Why there might not be enough Raspberry Pi to go around



The incendiary prediction

I'm starting this post off with an incendiary statement. You're probably not going to get your $25 linux computer. You should know that I have absolutely no hard information to back any of this up. This is all rampant speculation on my part, mixed with a healthy dose of cynicism, and just a pinch of disappointment from past hopes dashed. I don't know anyone at the project. I don't have any news other than what's publicly available. I hope I'm dead wrong about this.

If you are at all interested in hardware hacking, and don't know what a "Raspberry Pi" board is, then you really should go check it out. The short version is that its a $25 embedded linux computer based on a Broadcom ARM11 System-On-Chip. It sports video and audio out as well as usb and even ethernet (on a model that will cost $10 more). These systems exist only as prototypes at the time of this writing. They've made 100 prototypes, which they plan to auction to the highest bidders. They are not particularly unique in this space. There are many, many, many, many, other products out there that have similar capabilities.

A victim of its own success

This board looks like a great product. Seriously, it looks amazing. In fact, it looks like exactly what I've been trying to get my hands on since I started working with embedded linux more than 7 years ago. Its strong enough (enough ram, expandable storage) to be useful and cheap enough to make its use practical in a dizzying array of applications. This is where I think they are going to run into problems. My prediction is that they are going to underestimate the demand by multiple orders of magnitude. They'll prepare themselves to meet demand in the thousands, maybe even the tens of thousands. They should be preparing for demand in the hundreds of thousands, perhaps even the low millions. I suspect that they'll never even know what hit them. Based on the performance of a few other "open source" hardware micro-manufacturers (including one I've been directly involved in), I predict that they'll never really plumb the depth of this demand. What will happen is that they'll lack the resources (or the courage) to produce more than a few thousand at a time. The experience from the outside will be months of "out of stock, back-ordered" waits, punctuated by a few minutes of frantic sales depleting each production run as it arrives.

That's two bold predictions. First, there's going to be outsized demand for this product and second, the producers of this product, who are a non-profit from the UK focused on education will not be able to rise to the challenge presented by this demand.

"You're gonna need a bigger boat"

The demand is going to be huge. Seriously. Huge. The magic isn't the machine. Its the price. I've been using embedded systems for the better part of a decade now. They cost $100.(1) You can spend more, of course, but you really can't spend less. About the best you can do today is a "dev kit" from Globalscale called a SheevaPlug. This is a tiny embedded computer built into a wall plug. Its a neat product, but at $99 a pop, there's a limit to how small a project you might use one for. The RasPi is 4 times cheaper. Consider that for a moment. Ask the folks at HP what happens when you take a competent product that hardly gets any attention and make it 4 times cheaper than the only other alternatives.(2)

Where will all of this demand come from? I suspect that the maker community alone will consume hundreds of thousands of these provided they are able to get their hands on them. As this is written, an Arduino Uno from Newark is $25. Unless you need those few extra IO pins or an A/D converter, it seems to me that it would be much easier just to use a RasPi for the same price. In fact, with a bit of community support, I can't image that there won't be all sorts of "sketch" like software bundles that can simply be dumped onto a RasPi's sd card that let it do whatever an Arduino can plus much more. Even if you do find yourself needing the extra IO provided by an Arduino board, the perfect companion to Arduino is a RasPi. You can get a RasPi and a $6 USB wifi dongle from Amazon and put your Arduino on the internet for about a third of the cost of simply buying a "WiFly" shield for your Arduino. No fiddling with embedded code to get it to talk to your new WPA2 protected 802.11N router required. As the community grows, I would not be surprised to find "shield" like addons(3) that use the RasPi's GPIO to extend its capabilities even more cheaply than using Arduinos. Numbers? Massimo Banzi, part of the Arduino team, put the number of official Arduinos in the wild at 300,000 as of May 13, 2011. That's just the official ones.

Much more than that, I've started to look back at all of the "someday projects" I've been thinking about forever but never doing because they weren't quite practical. A simple example? I've always wanted to be able to open my garage door with my iPhone. This is a trivial project for any motivated geek but collecting an always-on web server, a relay board, a wireless interface to get it to the door opener itself and the time to learn the different protocols needed to interface all of those parts was a hump just a touch too high to ever get motivated to climb. My previous plan involved a Sheevaplug, Xbee modules, an Arduino and a relay. Total cost about $250. Just a bit much even for the awesome of having my phone open my garage door. If I could buy a RasPi and a wifi dongle, the web server could run on the RasPi and the code would be a matter of a 3 liner in PHP to echo a 1 to the GPIO, wireless would a non-issue and the whole mess would cost less than buying a replacement clicker from Sears. As I skimmed down my list I had two dozen more eureakas of equal magnitude. The RasPi and a wifi dongle was the missing piece in every one. Wireless webcam. Yep. Wireless motion sensor, just add Zoneminder. Web enabled switches for lights? RasPi and wifi is cheaper than the cheapest X10 or Insteon setup and tons less hassle. Remote control stuff? Why buy a transmitter and receiver just to move a few servos? RasPi and wifi is cheaper, you get to use your iPhone, and there's no limit to the number of channels you can send. Need a brain and wireless link for you world domination robot? Its the brains for all of those gizmos your RepRap is going to pump out. You get the idea. Everything you can do with Arduino, you can do ten times better with RasPi for the same price. The Makers are coming.

Beyond Makers, there are a few other killer apps that will send demand though the roof, starting with IT and computer security. In a previous life I was the co-founder of a company that sold software as a service "in a can". We essentially provided adaptable network security by placing small embedded linux severs at key points in the network so we could see what was going on and control network flows. It worked fabulously. We used hundreds of them. The only thing that prevented us from putting one on every single office hub or switch (or even PC) that connected to the network was the $280 price tag. If they had cost $25 instead, we'd have used thousands.

RasPi's are the ultimate IT Swiss army knife. Cheap enough to be disposable, you could carry a few with you everywhere and not worry too much about leaving them behind. Need a wireless print server for that old printer? RasPi is cheaper. Need a wifi bridge for your Xbox? RasPi is cheaper. Wireless file server? Yep. Media streamer? Yep. Pandora player? Yep. The list is endless. All you need is a handful of RasPi's and a few SD cards preloaded with various functions. They'll simply morph into whatever bit of IT goodness is required for the task at hand.

But all this is peanuts compared with the demand that will spring up once Android makes it onto the RasPi. Its a $25 android device with HDMI out. Pair it with a dirt cheap Chinese resistive touchscreen (I've seen 5 inchers for as low as $30) and it becomes cost effective for geeks everywhere to touch-screen and androidize everything from their fish tanks to their alarm clocks. The internet of things doesn't come from some giant company blessing your washing machine with a finicky protocol that only talks to their servers and feeds you data though their portal. It comes from millions of geeks everywhere doing it themselves because its just recently become cheap enough and easy enough. Communities form and open protocols develop. The marketplace keeps the whole community loosely united and Andriod explodes into, well, everything. Its a great big ball of win.

But you won't be getting one.

If you do manage to get one, I fear it will be just that. One or two. Only then, after a few weeks of back-ordered uncertainty. There's magic in a $25 linux computer, but the magic only works if you can repeat it. It doesn't matter how cool the $25 thing you can make with your RasPi is if you can only make one. I can make one with my $99 SheevaPlug, it costs more but I don't have to wait weeks to get one. My novelty only cost $74 more. If you can keep making them and everyone else can too, it goes from novelty to revolution. Communities can be supported and the awesome things that can be done will grow daily.

The RasPi platform shares a great deal in common with Arduino. Its openness and community focus are very exciting. There is however one aspect that is different. The design is still essentially closed. If the Raspberry Pi foundation can't meet the demand, no one will in the short term. The temptation will be enormous for them make a few at a time, congratulate themselves on their success as they sell out over and over, and leave most of the demand unmet. They are a non-profit but it is also possible that when they realize how much unmet demand there is they will be tempted to raise the price. That would be sad indeed. By the time the price doubles, the magic will be gone. It would be easy to use flagging demand at that point to declare the product a flop and stop producing altogether.

I hope I'm right and wrong

I hope very much I've underestimated these blokes across the pond and they'll meet the demand with the steely resolve (and the means) to make as many as it takes. I hope I'm right about the demand because it will mean that a community will form that will rival and perhaps even surpass the Arduino community. The beginning of the internet of things. Not GE and Intel's things, everyone's things. As for me, I'll be placing my order for 20 units as soon as they go on sale and keeping fingers firmly crossed. A houseful of linux for the cost of a netbook.


(1) The cost of an "embedded" version of the big computer on your desk (one that can run essentially the same software, just not as much at once) has remained remarkably stable over the years. Sure its possible to run "a web server" on a $20 micro, but if you want to apt-get apache, you've had to pony up.

(2) Every once in a while a new bit of vapourware floats by thats more photoshop than product. People see the falling price of ARMs and can't help rushing to announce the ultra-cheap computer they are going to build "next year". Its certain that someone will eventually build such a thing, I'm just really hoping that eventually is now and RasPi is it.

(3) This has already happened, announced while I was writing this.