SAN JOSE, Calif., March 17, 2026 /PRNewswire/ — Juice Labs today announced at NVIDIA GTC 2026 that its GPU-over-IP platform has joined the NVIDIA AI Grid ecosystem as a software partner. The company’s platform provides a software-defined fabric layer that enables telecommunications providers to pool and deliver distributed AI inferencing capabilities across existing fiber networks, without centralized cloud buildouts or hyperscaler-grade orchestration at every edge location.
A major U.S. cable operator operating across 41 states has validated the architecture through an initial pilot, confirming that service providers can bring GPU compute services to market in weeks, with dramatically lower capital requirements than traditional edge cloud deployments.
“NVIDIA AI Grid gives telecommunications providers a world-class foundation for distributed AI,” said Steve Golik, Co-Founder & CEO Juice Labs. “What we bring to the ecosystem is software fabric that makes distributed GPUs easily accessible, so a service provider can deploy capacity where it makes operational sense today and serve enterprise customers anywhere across their network.”
Extending NVIDIA AI Grid to the Telco Edge
NVIDIA AI Grid defines a distributed intelligence architecture spanning AI factories, regional POPs, central offices, and RAN sites, built on NVIDIA RTX PRO™ 6000 Blackwell GPUs, NVIDIA Spectrum-X Ethernet , and NVIDIA BlueField DPUs networking fabric. Juice Labs integrates with this architecture to address three deployment realities specific to the telco edge:
Simplified deployment. Service providers connect NVIDIA servers directly into the Juice fabric without building full-stack cloud infrastructure at each location, maximizing GPU density within the constrained power, space, and cooling environments of telco POPs. Deployments that traditionally took quarters now take weeks.
Dynamic resource pooling. GPU capacity is shared elastically across multiple edge sites, including east-west traffic between locations. Providers deploy capacity strategically across their footprint and allocate resources based on real customer workloads, transforming a collection of edge sites into a unified GPU portfolio.
Fractional GPU delivery. Juice’s virtualization layer allocates GPU capacity down to the GB of VRAM with runtime reallocation in seconds, enabling service providers to serve diverse enterprise workloads with precision and improve unit economics across the infrastructure.
The architecture also delivers a data sovereignty advantage: enterprise customers access distributed AI inferencing capabilities over existing fiber connections while keeping sensitive data on their own premises, supporting compliance requirements in healthcare, finance, and government sectors.
Available Now
Juice Labs’ GPU-over-IP platform for NVIDIA AI Grid is available now for telecommunications service providers and enterprise customers on x86 and Arm platforms.
To schedule a meeting at GTC with Juice Labs team please email [email protected] or visit our GTC landing page https://www.juicelabs.co/gtc
About Juice Labs
Juice Labs provides GPU-over-IP software that enables telecommunications providers, enterprises, and research institutions to pool, share, and deliver GPU resources across standard IP networks. The company’s platform serves as the connectivity fabric for distributed NVIDIA AI Grid deployments, enabling service providers to capture new GPU-as-a-Service revenue from existing edge infrastructure. Juice Labs is a member of the NVIDIA Inception Program. For more information, visit www.juicelabs.co or contact [email protected].
Media Contact
Jaymie Scotto & Associates (JSA)
+1 866.695.3629 ext. 4
[email protected]
SOURCE Juice Labs
