News: 0185407090

  ARM Give a man a fire and he's warm for a day, but set fire to him and he's warm for the rest of his life (Terry Pratchett, Jingo)

Nvidia Launches Free Tool That Links Idle Computers Into a Personal AI Data Center

(Thursday September 03, 2026 @05:00PM (BeauHD) from the beowulf-cluster dept.)


Nvidia has launched [1]PAIR , a free open-source tool that [2]links compatible computers on a home network so they can pool idle processing power for local AI inference and agentic workloads. "While the compatible devices are mostly Nvidia GeForce GPUs (PAIR works with RTX 20-series cards and newer, as well as RTX Pro GPUs and DGX Spark systems), Apple's M4 chips or newer will also work," reports The Verge. From the report:

> The key thing here is that PAIR uses your in-home systems when they're idle to avoid interfering with other tasks. And this disaggregated system of computers can work in parallel to chew through lots of processing requests -- which should be helpful for an agentic workflow that breaks complex tasks into smaller jobs. This should prevent large bottlenecks on a single GPU, and Nvidia says PAIR can adapt as devices join or leave the network -- including if a user does something like start playing a game on their desktop PC.

>

> [...] Nvidia says PAIR is secured by pairing all devices through a six digit code and then securing the channel via mTLS (Mutual Transport Layer Security), to create an encrypted communication line that's trusted in both directions between computers. The Nvidia PAIR beta is available today, with support for Windows, Linux, and macOS.



[1] https://www.nvidia.com/en-us/ai-on-rtx/personal-ai-router/

[2] https://www.theverge.com/ai-artificial-intelligence/989435/nvidia-pair-personal-ai-router-home-local-llm-compute-tool-rtx-macbook



imagine a beowulf cluster (Score:5, Funny)

by Anonymous Coward

imagine a beowulf cluster of these!

Whose work loads? (Score:4, Insightful)

by PPH ( 736903 )

> so they can pool idle processing power for local AI inference and agentic workloads

Where do I send my power bill to get it paid?

Re: (Score:1)

by DarkOx ( 621550 )

not on that; think about all the model propagation that has to occur. That is a lot of 20GB+ blobs to shovel around to endpoints.

We can fuss about over subscription etc, but at this end of the day this sort of thing shifts a lot of costs onto largely unsuspecting last mile ISPs too. I don't feel to bad for specutrum and comcast who will have peering with CDNs near their largest customer pools, but the few mom and pops and municipals that haven't been consolidated - well this kinda thing could suck for them

Re: (Score:2)

by PPH ( 736903 )

But bandwidth is free. Or so they tell me when they want me to load their tracking/advertising JavaScript.

Re: (Score:2)

by KiloByte ( 825081 )

Is your only connection a metered phone? Usage on the level of several TB per month on the lowest tier wired connection doesn't make an ISP angry, at least in the civilized world.

Re: (Score:3)

by sabbede ( 2678435 )

Well, since it says "local", I assume you'll pay your own power bill for your own inference and agents.

Re: (Score:2)

by fahrbot-bot ( 874524 )

> Well, since it says "local", I assume you'll pay your own power bill for your own inference and agents.

From the PAIR link in TFS:

> Private Local Inference Keep It Close to Home

> Run AI workflows at home with data that stays on your network. NVIDIA PAIR is built for private local inference, helping you use prompts, files, and agent context without sending them to the cloud.

I'm still dubious, however, of the pinky-swear promises from these companies that they won't somehow get your data.

Re: (Score:2)

by sabbede ( 2678435 )

Running it locally is probably the best you can hope for in that regard.

Good timing (Score:2)

by ufgrat ( 6245202 )

I've just retired the last of my NVidia GPU's because support, particularly under Linux, has become too painful, and new cards are too bloody expensive.

Got a ROCm version?

Re: Good timing (Score:2)

by drinkypoo ( 153816 )

It's sad how cuda is great but the graphics driver is poop.

Re: (Score:2)

by sabbede ( 2678435 )

As a personal favor to you, I'll take them off your hands. You only have to pay for shipping, unless you happen to live in Atlanta's southern suburbs.

Mattel x Skynet collab; Fyre and Forget. (Score:2)

by geekmux ( 1040042 )

> Got a ROCm version?

Wait..is that a..SoC design?

Huh.

One hell of a way to make a ROC'em SOC'em robot.

(Don't forget to tip your waitress-bot.)

Local SETI@HOME + Tailnet project (Score:3)

by nadass ( 3963991 )

It sounds like they're repurposing the SETI@HOME project's distributed computing processes by combining it with a local tailnet (private mesh network) whereby the main PC is the 'master' node for both functions -- unless the distribution is decentralized whereby it's akin to crypto mining with a decentralized tailnet.

Overall it seems like a neat idea... a private computing cluster.

Re: (Score:2)

by ffkom ( 3519199 )

But SETI@HOME dealt with completely non-sensitive data which was fine for everyone to observe, and even tampering with it would not have caused much damage. "AI", on the other hand, is currently the tool that most people happily hand over their most sensitive data to.

Re: (Score:2)

by sziring ( 2245650 )

Extraterrestrials might be sensitive, we just need to find them first.

grid computing for the home... (Score:4, Insightful)

by Ritz_Just_Ritz ( 883997 )

Meh. I'm going to pile on and say that most homes:

a. aren't going to have multiple higher end devices with fancy gpus or a number of Apple M4 or M5 devices with enough RAM "laying around" doing nothing

and

b. even if they did, they're not going to be willing to have them on 24/7 and sucking up electrons like nobody's business to crank on AI inference jobs. The power usage would dramatically go up along with their monthly electric bill.

I could potentially see this working at a smaller business if their fleet of devices was up to the task. Game development shop?

Seems like a real reach and a "solution" in search of a problem to solve.

Re: (Score:2)

by ffkom ( 3519199 )

The amount of noise coming from cooling systems could be enough to keep people from "sharing" their computing power. I agree this seems more like something a small company may consider to use.

Re: (Score:2)

by Mspangler ( 770054 )

A is true. The newest discrete GPU here is an AMD RX 960. Newest CPU is an M2. The Linux box is a Ryzen 4600G.

B is not a problem. As the man said, winter is coming. The house is all-electric and the mini split heat pump does not heat the bedrooms on either end. Those use resistors (baseboards). The cluster could be one M2, the 4600G, and the 2010 Mac Pro with the GPU and a six-core 3.33 GHz Xeon with 24 GB RAM.

Is that enough to do a useful job?

Back in 1995 I was running a neural network on a 100 MHz PPC 601

Re: (Score:2)

by Shades72 ( 6355170 )

If you pay for tokens from any cloud-AI provider or pay for power to run your own LLM(s) with this new software...I don't see that much difference to be frank.

In both cases I lose money and in both cases I have to hope that financial loss produced results that makes me more money than I spent.

Local is more Private and I happen to have a large roof...well, more than one, completely filled with solar panels. And I live in a place with a lot of sun-hours, so on sunny days I really don't feel that much of a fin

Boom for Corporations--Personal Computer Nightmare (Score:2)

by BrendaEM ( 871664 )

I would not be surprised if Windows is already doing this in the background.

If you lose your temper at a newspaper columnist, he'll get rich,
or famous or both.