Radian Arc is the AMD-exclusive GPU cloud for telcos. One rack, 42RU, ~10kW — deployed in your data centre, federated across 9 markets, in production with Jio and True. Low latency by construction, sovereign by design, CAPEX-free to the operator.
AMD is an investor in Radian Arc, and Radian Arc is AMD's go-to-market partner for telcos. AMD sells their cloud GPU silicon to two classes of customer: the hyperscalers, and us.
That commercial exclusivity is the moat. It's why a sovereign, in-network cloud engine is economically viable at telco scale — and why we can underwrite the hardware without passing CAPEX to the operator.
Every Radian Arc POP is the same shape: physical infrastructure at the bottom, compute above it, our control plane binding them, content licensed on top, and the subscriber experience at the surface. Five layers, one rack, federated across nine markets — and every layer is inspectable, swappable, and owned by a single control plane. Click any layer to inspect what sits inside.
A single 42RU rack fits in any Tier-III or carrier-grade data centre: ~10kW draw, standard power, standard cooling, standard optical uplinks. No exotic footprint, no new building, no new skills on the ops team. If you can land a router, you can land a Radian Arc POP.
Every POP runs AMD's cloud GPU. v520 ships today at 32 concurrent users per GPU; v620 lifts that to 48; v710 to 72 — all on the same chassis, via field-swappable card upgrades. The subscriber-to-CCU ratio is ~3.5%, so a single rack underwrites roughly 914 paying subs.
Radian Arc's federation layer is a single control plane over every deployed rack, everywhere. Sessions are routed to the nearest healthy POP; overflow spills to a partner POP or to public cloud. You manage a market, we manage the cloud.
The catalogue ships pre-cleared for streaming via Blacknut and direct publisher deals. No per-title licensing work on the operator side. Co-streamable with operator-owned content or regional titles when you have them.
White-label native apps on iOS, Android, Smart TV and WebOS, billed through the carrier, launched inside the operator's own app-home. Subscribers never see Radian Arc. You own the brand. We own the cloud underneath.
A Radian Arc POP is deliberately unexciting hardware: Juniper edge, Dell management, two Supermicro GPU chassis. Boring is the point — any telco ops team can land this in any data centre, this quarter.
Every rack, in every market, runs under one Radian Arc brain. Sessions route to the nearest healthy POP; if a market spikes, capacity spills sideways to a peer POP or up to public cloud. No one market ever runs alone.
Cloud gaming is the anchor tenant — but a Radian Arc POP is a GPU cloud. Once the rack is in-network, the operator unlocks AI inference, desktop-as-a-service, SD-WAN, and the long tail of GPU workloads that telcos have been unable to monetize locally.
500+ titles, white-labelled, carrier-billed. The anchor tenant that pays for the rack.
AMD ROCm on the same GPUs the games run on — used for CDN telemetry and AI inference for value-added services.
Desktop-as-a-service for CAD, medical imaging, ML training — streamed from the same rack.
ActivePort SDN/SD-WAN running inside the POP — the rack is the orchestration plane for the operator's own network.
Dynamic capacity allocation between services — cloud gaming at peak, enterprise workloads off-peak.
Sovereign VDI for regulated industries: healthcare, finance, government — where data must stay in-country.
Computer vision and edge AI for municipal deployments — traffic, safety, crowd analytics.
Radiology and pathology inference on sovereign silicon. Patient data never leaves the country.
Virtual-machine and bare-metal-as-a-service — the same rack, rented by the hour to enterprise customers.
A BOM, a commercial structure, a deployment timeline, and a first-year subscriber model — tailored to your network, your billing, your content regulator.