Upscale AI Secures $200M in Series A Funding
Upscale AI, Inc. did not arrive quietly. It landed with the kind of thud that rattles racks and wakes up legacy networking vendors in a cold sweat. Santa Clara based, born in September 2025 after incubating inside Auradine, Upscale AI stepped out of stealth with a clear message. AI workloads are breaking old networks, and polite fixes are not enough. On January 21, 2026, the market answered with a $200 million Series A and a valuation north of $1 billion, four months after launch. Unicorn status, yes, but earned with intent, not volume.
Barun Kar and Rajiv Khemani did not wander into this problem. They have been here before, building companies like Palo Alto Networks, Innovium, and Cavium, learning exactly where infrastructure snaps under real pressure. Barun Kar, Chief Executive Officer, and Rajiv Khemani, Executive Chairman and also Chief Executive Officer of Auradine, are not chasing trends. They are responding to physics. AI training and inference do not care about legacy architectures designed for polite traffic and predictable flows. They demand synchronization, brutal speed, and zero patience.
Upscale AI builds networking the way AI actually behaves. Full stack, turnkey, spanning silicon, systems, and software. SkyHammer is not a metaphor, it is the job description. The platform collapses distance between GPUs, accelerators, memory, storage, and networking until the rack behaves like a single, synchronized engine. Ultra low latency is not a feature, it is survival. This is scale up architecture built for racks that think together or fail together.
Investors did not miss the signal. Tiger Global, Premji Invest, and Xora Innovation led the Series A, joined by Maverick Silicon, StepStone Group, Mayfield, Prosperity7 Ventures, Intel Capital, and Qualcomm Ventures. This comes after a $100 million seed round announced in September 2025. Over $300 million in total funding, deployed toward commercial shipment in 2026, rapid team expansion, and execution, not theater.
Upscale AI is also betting on openness in a market addicted to lock in. Active across Ultra Accelerator Link, Ultra Ethernet, SONiC, Open Compute, and the Switch Abstraction Interface, the company is pushing standards forward instead of wrapping them in proprietary fog. AMD has already publicly endorsed early adoption of UALink. Hyperscalers are paying attention, even if names stay quiet for now.
The AI networking market is heading toward $100 billion by decade’s end. Upscale AI is not trying to own the future by shouting. It is doing it by engineering the present until the old systems can no longer keep up, and that pressure is only starting to build.