Last Updated: 22 June 2026
Upscale AI, a company focused entirely on AI networking infrastructure, has closed a $190 million Series A-1 financing round. The Santa Clara-based company was co-founded by CEO Barun Kar and Executive Chairman Rajiv Khemani, both repeat founders. Their core argument is simple: as AI clusters grow larger and more complex, the network connecting accelerators, memory, and storage becomes the limiting factor on what those systems can achieve.
AI training and inference workloads have grown far faster than the network infrastructure built to support them. General-purpose data center networks were designed to connect separate computing endpoints. They were not built to synchronize thousands of processors running frontier AI models simultaneously. Specialized AI networking is now one of the fastest-growing segments in the infrastructure market, expanding at over 30% annually. We previously covered how Upscale AI raised its $200 million Series A earlier this year, when the round brought the company to unicorn status.
The Series A-1 was led by Premji Invest. New investors NVIDIA, Salesforce Ventures, Seligman Ventures, and Temasek joined the round, alongside existing backers Maverick Silicon, Mayfield, Prosperity7 Ventures, StepStone Group, and Tiger Global. The investment brings Upscale AI's total funding to $500 million and its valuation to $2 billion.
Most of what Upscale AI is building comes down to an architectural choice. A significant part of today's AI networking infrastructure relies on proprietary systems that lock operators into a single vendor's ecosystem. For large cloud providers and neocloud operators, that creates real practical problems: higher costs, reduced flexibility, and limited ability to upgrade or change infrastructure over time.
Upscale AI is building around open standards, meaning components from different vendors can be connected and used together. It is an approach that competes directly with proprietary alternatives by offering interoperability rather than lock-in.
"Building the future of AI requires a rethink of underlying data center infrastructure, particularly as organizations demand scalable, interoperable systems that avoid vendor lock-in. Upscale AI addresses this challenge by combining a commitment to open networking standards with a zero-packet-loss architecture that maximizes hardware utilization."
Emily Zhao, Managing Director at Salesforce Ventures
The company participates in open-standard industry bodies and contributes to the development of shared networking specifications.
The new funding will go toward scaling operations and accelerating delivery of its networking technology. Upscale AI is already in conversations with multiple large cloud providers and neocloud infrastructure operators, with customer evaluations and deployments underway across scale-out and scale-up networking environments.
"AI infrastructure is being redefined at cluster scale, and networking is one of the most critical bottlenecks. Upscale AI is building a high-performance, open-standard AI fabric purpose-built for large-scale, synchronized workloads. This investment reinforces our vision and enables us to scale the business as we engage with leading neocloud and hyperscale infrastructure providers to meet growing demand for open, interoperable AI infrastructure."
Barun Kar, CEO of Upscale AI
The company's fundraising has moved quickly. Upscale AI went from a $100 million seed round in September 2024 to a $200 million Series A in January 2026, and has now added $190 million more within months. That kind of pace is uncommon even in a sector that attracts significant investment.
Upscale AI was founded by Rajiv Khemani and Barun Kar. The company is headquartered in Santa Clara, California. Its focus is entirely on networking for AI infrastructure, not as one product among many but as the company's sole purpose. The leadership team includes CTO Puneet Agarwal alongside senior executives across product, engineering, sales, and operations.
The product portfolio covers three layers: silicon, systems, and software. Its scale-up product, SkyHammer, is built from the chip level to deliver reliable, low-latency networking for synchronized AI workloads, from individual racks to larger cloud-scale clusters. It brings accelerators, memory, and storage together into one coordinated system. The goal is to reduce the communication delays that slow down large-scale training and inference jobs. On the scale-out side, Upscale AI offers Ethernet systems built on NVIDIA Spectrum-X switch silicon and a SONiC-based network operating system. SONiC is an open-source operating system widely used in large data centers. These systems are designed for distributed AI clusters that use hardware from multiple vendors.
One specific design choice sets the architecture apart: zero-packet-loss networking. In AI training, when a data packet is lost or delayed, the entire cluster has to wait and resend. Upscale AI builds its network fabric to prevent that from happening. The result is better use of available hardware and fewer interruptions during intensive workloads.
Upscale AI has now raised across three rounds. The $100 million seed round closed in September 2024. The $200 million Series A followed in January 2026, with Maverick Silicon, StepStone Group, Mayfield, Prosperity7 Ventures, and Tiger Global participating. All five returned for the Series A-1, joined by Premji Invest as lead and new investors NVIDIA, Salesforce Ventures, Seligman Ventures, and Temasek.
"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. The progress we've seen since our initial investment has only deepened our conviction, and we're thrilled to lead this round and double down on the company's vision."
Sandesh Patnam, Managing Partner at Premji Invest
NVIDIA's participation carries strategic weight. Upscale AI's scale-out systems are built on NVIDIA Spectrum-X switch silicon, and the company joined the NVIDIA Partner Network in March 2026. Singapore-based investment firm Temasek and Salesforce Ventures round out a group that includes technology-focused venture firms, long-term institutional investors, and strategic corporate backers.
