AI networking company Upscale AI has launched with a $100 million seed round.

The round was co-led by Mayfield and Maverick Silicon, with participation from the likes of Qualcomm Ventures and Stanford University. The funding is designed to support Upscale AI's efforts in developing a full-stack AI networking infrastructure.

The firm focuses on specialized xPU clusters, secure, energy-efficient systems designed to link large-scale, heterogeneous compute environments with ultra-low latency.

Underlying the build of its network operating system (NOS) are open-source technologies and standards. These include the Linux-based software for open networking in the cloud (SONiC), ultra accelerator link (UAL), which is designed to support scale-up interconnects for next-generation AI and high-performance computing (HPC) workloads, alongside Ultra Ethernet (UE), and switch abstraction interface (SAI).

With the funding, Upscale will focus on xPU maximization and target what it described as the $20 billion-plus AI networking market, catering to AI training, inference workloads, edge computing, and hyperscale cloud deployments.

The firm aims to reduce the total cost of ownership (TCO) at the data center level and introduce “disruptive” AI networking rack platforms designed for end-to-end multivendor networking.

From blockchains to networks

Upscale AI was founded this year by CEO Barun Kar and executive chairman Rajiv Khemani. The company was spun out of blockchain and AI player Auradine, also founded by the pair.

Kar founded Auradine after a 15-year stint as SVP for Engineering at Palo Alto Networks, with earlier experience including Juniper Networks and Motorola. Khemani, meanwhile, has served in C-suite roles at the likes of the Marvell-acquired Innovium and Cavium businesses, alongside management tenures at Intel, NetApp, and Sun Microsystems.

On the day of the funding announcement, Upscale AI’s senior director of product and technology, Rohit Mittal, spoke at a DCD Compute 2025 panel with SDxCentral in attendance.

Mittal told the Compute audience that the global AI market was currently driven by tokenomics, with the U.S. market more capitalist than others, as American firms work out how to provide clients with “the cheapest tokens without impacting the user experience”.

Described as a founding member of Upscale, Mittal also shared his experience that AI build-outs work better targeting one strength instead of many, advising Europe to concentrate on its current leadership in sustainability and “build your workforce for low power, and build the story from there.”