Perplexity and Nvidia have launched Portable Computer, a local-first AI agent that runs on Linux hardware.
According to VentureBeat, Perplexity and Nvidia are pitching it as a way to run AI models, files, tools, and workflows directly on an Nvidia-powered Linux kit. The idea is to keep routine work on the machine rather than punting everything to a cloud model with a meter running.
Portable Computer is a local-first version of Perplexity’s agent platform. Local tasks are supposed to carry no token charges, which should please anyone fed up with every AI button coming with a little invisible taxi meter. Data stays on the device by default. When the system thinks it needs a cloud model, it asks the user for permission before sending the job upstairs.
For Perplexity, that gives its agent platform a more serious privacy story. For Nvidia, which has spent the past two years selling the world on trillion-dollar AI data centres, the announcement signals something subtler but strategically important: the chipmaker believes local AI has crossed a threshold from hobbyist curiosity to practical tool and it wants to sell the hardware that runs it.
Nvidia has spent ages telling everyone that the future lives in giant AI data centres, but the Portable Computer plan suggests the firm still sees plenty of cash in beefy machines under desks.
Nvidia director of developer technology Nader Khalil, who works on developer tooling and open source, said local AI had changed. The old local model scene was mostly enthusiasts squeezing tiny models into shape and pretending the results were not occasionally daft.
“Local AI reached an inflexion point. For the longest time, it was hobbyists and enthusiasts, running these quantised models reduced to super tiny… And while that’s cool, it’s not super practical. But all that changed with a lot of these new open-source models that have come out that are super useful.”
Portable Computer launches for Pro, Max, Enterprise Pro and Enterprise Max subscribers on Linux. Windows support is due in September, which means the usual Linux-first crowd gets to smirk for a few weeks.
The hardware bar is not bargain-bin stuff. Any RTX GPU with at least 24GB of VRAM qualifies, which means roughly a GeForce RTX 3090 or newer.
Any RTX GPU with at least 24GB of VRAM, roughly a GeForce RTX 3090 or newer, meets the threshold where the two can deliver a great experience, but balance that with ensuring it’s broadly available.
That is “broadly available” in Nvidia land, where the floor still involves a GPU that many normal humans would consider a fairly serious purchase. Still, 24GB of VRAM is a lot less ridiculous than asking every user to build a private hyperscale bunker.
Whether Portable Computer becomes a serious local agent platform or another shiny AI launch depends on how well it handles real files, messy workflows and tool use without constantly phoning home.
