The Amazing Raspberry Pi Story

Eben Upton gave a spellbinding keynote at ARM TechCon on the history of Raspberry Pi. He is a founder of the Raspberry Pi foundation and also an ASIC designer at Broadcom. Unusually for a keynote like that, he had no powerpoint. He just stood on stage and talked for forty minutes about the amazing story of how Raspberry Pi came into existence. They expected to sell, maybe, ten thousand. They have sold ten million and counting. They easily sell 10,000 every day.

In case you don't know already (where have you been hiding?), the Raspberry Pi is a credit card-sized computer that sells for about $30. The current version contains a quad-core 64-bit ARMv8 CPU with 1GB RAM. It has 802.11n wireless LAN, Bluetooth LE, 4 USB ports, HDMI, audio jack, camera interface, micro SD slot, and more. It is a fully configured computer without all its peripherals.

You can buy it on Amazon (Prime eligible, one click and you can have one in two days).

Here is a picture of the current Raspberry Pi model 3 (my business card for scale):

The Story Begins: Who Will Study CS at Cambridge?

Story starts when Eben was director of studies at St John’s College, Cambridge. That role is to organize undergraduate teaching at one of the 30 colleges, and also ensure a supply of undergraduates for the future. Every December, a lot of high school students come in to enquire about the computer science (CS) course.

When Eben went to Cambridge in 1996, they had a typical over-subscription ratio of about 6:1. So to get on the course, you had to beat out 5 other people, people with a deep understanding, who had probably been programming since they were 10 on computers like the BBC Micro Computer (yes, BBC is the British Broadcasting Corporation), which is what he grew up on. These are kids who bought their machines, or got given them, to play games on. But they all had one thing in common. When you turned them on, they gave you a programming prompt. So if you wanted to play a game, or do your homework, the first thing you had to do was choose not to program. Almost everyone Eben knew could write at least simple programs.

And somewhere along the way, Cambridge got fat and happy on this funnel of intelligent, talented kids. Cambridge University is famous for its short terms (made up for, I have to point out, by having six day weeks with lectures on Saturdays) meaning that a three-year computer science course just has 60 weeks to turn people into professional computer scientists. So it was very useful to be able to rely on a certain level of ability when high school students came in the door (“matriculated” in Cambridge-speak). In fact, the first problem was to convince them they didn’t already know everything. Standard ML functional programming was perfect for this, Eben joked (originated at my other alma mater, Edinburgh University CS).

By the time Eben did his PhD and got to be director of studies and was interviewing people, the picture had changed. The university had gone from having 500 applicants for our 80 or so places, to more like 200. There were still plenty of smart, intelligent kids but they lacked the built-in hacker knowledge of what the machine is doing, so the university would spend the first part of the 60 weeks building students up to a level that they’d been able to assume. I studied CS at Cambridge, too, but in those days it was a one-year course in third year, with only with about 20 students, and I was unusual, in that mainframe era, since I had been programming since I was 14.

Back in 2006, Eben and a few others started to wonder why this had happened and if there was anything they could do about it. One theory, and it is still a theory, is that those 8-bit machines that everyone had had as children had gone away and were replaced, from below, with game consoles whose business model required them not to be programmable, and from above by the PC. Of course, the PC is programmable, but you have to choose to program it, to get the tools and documentation. It doesn't stare you in the face every time you turn the computer on.

They felt that there was a niche, a niche for things to go in kid’s bedrooms. The four attributes it would need were:

1. Programmable, obviously
2. Had to be interesting, able to do other things than just programming (that meant games, graphics, video)
3. Robust (capable of being throw in a schoolbag day after day)
4. Cheap (benchmark was a school textbook, so picked $25…which shows how little they knew about how much textbooks cost)

Around then, Eben went to work for Broadcom. Mobile chips have a lot of processing power so they went that route, and by 2008 they had a device. It had a closed proprietary RISC core on it, with CPython programming language ported to it. It was close to what they wanted: a machine that booted to the programming prompt of a widely used educational programming language.

The group decided to set up a foundation to promote it. It got named Raspberry Pi. The “pi” came from the programming language Python (and they thought it would make a great logo, although in fact they have never used it, the actual logo is a raspberry). The “raspberry” came from the tradition of fruit-named computer companies. At the time there was not just the famous one in Cupertino, but also Apricot, Tangerine, and Acorn (the fore-runner of ARM, and technically a fruit).

But there was a big problem. The group had to write everything themselves: network drivers, SD-card drivers, and so on. It was too much with too little resource.

Fortuitously, in 2011, Broadcom produced a version of the chip with an ARM-11 in. They suddenly realized this was what they wanted. ARM was an open platform, something that could run Linux, so they would be able to leverage the huge investment of other people.

But they were still thinking in a very parochial way. Their dream was just another 250 computer science applicants a year at Cambridge. However, many people on the foundation had grown up with the BBC Micro and thought, nostalgically, it would be cool to put a BBC sticker on it. They had a brand, Raspberry Pi, that they never wanted but they kept going to the BBC and trying to get them to adopt it as the BBC Nano. Auntie (as the BBC is known in Britain) kept saying no for various legal reasons.

The Video

Their last attempt was in May 2011, when went to see Rory Cellan-Jones, the senior technology correspondent for the BBC. They said “can we put a BBC logo on it?” and he said “no.” But “I think it's a great idea,” he said. “Can I take a video of one of you holding up the prototype and talking about it?” So they made a two-minute video of one of the foundation holding up something—that looks nothing like the eventual Raspberry Pi—and talking about the project.

It was their "Turn again Dick Whittington, Lord Mayor of London" moment. That video (above) got 600,000 YouTube views in two days.

For a couple of days, Eben was on cloud nine and then it hit him. “We’d promised 600,000 people we’d build them a $25 computer, and we had no idea how to do it.”

If you add up the cost of the chips, it looks OK, but there are 180 components on the Raspberry Pi, only two are the chips. They spent the year working out how to get the cost down to $25. They scraped together $250K of capital, mainly from high-net-worth individuals in the Cambridge area, enough working capital to build 10,000 Raspberry Pi boards.

But they were getting worried. The traffic to their website kept going up. They put a buggy image of the software that you could run in QEMU (there was no hardware yet) and 50,000 people downloaded it. Remember, this is alpha quality code for a computer you can’t buy. So 10,000 units would only last a week and they take three months to build so…not good.

Manufacturing Begins

Two companies, Premier-Farnell (as of October this year, part of Avnet) and RS Components, shared the same concern about lack of computer literacy. They agreed not just to distribute them the boards, but to manufacture them. They would provide the capital and the first-line customer support. Presumably they thought they were making a philanthropic decision, but it has turned out to be a great business decision, too.

So Raspberry Pi switched from a tiny capital-constrained foundation trying to manufacture, into an IP licensing company. Nobody in Cambridge has ever had that idea before!

They took 100,000 orders on the first day it was on sale, February 29, 2012. So the answer to how long would 10,000 units last was “About an hour.” They would probably have sold more if all the websites had stayed up,

On February 29 this year, their fourth birthday (or first if you take the February 29 thing as seriously as The Pirates of Penzance), they announced the model 3.

On September 8, they announced that they had shipped the ten millionth Raspberry Pi, beating their wildest dreams of 10,000 by three orders of magnitude. As Eben puts it on the Raspberry Pi website:

When we started Raspberry Pi, we had a simple goal: to increase the number of people applying to study Computer Science at Cambridge. By putting cheap, programmable computers in the hands of the right young people, we hoped that we might revive some of the sense of excitement about computing that we had back in the 1980s with our Sinclair Spectrums, BBC Micros, and Commodore 64s. At the time, we thought our lifetime volumes might amount to ten thousand units—if we were lucky.

Today

Today, the focus of the Raspberry Pi foundation is more educational. They take the royalty money and re-invest it in teaching materials, training courses for teachers, and so on. In November 2015, they merged with Code Club.

Eben said there are two big lessons he draws from the Raspberry Pi story:

• The community is really important. With just a few of them, they couldn't do much. They needed to build a critical mass of contributors around the product to become successful.
• Standards. Without standards you won't build a community, and you can't leverage the work of others. For example, by running Linux on ARM, they are benefiting from work done by a lot of professionals in companies like IBM, Red Hat, and ARM themselves.

By the way, the BBC did eventually launch its own computer project too, which is given free to all year 7 pupils in Britain (11-12 year olds). For that story, see I Danced With a Nun in a Disco...And the micro:bit.

Next: Who Wrote the Book on Formal Verification?

Previous: Automotive Security: A Hacker's Eye View