Inference

Aviz podcast says GPU shortages, privacy limits, and aging workforces are pushing Japan toward on-premise AI networking with Kubernetes, GPUs, SONiC.

Aviz Networks podcast discusses how SONiC adoption shifted from evaluation to production planning for AI-era networking.

The blog argues that AI depends on networking, outlining scale-up, scale-out, open architectures, and vendor-agnostic planning to control AI cost per token.

Says AI traffic has shortened networking refresh cycles to about three years, and cites Cisco’s “no AI without a secure network” quote.

Podcast discussion on building AI-era networking: moving from one fabric to multiple networks for AI factories, with openness and standardization to manage complexity and cost.

Aviz Networks podcast episode 3 says AI changes networking via AI-ready infrastructure and AI-assisted network operations, shifting KPIs toward utilization.

Scality and WEKA expanded their France partnership with a joint customer support agreement and local French-language tier-one support.

Crusoe expanded Crusoe Intelligence Foundry with serverless fine-tuning and self-serve deployments, exporting .safetensors weights and enabling one-click inference.

ABI Research forecasts TinyML AI chipset shipments above 4.1 billion units by 2031, with revenue exceeding $7.8B.

Teleport added LLM Proxy and Delegated Identity to Beams public beta for auditing agent requests and scoping agent permissions.

The post contrasts AI conversations in applications with performance changes in data movement, network design, telemetry, and stable operations for agents.

After Cisco Live 2026, Sameh Boujelbene analyzes how Cisco is positioning validated AI infrastructure platforms beyond standalone hardware performance and networking speeds. The post Cisco’s AI Data Center Strategy Expands Beyond Hardware Toward Validated Infrastructure appeared first on Dell'Oro Group.