Upscale AI raises $200M Series A to build AI networking platforms

Upscale AI has closed a landmark $200 million Series A round that pushes the startup’s total financing past $300 million. The oversubscribed raise is led by Tiger Global, Premji Invest and Xora Innovation, with active participation from Maverick Silicon, StepStone Group, Mayfield, Prosperity7 Ventures, Intel Capital and Qualcomm Ventures. The new capital will accelerate product rollout and commercial deployments for the company’s rack-scale AI networking stack.

The round underscores a widely held view among infrastructure investors: networking is now a primary limiter on AI scale. Where legacy data-centre networks were designed to link general-purpose compute and storage, AI workloads require tightly synchronized interconnects that collapse latency across accelerators, memory and storage. Upscale AI was founded to target that gap with a purpose-built silicon-plus-systems approach.

At the heart of Upscale’s offering is SkyHammer, a turnkey scale-up solution that the company describes as a means to unify GPUs, AI accelerators, memory, storage and switching into a single, synchronized engine. The platform is based on open protocols and software, and the company participates in industry groups and open-source efforts to drive interoperability as heterogeneous AI hardware proliferates.

Investors said they backed the team for its speed of execution and strategic positioning. Barun Kar, Upscale AI’s CEO, said the funding will expand engineering and commercial teams as the company moves into customer deployments. Executive Chairman Rajiv Khemani highlighted customer demand at hyperscalers and AI infrastructure operators seeking alternatives to closed, proprietary networks.

Market context strengthens the case. Independent analysts project the addressable opportunity for AI-optimized networking to grow sharply as model sizes and interconnect demands rise — some estimates put the TAM for AI networking at roughly $100 billion of annual spend by the end of the decade. That outlook is driving both specialist startups and strategic corporate investors to double down on new interconnect approaches that reduce latency and improve utilization for large-scale training and inference.

Operationally, Upscale plans to use the cash to accelerate shipments this year, hiring across R&D, systems engineering and go-to-market. The company positions itself as a full-stack supplier — combining proprietary silicon, integrated systems and orchestration software — with an emphasis on open standards to ease adoption and avoid vendor lock-in.

For the wider market, Upscale’s momentum is another signal that networking will be a major battleground for next-generation AI infrastructure. If the company can convert traction with early hyperscaler trials into production deployments, it could help define a new class of rack-scale architectures that tackle the synchronization and latency constraints limiting current AI compute fabrics.