A Pure-Play Bet on AI Networking
Upscale AI, a Santa Clara-based startup building networking infrastructure for the AI era, has raised $190 million in a Series A-1 extension that brings its total funding to $500 million and values the company at $2 billion. For more context on this story, see our ongoing breaking AI news.
The round was led by Premji Invest and included new funding from NVIDIA, Salesforce Ventures, Seligman Ventures and Temasek, alongside participation from existing investors Maverick Silicon, Mayfield, Prosperity7 Ventures, StepStone Group and Tiger Global, according to a company announcement issued June 22.
The backing from NVIDIA is particularly notable, lending the credibility of the dominant AI chipmaker to a startup pitching itself as the connective tissue that knits together large-scale AI compute. For a networking upstart, NVIDIA's imprimatur is the kind of validation that can open doors with the very hyperscalers it hopes to sell to.
Fixing the Network Bottleneck
Upscale AI describes itself as a pure-play AI networking infrastructure company, reimagining how data moves between the components of an AI system. Its full-stack portfolio spans silicon, systems and software, woven together through what the company calls a unified, open-standard AI fabric.
That fabric is designed to connect accelerators, memory and storage into a single high-performance engine, purpose-built to eliminate network bottlenecks across large-scale, synchronized AI workloads — including the training and inference of frontier models. The emphasis on synchronization is key: when thousands of chips must exchange gradients in lockstep during training, any latency or congestion in the network throttles the entire cluster.
"AI infrastructure is being redefined at cluster scale, and networking is one of the most critical bottlenecks," said Barun Kar, CEO of Upscale AI. "Upscale AI is building a high-performance, open-standard AI fabric purpose-built for large-scale, synchronized workloads."
Why Networking Is Suddenly Strategic
As AI models have grown, attention has shifted from the GPUs themselves to everything around them. Once a cluster reaches tens of thousands of accelerators, the network that shuttles data between them becomes a decisive factor in both performance and cost. Interconnects that once seemed adequate for general-purpose cloud computing now struggle to keep frontier training runs fully fed, leaving expensive compute idling while it waits for data.
That dynamic has turned networking into one of the most contested frontiers in AI hardware, with startups and incumbents alike racing to build faster, more efficient fabrics. The stakes are financial as well as technical: every percentage point of idle compute on a billion-dollar cluster represents capital burned for no return, which is why networking efficiency now translates directly into margin for the operators footing the bill. Upscale AI's open-standards pitch positions it against proprietary approaches, betting that hyperscalers and the so-called neocloud providers that specialize in AI compute will favor interoperable infrastructure over locked-in ecosystems.
The company said it is actively engaged with multiple hyperscalers and leading neocloud infrastructure providers, with customer evaluations and deployments underway across both scale-out and scale-up networking environments — the two dominant architectures for distributing AI workloads across clusters.
Doubling Down
For lead investor Premji Invest, the round is a doubling down on an existing bet. The firm previously backed the company and said its progress since has only deepened its conviction.
"Rajiv, Barun, and the Upscale AI team are tackling one of the most critical bottlenecks in AI infrastructure, and they're doing it with a world-class team and a differentiated architectural approach," said Sandesh Patnam, managing partner at Premji Invest. "We're thrilled to lead this round and double down on the company's vision."
Upscale AI plans to use the new capital to scale the business and accelerate delivery of its AI-native networking technology, pushing its fabric from evaluation into broader production deployment.
The Funding Environment
The raise is the latest evidence that investors remain eager to fund the pick-and-shovel layer of the AI boom, even as scrutiny intensifies around the valuations of model makers themselves. Startups that solve concrete infrastructure problems — power, cooling, memory and now networking — continue to command premium valuations on the strength of customer demand rather than speculative promise.
Crunchbase recently noted that venture funding for infrastructure and robotics startups has surged to record numbers in 2026, a trend that Upscale AI's round exemplifies. The willingness of a deep-pocketed strategic like NVIDIA to back a networking challenger also suggests the chip giant sees complementary, rather than competitive, value in an open fabric that makes its GPUs more useful — a signal that the AI hardware stack is bifurcating into specialist layers rather than consolidating under any single vendor.
With $500 million now raised and a $2 billion price tag, Upscale AI has the runway to push its open-standard fabric into production at hyperscale. Whether it can convert evaluations into large, multi-year deployments will determine whether the networking bottleneck becomes its defining opportunity — or someone else's.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
