// hardware architecture

All signals tagged with this topic

Cisco and Nvidia push enterprise AI beyond the hyperscaler stranglehold

The AI infrastructure market is bifurcating. Hyperscalers no longer have exclusive access to the rack-scale hardware and software stacks needed to run serious language models and foundation models at enterprise scale. Cisco's networking layer plus Nvidia's accelerators and software are creating a path for mid-market companies to run their own AI workloads without renting compute from AWS, Google, or Microsoft. This threatens the hyperscalers' margin capture on inference and fine-tuning workflows. The shift redistributes AI economics: enterprises keep their data and model improvements in-house, while Cisco and Nvidia move from component suppliers to full-stack orchestrators competing directly with cloud providers.