Upscale AI is part of a new generation of AI infrastructure startups addressing a key challenge in data centers. The three-year-old startup has the backing of top investors to further develop its disruptive AI-scaling technology.
Upscale AI, which is tackling networking, has raised an oversubscribed $200 million Series A, bringing its total funding to over $300 million. The company reached unicorn status just months after raising a $100M seed round. It is a long way from incubation by Auradine, an infrastructure player in the blockchain and AI compute space.
Serial entrepreneurs Barun Kar (CEO) and Rajiv Khemani (Executive Chairman) launched Upscale AI in 2022. They assembled a world-class founding team comprising over 100 influential technologists. Their mission was to introduce a disruptive technology for scaling AI.
The two gained early investor support by successfully co-founding ventures such as Palo Alto Networks, Innovium, and Cavium.
Tiger Global, Premji Invest, and Xora Innovation led the current round, with additional participation from Maverick Silicon, StepStone Group, Mayfield, Prosperity7 Ventures, Intel Capital, and Qualcomm Ventures.
Upscale AI’s Solution and Why This Matters Beyond the Data Center
Data center operators currently face suboptimal choices to scale: retrofitting legacy networks, adopting proprietary solutions, or accepting inefficient approaches that raise costs and reduce performance. Upscale AI is offering an alternative.
The company seeks to replace legacy technology designed for general computing and storage.
Upscale’s AI strategy is based on its purpose-built SkyHammer™scale-up solution. A unified rack is achieved by collapsing the distance between accelerators, memory, and storage, transforming the entire stack into one cohesive, synchronized system. By unifying GPUs, AI accelerators, memory, storage, and networking, a single, synchronized AI engine is achieved.
The co-founders are committed to open standards and open source for enhanced networking that brings key benefits to operators:
- Reduced time-to-deployment for new AI capabilities
- improved utilization of costly accelerators, with less time spent waiting for data movement
- lower infrastructure costs per unit of AI output for enterprises beyond the largest hyperscalers.
To advance its mission, the company actively participates in developing these standards with organizations such as the Ultra Accelerator Link Consortium, Ultra Ethernet Consortium, Open Compute Project, and SONiC Foundation. Its goal is to “shape how AI systems scale with openness, interoperability, and performance at the core”.
Industry analysts bolstered their plan, in a recent press quote indicating that the category could grow significantly. One projected AI networking as having a potential $100 billion annual market by the end of the decade. Another noted the market has reached a clear “inflection point.”
A View of the Future
Upscale AI is now expanding engineering, sales, and operations as it moves toward commercial deployment.
If milestones and expectations are met, Upscale AI’s Series A will represent more than a major funding event. It suggests that the next phase of AI competition may be powered by advances in networking.
