# Radiant

> Radiant is an integrated AI infrastructure platform. It finances, builds and operates AI factories as one system: datacenters, AI hardware, managed compute and managed services, run through a single control plane called FlightDeck. Radiant is backed by Brookfield and serves hyperscalers, enterprises, governments and sovereign AI programs.

Radiant owns capital, powered land, compute, software and operations and runs them as a single platform with one contract, one control plane and one team accountable end to end. It is an NVIDIA Cloud Partner and builds clusters to the NVIDIA Cloud Partner reference architecture. Radiant was formerly Ori Industries.

Key facts:

- Backed by Brookfield, which holds one of the world's largest portfolios of powered land and operating data centers
- NVIDIA Cloud Partner deploying Blackwell Ultra (GB300 NVL72), Hopper (H200) and Vera Rubin (NVL72) systems
- Hyperscale capacity ready in 6 months on powered sites and 9 to 12 months on greenfield sites, delivered as single-tenant campuses under long-term agreements
- Power portfolio spans hydro, wind, geothermal, biomass and gas, with behind-the-meter generation available

## Platform

- [Platform Overview](https://radiant.co/platform-overview): How Radiant designs, builds and operates the full AI infrastructure stack as one system, from datacenter to workload
- [FlightDeck](https://radiant.co/flight-deck-and-flight-operations): The operational plane for the AI fleet, covering capacity, health and break-fix, topology, networking, identity, audit and lifecycle through UI, API, CLI and SDK

## Managed Compute

- [Bare Metal](https://radiant.co/bare-metal): Dedicated GPU clusters with direct hardware access, API-driven node lifecycle, automated health remediation and single-tenant isolation, scaling to tens of thousands of GPUs
- [Virtual Machines](https://radiant.co/virtual-machines): Hardware-isolated GPU instances on dedicated compute fabric, with fractional GPUs, suspend and resume, and lifecycle control by API, CLI, Terraform and SDK

## Managed Services

- [Managed Kubernetes](https://radiant.co/managed-kubernetes): GPU-native upstream Kubernetes where Radiant operates the control planes, node pools and accelerator stack
- [Managed Slurm](https://radiant.co/managed-slurm): Familiar Slurm queues and batch workflows on Kubernetes-operated infrastructure, with self-healing nodes and one shared GPU pool
- [Managed Storage](https://radiant.co/managed-storage): Parallel file, S3-compatible object and block storage provisioned with compute, with residency and isolation enforced at the storage layer
- [Managed Networks](https://radiant.co/managed-networks): Tenant-isolated network fabric with DPU-enforced flow-level policy, private interconnects and topology-aware placement

## AI Hardware and Datacenters

- [GPUs](https://radiant.co/gpus): NVIDIA GB300 NVL72, H200 and Vera Rubin NVL72 delivered as validated clusters across bare metal, instances, Kubernetes and Slurm
- [Storage Infrastructure](https://radiant.co/storage-infrastructure): The physical storage fleet, built into the cluster and tracked down to the drive through FlightDeck
- [Networking](https://radiant.co/networking): Edge to GPU networking on NVIDIA Quantum-X800 InfiniBand and Spectrum-X Ethernet, with four dedicated fabric planes and encrypted inter-site links
- [Datacenters](https://radiant.co/datacenters): A portfolio of powered, permitted, fiber-connected sites with high-density power, liquid cooling and a hybrid pre-fab build method

## Solutions

- [Hyperscale Solutions](https://radiant.co/hyperscale-solutions): Single-tenant AI factory campuses for hyperscalers and large enterprises, covering site selection, power development, supply chain, construction and operations after ready for service
- [Sovereign Solutions](https://radiant.co/sovereign-solutions): Sovereign AI infrastructure for nations, telcos and regulated industries, built on six pillars with a control plane that runs entirely within national boundaries

## Company

- [About Radiant](https://radiant.co/about): Mission and leadership team
- [Contact](https://radiant.co/contact): Sales and general enquiries
- [Talk to an Architect](https://radiant.co/talk-to-an-architect): Speak with the engineers who build and run the platform

## Leadership

- Mahdi Yahya, President. Founder and former CEO of Ori Industries
- Vishal Padiyar, Executive Chairman. Managing Director in Brookfield's Infrastructure Group
- Chris Galloway, Chief Technology Officer. Previously CTO at Ori
- Jonathan Symonds, Chief Marketing Officer. Previously CMO at MinIO and Ayasdi
- José Domingos, Chief Customer Officer. Previously at Ori and Chief Platform Architect at BT
- João Coelho, Chief Product Officer. Previously Head of Product at Ori and co-founder of Multy

## Documentation

- [Documentation](https://docs.radiant.co/): Product and API documentation

## Recent Blog Highlights

- [Agentic AI May Finally Be Here. It Is Changing How We Build AI Factories](https://radiant.co/blog/agentic-ai-is-changing-how-we-build-ai-factories): How AI agents turn single prompts into complex workloads and what that means for AI factory design
- [Optimizing for Inference: Six Hardware Characteristics of a High-Performance Cluster](https://radiant.co/blog/characteristics-of-a-high-performance-inference-cluster): The hardware priorities that shape inference performance
- [Put Every GPU to Work: Fleet Management with Radiant FlightDeck](https://radiant.co/blog/put-every-gpu-to-work): Fleet visibility, allocation and recovery that keep capacity productive
- [Private Networking Built to Protect Workloads and Keep AI Running](https://radiant.co/blog/private-networking-to-protect-workloads-and-keep-ai-running): Isolation and network automation in FlightDeck
- [AI Compute as a Market](https://radiant.co/blog/ai-compute-as-a-market): What makes an AI factory investable and how standardization enables a market in compute
- [Demystifying Behind-the-Meter: What It Actually Means and Why It Matters](https://radiant.co/blog/what-is-behind-the-meter-power): How behind-the-meter power helps AI buildouts overcome grid delays
- [The Boundary That Makes an AI Factory Repeatable](https://radiant.co/blog/the-boundary-that-makes-an-ai-factory-repeatable): The electrical boundary that lets an AI factory add power sources without redesign
- [A Deep Dive into NVIDIA's Vera Rubin NVL72 Platform](https://radiant.co/blog/a-deep-dive-into-nvidia-vera-rubin-nvl72): Architecture and performance of Vera Rubin NVL72
- [NVLink Partitioning: Turning Vera Rubin Performance into Superior GPU Economics](https://radiant.co/blog/nvlink-partitioning-for-superior-gpu-economics): How FlightDeck uses NVLink partitioning for utilization and multi-tenant isolation
- [The Control Plane is the Country: Why Sovereign AI Breaks at the Operational Layer](https://radiant.co/blog/the-control-plane-is-the-country-why-sovereign-ai-breaks-at-the-operational-layer): Why sovereignty depends on who operates the control plane


## Social
- [X](https://x.com/Radiant_Infra)
- [LinkedIn](https://www.linkedin.com/company/radiant-co)