AI-RAN: shared accelerated infrastructure.

7 technologies in AI-RAN and Open RAN.

Technologies

AI-for-RAN Using AI models to improve radio performance, spectral efficiency and operations.
AI-and-RAN (shared GPU for RAN and AI) Running the RAN and AI inference on the same accelerated servers, with RAN given priority when traffic needs it.
AI-on-RAN (edge AI services on RAN sites) Hosting third-party and site AI applications on RAN compute close to users and sensors.
Software-defined L1 on GPUs Running the full physical layer in software on GPUs instead of fixed-function hardware.
RAN-first workload scheduling Schedulers that protect RAN deadlines and lend idle accelerator time to AI jobs.
Isolation between RAN and AI tenants Hardware and software isolation so AI tenants cannot affect RAN timing or reach RAN data.
AI-RAN site power and cooling Power and thermal design for accelerated cell sites and hubs, measured against energy efficiency metrics.