Forgot your password?
Close
typodupeerror
AI Businesses Build

Raspberry Pi Stock Rises Over Its Possible Use With OpenClaw's AI Agents (reuters.com) 46

Posted by EditorDavid from the feeling-board dept.
This week Raspberry Pi saw its stock price surge more than 60% above its early-February low (before giving up some gains at the end of the week). Reuters notes the rise started when CEO Eben Upton bought 13,224 pounds worth of shares — but there could be another reason. "The rally in the roughly $800 million company has materialised alongside social-media buzz that demand for its single-board computers could pick up as people buy them to run AI agents such as OpenClaw."

The Register explains: The catalyst appears to have been the sudden realization by one X user, "aleabitoreddit," that the agentic AI hand grenade known as OpenClaw could drive demand for Raspberry Pis the way it had for Apple Mac Minis. The viral AI personal assistant, formerly known as Clawdbot and Moltbot, has dominated the feeds of AI boosters over the past few weeks for its ability to perform everyday tasks like sending emails, managing calendars, booking appointments, and complaining about their meatbag masters on the purportedly all-agent forum known as MoltBook... In case it needs to be said, no one should be running this thing on their personal devices lest the agent accidentally leak your most personal and sensitive secrets to the web... In this context, a cheap low-power device like a Raspberry Pi makes a certain kind of sense as a safer, saner way to poke the robo-lobster...
The Register argues Raspberry Pis aren't as cheap as they used to be "thanks in part to the global memory crunch. Today, a top-specced Raspberry Pi 5 with 16GB of memory will set you back more than $200, up from $120 a year ago."

"You know what's cheaper, easier, and more secure than letting OpenClaw loose on your local area network? A virtual private cloud..."
This discussion has been archived. No new comments can be posted.

Raspberry Pi Stock Rises Over Its Possible Use With OpenClaw's AI Agents More

Raspberry Pi Stock Rises Over Its Possible Use With OpenClaw's AI Agents

Comments Filter:

  • That is about the worst possible platform. But I guess too many people think LLMs are the second coming or something.

    • People believe LLMs and machine learning are the future. There is no doubt that the hype is there right now. That's where the market is at right now, even if it is a stupid place to be.

      You can run inference on just about anything if your model size is small enough (e.g. tinyllamas). An RPi is relatively powerful compared to some of the things we already run inference on. But these small systems running small models tends to be less impressive. You could do something simple like process log files in less tim

      • Re: Very. (Score:5, Insightful)

        by AvitarX ( 172628 ) <me@NoSpAm.brandywinehundred.org> on Monday February 23, 2026 @12:07AM (#66004840) Journal

        I thought clawbot used external LLMs and it's job was connecting them to your systems to do things.

        It's the agent part, not the AI part.

        • Yeah, you're right. I totally misunderstood what they meant by run locally. It's more like giving Claude or ChatGPT a remote shell login to your machine and letting it automate stuff. I'm guessing there are non-zero odds that it erases all your files, given AI's track record right now.

          • Re: OpenClaw seems boring (Score:5, Interesting)

            by AvitarX ( 172628 ) <me@NoSpAm.brandywinehundred.org> on Monday February 23, 2026 @12:55AM (#66004862) Journal

            I'm not doing it (just to clarify), but my understanding of what an agent does seems perfect for a raspberry pi.

            I could be wrong though, because the agent is taking LLM text and turning it into action, and honestly I have no idea how much CPU or GPU that takes.

            But the concept of the bot is that it's always on and you can text it.

            I agree with you though. I'm not ready to let go of control of important stuff to a bot controlled by an LLM. The failure mode seems extreme while the convenience meh.

            The fact that a bot wrote a blog post calling someone trash for rejecting a patch (that likely sucked) from the same bot tells me I don't want it.

          • To be fair...

            I have had several people in my circles assume they're 'running locally' in the same way they assumed bitcoin was 'decentralised'

            Part of the sales pitch obfuscates what this term actually means (because "it's constantly sending every keypress to Anthropic" sounds scary, even though that's how autocomplete has worked for 20+ years)... The pitch around the technology is, without prejudice, primarily intended for people who can't determine the quality of the output.

        • Absolutely!

          As you say, all the inference and heavy lifting happens on the remote LLM. I just had this discussion with ChatGPT earlier today. I asked it about running Claude Code and Yegge's Beads. It says the biggest issue is whether your agents and their local tools are compilable for Arm64.And both of those are.

          CPU and RAM needs are no problem for a Pi 5. Possibly a Pi 4 even. 4G of RAM is enough in a pinch. Disk I/O is the first place to watch for a bottleneck. And even that is not heavyweight exce

    • Re: (Score:2)

      by allo ( 1728082 )

      A Pi is (thought to be) a prototyping device. Nobody would want to deploy stuff on a Pi. Yes, they are often (ab)used for KODI, but in principle you should use them to learn python and let LEDs blink, and now maybe experiment a bit with AI without installing Linux on your computer.

      • Re: How stupid can you be? (Score:5, Informative)

        by madbrain ( 11432 ) on Monday February 23, 2026 @06:03AM (#66005092) Homepage Journal

        Wrong. There are plenty of production uses for low power computers in embedded applications. I have two Raspberry Pi 3B+ running headless attached to the RS-485 bus of each of my Carrier HVAC system (one furnace, one heat pump). The low power consumption is a feature. Compute power is not needed to run the Go infinitive service.
        And the Pi is indeed a computer. And Raspberry Pi OS is Linux.

        • Re: (Score:2)

          by allo ( 1728082 )

          There are better boards for low power tasks. Even in the same niche there is competition. Did you try orange pi or odroid?

          • Re: (Score:2)

            by madbrain ( 11432 )

            I tried an Odroid XU4 before, and there were lots of driver/kernel bugs. I had an Odroid N2+ at some point, but the USB connector failed. I can't say that I'm impressed with Odroid. I have not looked into Orange Pi. I'm just not familiar.

            The Pi3B+ idles at about 3W. It's fanless. includes everything that's needed. Parts for it are easy to find. I had spare SD cards for it.
            The Orange Pi 3B looks like it is superior hardware. But overkill to run a simple service like Infinitive. I bought the pair of Pi 3B+ fo

            • Re: (Score:2)

              by allo ( 1728082 )
              I think I had the XU with fan and regretted not getting the passive cooled one. In the end these devices are nice because they are small and draw little power, but if you would have some real use for such an agent, the better solution is probably a cheap VM somewhere. Currently I would just suggest to keep them experimenting with the agents and wait for the usable next generation after all people noticed what the flaws of the current approaches are. To me it looks like nobody seems to care about giving the

              • Re: (Score:2)

                by madbrain ( 11432 )

                I still have the XU4Q. But again, too many bugs in their ported distro. And it's 32-bit only, so now it's a paperweight with Linux having dropped 32-bit support.

                A VM has no applicability to my embedded applications. For AI agent, I agree a VM is OK. But the security problems with agents go far beyond VMs. They need access to the Internet to access remote LLMs. But that means they also browse the web, post stuff, and so on. One agent yesterday uploaded stuff to github as it was working, when I did not ask it

      • Oh man are you out of date with how Pis are being used these days. Yeah they were supposed to be fun little learning gadgets but they have deviated massively from the original path. Pis can be found in among other things: Building management systems, consumer products, DIN mounted industrial settings including IP rated Pis to do things like data aquisition, or IoT gateway devices in factories, and of course they are in many MANY kiosk or digital signage devices.

        They are very very far from the original proto

      • A Pi is (thought to be) a prototyping device. Nobody would want to deploy stuff on a Pi.

        A raspi is orders of magnitude more powerful than computers I've done a lot of real work on. Is it the best platform no, they make stupid mistakes with every version of the hardware. Is it more than powerful enough to do real work yes, of course it is.

    • Why? Are you confusing running an actual LLM with what OpenClaw is? Maybe you should understand software before you define hardware requirements for it out of ignorance.

      Hint: Openclaw is a gateway to other AI tools, it does no AI processing. You could run it on a TI84 calculator.

  • Since memory is now dominating the price of SBCs, an N150 is a better deal than a RPi 5. It will include more ports, an M.2 slot for an SSD, and have more compute power. It will also come with Windows if you prefer an inferior OS.
  • This is what democracy looks like. The humble Pi has done more for humanity than we are prepared to admit. You do not own enough SBCs.

  • You can buy an Intel or AMD-based mini-PC with 16Gb RAM for about $200 on Amazon that performs way better than a Pi 5 and can run Linux.

  • stopped looking at RPi a while ago now, price to performance wise they used to be awesome, now they are more expensive than cheap lowend mini PC's and NUC type devices that have considerably more compute and options.

    • If you need compute you're correct. But there are far more to a statement of requirements than just price / Gflop (or whatever unit is used in computing power). There's no NUC type devices as small as the Pi, there's no NUC type devices that run on 5V, there's no NUC type devices that run at the same low power.

      Also the cheapest NUC type device I found was still 8x the cost of a Pi, and then I need to buy a powersupply for it too.

    • I stopped looking at RPi when they gave their initial user base the finger during COVID so they could focus on commercial customers.
  • If I get an email from your 'AI assistant'? You're going in my spam folder. If I see a social media post from your AI assistant? You're getting an unfollow

Slashdot Top Deals

We all agree on the necessity of compromise. We just can't agree on when it's necessary to compromise. -- Larry Wall

Close