Upscale cofounders Barun Kar (left) and Rajiv Khemani (right). · Fortune · Courtesy of Upscale AI
Upscale AI Raises $190 Million at $2 Billion Valuation With Nvidia and Salesforce Ventures to Solve AI's Networking Bottleneck
Santa Clara-based Upscale AI raised $190 million in a Series A-1 round on June 22, 2026, bringing its total funding to $500 million and its valuation to $2 billion - all in under 18 months since founding. The round was led by Premji Invest and included new investors Nvidia, Salesforce Ventures, Seligman Ventures, and Temasek, alongside returning backers including Tiger Global, Mayfield, StepStone Group, and Maverick Silicon. The company is building what it calls an open-standard AI networking fabric - a full-stack platform spanning silicon, systems, and software designed to connect accelerators, memory, and storage into a unified AI engine.
The Santa Clara startup raised $190 million in a Series A-1 round, bringing its total funding to $500 million and its valuation to $2 billion - all in under 18 months. Premji Invest, a $15 billion investment fund that previously backed one of the founders' earlier exits, led the round. New investors include Nvidia, Salesforce Ventures, Temasek, and Seligman Ventures, alongside returning backers Mayfield, Tiger Global, StepStone, Maverick Silicon, and Prosperity7. BetaKit
The Problem Upscale Is Solving
Networking is becoming the hidden bottleneck in AI infrastructure. As AI clusters scale to hundreds of thousands of GPUs, the speed at which those chips communicate with each other - and with the memory and storage holding the data they need - increasingly determines the actual training and inference performance, not the raw compute capability of each individual chip.
Upscale AI is reimagining networking for the AI era with a full-stack portfolio spanning silicon, systems, and software. Through a unified, open-standard AI fabric, the company connects accelerators, memory, and storage into a single high-performance AI engine. CEO Barun Kar said: "To be able to go to our suppliers and lock in that supply, people are interested in seeing how strong our balance sheet is." The opportunity Upscale is chasing is enormous. Spending on AI data center switches is forecasted to surpass $100 billion annually by 2030, according to Dell'Oro Group, as Microsoft, Google, Meta, and Amazon race to build out AI infrastructure. FinTech Magazine
The company's thesis is that the networking architectures built for traditional data centers - designed to connect general-purpose compute and storage endpoints - are fundamentally unsuited for AI clusters. Traditional networking connects endpoints. AI networking, in Upscale's framing, unifies the cluster.
The Open-Standard Bet
The strategic bet that distinguishes Upscale from competitors is its commitment to open networking standards. The open networking standard Upscale is building on is backed by AMD, Intel, Google, Meta, Microsoft, and 80+ other companies. However, it still needs to prove it can match Nvidia's proprietary networking performance. FinTech Magazine
Nvidia's proprietary NVLink and InfiniBand networking have dominated AI cluster connectivity because they deliver exceptional performance when paired with Nvidia's own GPUs. As AI infrastructure becomes more heterogeneous - mixing different chip architectures, different memory types, and different storage systems - the argument for open-standard networking grows. An open fabric that works across AMD, Intel, and Nvidia hardware is more valuable than a proprietary one that locks customers into a single vendor's ecosystem.
Nvidia's participation as an investor in Upscale is strategically interesting in this context. The chip giant is simultaneously defending its proprietary networking stack and investing in the open-standard competitor. That positioning hedges Nvidia's bets in a market where the outcome is genuinely uncertain.
The Competitive Stakes
Competitive pressure is mounting. Nexthop AI, a direct rival focused on custom AI networking for the same hyperscaler and neocloud customers, closed a $500 million Series B in March at a $4.2 billion valuation. Broadcom and Nvidia have effectively unlimited R&D budgets. FinTech Magazine
For business leaders tracking AI industry infrastructure trends, the Upscale raise validates something worth understanding: the next major competitive battleground in AI infrastructure is networking, not just compute. The $100 billion annual market projected for AI data center switches by 2030 represents a new layer of the AI stack where the winners have not yet been determined. That uncertainty is exactly what makes $500 million in venture funding at a $2 billion valuation look reasonable to investors like Nvidia, Temasek, and Tiger Global.
Cut Through the Noise
What is Upscale AI and what did it raise?
Upscale AI is a Santa Clara-based AI networking infrastructure company that raised $190 million in a Series A-1 round on June 22, 2026, at a $2 billion valuation, bringing total funding to $500 million in under 18 months. The round was led by Premji Invest and included Nvidia, Salesforce Ventures, Temasek, and Seligman Ventures as new investors. Upscale builds open-standard AI networking fabric connecting accelerators, memory, and storage into unified high-performance AI infrastructure.
Why is AI networking a bottleneck in 2026?
As AI clusters scale to hundreds of thousands of GPUs, the speed at which chips communicate with each other and access memory determines real-world performance, not raw compute alone. Traditional networking architectures designed for general-purpose data centers cannot efficiently handle the communication patterns of large-scale AI training and inference workloads. Spending on AI data center networking is projected to exceed $100 billion annually by 2030, according to Dell'Oro Group.
What is Upscale AI's open-standard approach and why does it matter?
Upscale builds on an open networking standard backed by AMD, Intel, Google, Meta, Microsoft, and over 80 other companies, rather than Nvidia's proprietary NVLink and InfiniBand architecture. An open standard allows AI networking to work across heterogeneous chip environments rather than locking customers into a single vendor's ecosystem. As AI infrastructure diversifies beyond Nvidia-only clusters, open-standard networking becomes increasingly valuable.
Who are Upscale AI's main competitors?
Nexthop AI, a direct rival targeting the same hyperscaler and neocloud customers, closed a $500 million Series B in March 2026 at a $4.2 billion valuation. Broadcom and Nvidia also compete in the AI networking space with effectively unlimited R&D budgets. Upscale's differentiation is its open-standard, full-stack approach spanning custom silicon, systems, and software purpose-built for AI workloads.
