Auradine, the AI and blockchain infrastructure company, and is led by CEO Barun Kar with Rajiv Khemani as executive chairman. Both founders have deep networking-silicon pedigrees, with prior roles at companies such as Palo Alto Networks, Innovium, and Cavium, the latter two acquired by Marvell.
The company has scaled funding rapidly. After emerging from stealth with more than $100 million in seed financing, it closed an oversubscribed $200 million Series A in January 2026 led by Tiger Global, Premji Invest, and Xora Innovation, with participation from Intel Capital, Qualcomm Ventures, Mayfield, and StepStone, pushing total funding above $300 million and reaching unicorn status.
As a young company shipping its first commercial products in 2026, Upscale AI's traction is still early relative to entrenched networking incumbents, and adoption will depend on proving performance and openness at production scale.
The company is best suited to operators of large GPU clusters and AI infrastructure builders evaluating next-generation interconnect. It is not relevant to teams that simply consume cloud AI APIs and do not operate their own clusters.