Adel El Hallak, Nvidia's VP of Product, told Tom's Guide he keeps a DGX Spark at home that runs AI jobs overnight, treating it like an appliance. He said local AI could change how people use AI, since running a model on hardware you own keeps data in your house with no monthly subscription. Nvidia's DGX Spark retails for $4,699, while a ChatGPT Plus subscription runs $20 per month, or $240 per year, totaling $1,200 over five years without ownership; the Spark runs unlimited queries with no per-token billing, no usage caps, and no internet required. The article notes GPT-6 and Claude are significantly more capable than open-source models run locally on a Spark, but says a local model with 128GB of memory and access to Nvidia's full CUDA stack suffices for tasks like summarizing documents, drafting emails, searching your own files, and running a coding assistant. The DGX Spark launched last October as what Nvidia called "the world's smallest AI supercomputer," aimed at developers, researchers, and data scientists; it runs Linux, weighs 2.6 pounds, and is the size of a Mac Mini.