Nvidia Launches $249 Desktop AI Computer for Decentralized Projects

Nvidia CEO Jensen Huang pulled out a $249 desktop AI computer from his kitchen oven during a recent presentation, showcasing the Jetson Orin Nano Super Developer Kit. The device can run large language models locally without the need for cloud subscriptions.

The kit's processor pushes 67 to 70 TOPS (trillion operations per second) while consuming just 25 watts of power, roughly equivalent to a standard lightbulb. This is a significant improvement over its predecessor, with a 50% boost in memory bandwidth and double the performance for the same price.

Nvidia claims that this device can handle complex workflows previously only possible on its larger HGX AI supercomputers. The company sees this as an opportunity to support decentralized AI projects, such as Bittensor, Render, or Akash, which rely on distributed computing marketplaces and local inference for data privacy and decentralization.

This $249 entry point could dramatically expand the pool of participants in these networks by eliminating recurring API fees. For example, OpenAI, Anthropic, and Google charge per-token for their cloud-based models. A one-time investment in a device like this can run comparable open-source models locally, making it a fundamentally different economic proposition.