Perplexity, Nvidia partner to run AI directly on desktop instead of major cloud providers

Summary

Kate Rooney reports on Nvidia and Perplexity launching an AI platform that runs locally on desktop hardware instead of major cloud providers. The pitch is lower compute and token costs, with frontier tasks optionally routed to the cloud. The system will initially use Alibaba's open-source Qwen model and later Nvidia's open-source model, with no mention of Anthropic or OpenAI.

• Nvidia and Perplexity announce an AI platform that can run entirely on local hardware.
• The platform runs on Nvidia's desktop-sized AI computer designed for large models.
• The cost pitch is lower compute and token costs versus cloud data centers.
• Frontier-level reasoning tasks can be routed to the cloud with user permission.
• The system initially uses Alibaba's open-source Qwen model, then Nvidia's open-source model.
• Anthropic and OpenAI are not mentioned in the implementation.
• Apple is cited as another local AI example with its newest Mac Mini.