Upscale AI, a startup specializing in data center networking tailored for AI compute environments, has raised an additional $190 million in funding, extending its Series A round to $390 million and valuing the company at $2 billion. The investment will help accelerate commercialization of their proprietary AI-focused networking hardware and software.
- Raised $190M extension, reaching $2B valuation
- Develops AI-optimized scale-up and scale-out switches
- Supported by major strategic investors including Nvidia
Market signal
Upscale AI’s latest $190 million funding injection extends its prior $200 million Series A, underscoring strong investor confidence in AI-centric networking solutions. This new capital arrives amid growing recognition of network latency as a critical bottleneck in scaling advanced AI workloads, driving demand for specialized hardware tailored to machine learning operations.
Backers include major tech stakeholders such as Nvidia and Salesforce Ventures, highlighting strategic alignment between AI chip makers and data center networking innovations. The company’s $2 billion valuation signals a significant market opportunity within enterprise data center infrastructure that supports complex AI applications requiring ultra-low and deterministic latency.
Operator impact
Upscale AI targets data center operators and cloud providers dealing with intensive GPU-to-GPU communications. Their scale-up switches incorporate custom SkyHammer chips that guarantee predictable latency by supporting AI-optimized protocols such as UALink and ESUN, enabling GPUs to directly access one another’s memory. This reduces computation stalls from latency spikes, improving overall AI model throughput and efficiency.
For operators deploying large racks of GPUs, Upscale AI’s scale-out switches based on Nvidia’s Spectrum-X silicon offer enhanced traffic management by bypassing CPUs via RoCE technology, reducing latency further and simplifying network telemetry. The integration of an AI-enhanced version of the SONiC operating system aims to ease network management complexity for end-users, improving operational workflows during AI cluster scaling.
What to watch next
Monitor Upscale AI’s commercialization progress and adoption rates among hyperscalers and emerging cloud providers as they validate the operational benefits of AI-specialized networking technologies. Expansion of their scale-out switch product line and continued software improvements around the SONiC platform will be key indicators of market traction and customer readiness.
Industry response to interoperable open-standard protocols for AI traffic will also shape competitive dynamics. Upscale AI’s strategy to combine custom silicon with community-driven software may influence broader ecosystem approaches to AI infrastructure networking, potentially accelerating innovation cycles in the enterprise technology market.