Nvidia squeezes a petaflop of AI compute into thin Windows laptops
RTX Spark pairs a 20-core Grace CPU with a Blackwell RTX GPU over NVLink-C2C, with up to 128GB of unified memory — no desktop tower required.
Nvidia's RTX Spark platform brings genuinely heavyweight local AI compute to thin-and-light laptops and compact desktops, linking a 20-core Grace CPU to a Blackwell RTX GPU over a high-speed NVLink-C2C interconnect — up to one petaflop of AI compute and 128GB of unified memory in a form factor that doesn't need a desktop tower.
That unified-memory figure is the headline for developers: it's large enough to run serious local inference workloads, including sizable open-weight models, entirely on-device — no API calls, no data leaving the laptop.
As open-weight models keep shrinking their effective footprint and platforms like RTX Spark keep growing local compute headroom, the two curves are starting to meet in the middle — local, private, laptop-class AI inference is quietly becoming a mainstream option rather than a hobbyist niche.