AI Networking Startups Power Next-Gen Data Centers Now

Big moves are shaking up AI infrastructure today. Three startups—Netris, Upscale AI, and Coval—just raised huge funding rounds. They’re racing to build the networking and evaluation tools that AI data centers need right now.

Netris: Revolutionizing GPU Cluster Networking

Netris just secured $15 million in Series A funding from Andreessen Horowitz. Their CEO, Alex Saroyan, says traditional software-defined networking (SDN) falls short for GPU clusters. Why? Because SDN is software-based and can’t keep up with daily link changes in massive AI data centers.

Netris offers a vendor-agnostic platform that works with networking gear used in Nvidia and AMD-powered data centers. Nvidia noticed Netris two years ago after a demo impressed them, and they started recommending Netris to customers.

Today, Netris powers more than 35 GPU clusters worldwide. Operators include Lightning AI, Foxconn, Visionbay, Hewlett Packard Enterprise, TensorWave, and Telus. These clusters run about a million GPUs total. That’s a lot of AI horsepower managed by one platform.

Upscale AI: Custom Switch Silicon Meets AI Networking

Upscale AI raised a massive $190 million in a Series A-1 round, led by Premji Invest. This brings their total funding to $500 million and values the company at $2 billion. Founded only in 2025, Upscale AI is already making waves.

The startup spun out of Bitcoin mining hardware maker Auradine. Now, they’re building a custom scale-up switch ASIC called Skyhammer. This silicon is designed to connect GPUs and XPUs inside AI systems at blazing speeds.

Rajiv Khemani, Upscale AI’s co-founder, explains Skyhammer doesn’t work alone. It must pair with XPUs and GPUs that have scale-up capabilities. This collaboration creates a powerful AI-optimized network fabric.

Upscale AI is also aligned with Nvidia’s Spectrum X Ethernet platform. Nvidia is a strategic investor and partner. Together, they’re crafting next-gen Ethernet switches and software for hyperscalers and neocloud providers. The goal? To optimize AI token usage, adapting to the evolving AI workload demands.

Upscale AI plans to reveal product details on Skyhammer later this year. Deployment will sync with GPU and XPU readiness, promising a perfect match of hardware and software.

Coval: The Voice AI Evaluation Powerhouse

Coval raised $28 million in Series A funding led by Norwest, adding to a total of $31 million raised since launching in 2024. This startup focuses on testing and monitoring AI voice and chat agents at scale.

Voice AI is booming, but it needs reliability. Coval’s platform covers the entire voice agent lifecycle:

• Pre-deployment simulation
• Live monitoring
• Human review and performance evaluation

The startup runs tens of millions of AI agent evaluations. Brooke Hopkins, founder of Coval, brought her evaluation expertise from Waymo’s self-driving car division to voice AI. Industry leaders recognize her impact. One expert said, “She helped prove self-driving cars could work, and now she’s tackling voice AI.”

AI Networking Is Booming—And Just Getting Started

The AI infrastructure market is exploding. Dell’Oro Group forecasts AI data center switch spending could top $100 billion annually by 2030. Enterprise network traffic driven by AI workloads is expected to surge over 30% by 2028.

Major cloud players like Microsoft, Google, Meta, and Amazon will nearly double their infrastructure spending in 2026. They need networking solutions that scale and perform under the intense demands of AI.

Nvidia and Broadcom continue heavy investments in AI networking tech. Nexthop AI, another competitor, raised $500 million in Series B funding at a $4.2 billion valuation earlier this year. The race to lead AI networking is fierce and fast.

Upscale AI’s CEO, Barun Kar, sums it up: “This investment reinforces our vision and enables us to scale as we engage with leading neocloud and hyperscale providers to meet growing demand for open, interoperable AI infrastructure.”

Meanwhile, Netris’s Saroyan says, “As a GPU cluster operator, you need to make configuration changes to every link, every day. Traditional SDN is falling short because it’s software technology.” Their platform fills that critical gap.

Across the board, these startups are building the backbone for AI’s future. From smarter networking hardware to rigorous AI agent evaluation, the groundwork is being laid right now. The AI data centers of tomorrow will run faster, smarter, and more reliably because of them.