Perplexity and Nvidia partner for local-first AI platform

Summary

Kate Rooney breaks down NVIDIA and Perplexity's local-first AI partnership using NVIDIA's DGX Spark to run AI directly on desktop devices rather than through the cloud. She explains the pitch around lower compute and token costs and improved data privacy, while noting frontier tasks still require cloud use. The segment also covers Apple and Microsoft local AI hardware pushes and reports that NVIDIA may be looking to buy Perplexity for about $30 billion. Rooney then discusses OpenAI's custom chip benchmark claims against NVIDIA and potential implications for OpenAI's IPO and Nvidia competition.

• NVIDIA and Perplexity announced a local-first AI platform on DGX Spark.
• The platform aims to reduce compute and token costs and keep sensitive data local.
• Frontier reasoning tasks still require cloud use with user permission.
• The local platform launches on Alibaba's Qwen model and will later add NVIDIA's open-source model.
• Apple and Microsoft are also pushing local AI hardware.
• NVIDIA is reportedly looking to buy Perplexity for about $30 billion; no company comment.
• OpenAI claims its custom chip is more performant and energy efficient than NVIDIA's counterpart.
• Rooney says OpenAI's chip could become a revenue stream and is potentially bullish for its possible IPO.