Upscale AI Taps NVIDIA to Forge Open Superhighways for AI Clusters

📊 Key Data
  • 1.7x higher performance: NVIDIA's Spectrum-X Ethernet switch silicon delivers up to 1.7 times higher performance and efficiency compared to traditional Ethernet.
  • 95% data throughput: Spectrum-X achieves 95% data throughput in large-scale AI environments, significantly improving network efficiency.
  • 60% to 95% improvement: Standard Ethernet often sees only 60% throughput, leading to network congestion and performance issues.
🎯 Expert Consensus

Experts view this partnership as a critical step toward enabling scalable, open, and interoperable AI infrastructure, addressing the growing need for high-performance, vendor-neutral networking solutions in the AI era.

7 days ago
Upscale AI Taps NVIDIA to Forge Open Superhighways for AI Clusters

Upscale AI Taps NVIDIA to Forge Open Superhighways for AI Clusters

SANTA CLARA, CA – March 11, 2026 – In a significant move to reshape the backbone of artificial intelligence, networking startup Upscale AI announced today it will build a new class of open, high-performance data center systems powered by NVIDIA's industry-leading Spectrum-X Ethernet switch silicon. The partnership aims to deliver the networking equivalent of an open superhighway for AI, enabling companies to build massive, heterogeneous AI clusters without being locked into a single vendor's ecosystem.

This collaboration places Upscale AI, a well-funded newcomer, at the heart of a critical industry shift. As AI models scale into the trillions of parameters, the underlying network—the nervous system of the AI data center—has become a primary bottleneck. Upscale AI's plan to pair NVIDIA's purpose-built AI silicon with its own AI-optimized version of the open-source SONiC network operating system addresses a growing demand for infrastructure that is both immensely powerful and fundamentally interoperable.

The Revolution for an Open AI Fabric

The relentless growth of AI has created a paradox in data center design. While companies are racing to deploy diverse hardware from multiple vendors—including various GPUs, accelerators, and storage solutions—they are often stymied by proprietary networking fabrics that restrict flexibility and inflate costs. Upscale AI's strategy directly confronts this challenge.

At the core of its approach is SONiC (Software for Open Networking in the Cloud), an open-source network operating system born in the hyperscale data centers of Microsoft and now governed by the Linux Foundation. SONiC's primary benefit is disaggregation, decoupling the network software from the underlying hardware. This allows customers to avoid vendor lock-in, foster innovation, and select best-of-breed components. For AI, this means an enterprise could run NVIDIA GPUs alongside other accelerators on a network fabric that is not controlled by a single hardware manufacturer.

However, deploying open-source software in mission-critical, high-performance environments presents its own challenges. Upscale AI aims to bridge this gap by delivering an “enterprise-grade, focused SONiC” version. This involves hardening the software for the specific demands of AI traffic, ensuring the deterministic, lossless behavior required for distributed training workloads. Inefficient networking can cause expensive GPUs to sit idle, and Upscale AI's focus on integrating ASIC-native telemetry and predictable performance is designed to maximize the utilization of these costly assets.

“As AI infrastructure evolves toward increasingly heterogeneous architectures, scalable and operationally sound Ethernet fabrics are becoming essential,” said Alan Weckel, Co-Founder and Technology Analyst at 650 Group. “By combining NVIDIA Spectrum-X Ethernet switch silicon with a full-stack approach that integrates systems, software, and support, Upscale AI is addressing a key gap in making open, scale-out AI networking practical for large scale AI deployments.”

Taming the AI Beast with a Full-Stack Solution

While openness is a powerful ideal, enterprise customers demand reliability and simplicity. Upscale AI's strategy is not merely to provide software but to deliver a fully integrated, turnkey solution that combines hardware, software, and lifecycle services. This “full-stack” approach is designed to lower the barrier to entry for organizations wanting to build sophisticated AI infrastructure without the massive engineering overhead typically required to manage complex, disaggregated systems.

The choice of NVIDIA's Spectrum-X platform is central to this value proposition. Purpose-built for AI, Spectrum-X has demonstrated the ability to deliver up to 1.7 times higher performance and efficiency compared to traditional Ethernet. Research shows it can achieve 95% data throughput in large-scale AI environments, a dramatic improvement over the 60% often seen with standard Ethernet, where network congestion can cripple performance.

“NVIDIA Spectrum-X Ethernet switch silicon is setting a new standard for Ethernet-based AI performance,” said Barun Kar, CEO of Upscale AI. “Pairing this technology with our purpose-built systems and AI-optimized SONiC software allows us to deliver the best of both worlds: an open, highly scalable and interoperable architecture with operational simplicity.”

This translates directly to economic benefits. By accelerating training times and maximizing hardware utilization, the combined solution promises a lower total cost of ownership (TCO) and a faster return on massive AI investments. According to Sameh Boujelbene, Vice President at Dell'Oro Group, the move addresses a core market need. “As scale-out Ethernet becomes the foundation for AI infrastructure, organizations are looking for solutions that combine openness with operational maturity,” she noted.

NVIDIA's Expanding Influence in the AI Data Center

The partnership is also a significant strategic victory for NVIDIA. While best known for its GPUs, NVIDIA has been aggressively building out a full-stack portfolio to dominate every aspect of the AI data center. Its Spectrum-X platform is a direct challenge to established networking giants like Arista and Cisco, and its success has been swift. Propelled by adoption from hyperscalers like Meta and Oracle, NVIDIA has seen its Ethernet switch revenue surge, capturing significant market share in the data center segment.

By enabling partners like Upscale AI to build solutions on top of its silicon, NVIDIA is fostering a broader ecosystem around Spectrum-X. This strategy accelerates its adoption and solidifies its position as a foundational technology for the AI era, competing not only on a product level but on an ecosystem level. The collaboration allows NVIDIA to extend its reach into enterprises that prioritize open, multi-vendor architectures.

“To lead in the trillion-parameter model era, scalability and efficiency are paramount,” said Gilad Shainer, SVP at NVIDIA. “We look forward to collaborating closely with Upscale AI as the team leverages the NVIDIA Spectrum-X Ethernet platform to help companies build the world's most advanced open AI infrastructure.”

As part of the initiative, Upscale AI has joined the NVIDIA Partner Network, signaling deep collaboration on reference architectures and validated designs. With plans to bring its Spectrum-X based systems to market later this year, Upscale AI is positioning itself as a key enabler in a new chapter of AI infrastructure, one where extreme performance and strategic openness are no longer mutually exclusive.

Sector: Software & SaaS AI & Machine Learning Venture Capital
Theme: Artificial Intelligence Generative AI Machine Learning Cloud Migration
Event: Private Placement
Product: AI & Software Platforms Connectivity & Infrastructure
Metric: Revenue EBITDA Net Income

📝 This article is still being updated

Are you a relevant expert who could contribute your opinion or insights to this article? We'd love to hear from you. We will give you full credit for your contribution.

Contribute Your Expertise →
UAID: 21025