# PaletteAI and BE Verity: one workflow for GPUs and fabric | Spectro Cloud

> Source: <https://www.spectrocloud.com/blog/paletteai-be-verity-gpu-and-fabric-orchestration>
> Published: 2026-11-17 15:00:00+00:00

**Key takeaway:** PaletteAI now calls Verity so that assigning GPU servers to a tenant in PaletteAI automatically configures that tenant's segmented network on the fabric, removing the manual compute-to-network handoff.

# Spectro Cloud and BE unite GPU cluster and fabric orchestration for AI factories

## The handoff that slows down AI builds

Standing up an [AI factory](https://www.spectrocloud.com/ai/ai-factories) takes two kinds of infrastructure work. One team racks GPU servers, installs Kubernetes, and gets workloads running. Another team builds the switch fabric those servers depend on.

In most organizations, those teams use separate tools and coordinate through tickets and spreadsheets. When the compute side changes, someone on the network side has to make a matching change by hand. As clusters grow, that handoff becomes one of the slowest steps in getting new capacity online.

Today Spectro Cloud and BE are announcing an integration between [Spectro Cloud PaletteAI](https://www.spectrocloud.com/platform/paletteai) and [BE Verity](https://be-net.com/verity/). PaletteAI manages the compute layer. Verity orchestrates and observes the network fabric. With the integration, platform teams can bring a new tenant’s GPU servers online in PaletteAI, and Verity applies the matching network configuration on the fabric automatically, without a separate request to the networking team.

## A single build for two different layers

PaletteAI handles compute. It provisions bare-metal GPU servers, deploys Kubernetes clusters and the [GPU operator](https://www.spectrocloud.com/blog/the-real-world-guide-to-the-nvidia-gpu-operator-for-kubernetes-ai), and runs AI stacks and workloads on top.

Verity handles the network. It orchestrates multi-vendor Ethernet fabrics across multiple versions of SONiC and Cumulus, from Day 0 design and zero-touch provisioning through Day 2 operations.

It has three parts:

- Verity orchestration builds the GPU and data center fabric to validated reference architectures, including multi-tenant segmentation, and serves as the network's source of truth.
- [Satori](https://be-net.com/satori) observability collects telemetry across compute, storage, and network, and uses machine learning to flag and predict problems before they slow down workloads.
- [SensAI](https://be-net.com/sensai/) is an AI agent that helps operators run the fabric day to day. Teams can use its built-in model or bring their own LLM (such as Claude or Codex) through MCP.

Verity is already running in production, including at one of the world’s largest neoclouds.

## What the integration does today

The integration is built for infrastructure and platform teams at [neoclouds](https://www.spectrocloud.com/blog/five-things-every-neocloud-and-msp-needs-to-turn-ai-factories-into-a-business-that-scales), enterprises, telcos, service providers, and public sector organizations that own and operate their own GPU servers and network fabric.

With the integration, PaletteAI works with fabric that Verity has already built. Tenant provisioning runs as a single workflow across the compute and network layers.

Here's an example.

A neocloud signs a new customer that needs a block of GPU servers. The platform team assigns the servers to that customer in PaletteAI, and PaletteAI calls Verity, which sets up that customer’s network across the fabric and connects the servers to it. PaletteAI then deploys Kubernetes and the AI stack on top. Compute and network are ready at the same time, and the customer’s traffic stays separate from other tenants on the fabric from the start.

## What the partners are saying

“In an AI factory, the network is the computer. GPUs only perform as well as the fabric connecting them. Verity already runs that fabric from design through day-to-day operations, and connecting it to PaletteAI means tenant changes on the compute side are matched on the network automatically.” Amir Elbaz, CEO and founder, BE.

“Teams running their own GPU infrastructure need to turn new capacity into tenant-ready environments quickly. With Verity, PaletteAI provisions a tenant’s compute and network in one workflow, which removes one of the slowest handoffs in getting a cluster into production.” Saad Malik, CTO and co-founder, Spectro Cloud.

## See it at SC26

BE will demo Verity at the Broadcom booth #4626 at [SC26](https://sc26.supercomputing.org/), and the Spectro Cloud team will be available at #7202 to dive deeper into the integration. If you're building an AI factory, [meet with both teams](https://www.spectrocloud.com/get-started) to see how PaletteAI and Verity connect compute provisioning with network automation.
