Raspberry Pi founder Eben Upton: ‘I’m an omni-geek’

Not many Lunch with the FT guests would suggest skipping dessert at a fine restaurant and grabbing a Welsh cake on the run before conducting a whistle-stop tour of an electronics factory. Then again, not many interviewees are as obsessed as Eben Upton is about his life’s work: the creation of Raspberry Pi, one of Britain’s most successful listed technology companies, which has sold more than 75mn of its low-cost computers.

But the 48-year-old Upton is not just a hard-charging entrepreneur intent on proving that a European tech company can compete against US and Chinese rivals in the hyper-combative era of AI. He is also a former Cambridge academic and philanthropist who is passionate about opening up access to the wonders of computing to everyone. To that end, he and several other co-founders launched the Raspberry Pi Foundation in 2008. With additional outside donor money, the foundation has spent £112mn since its launch up to the end of last year.

Initially, this charity fully owned its commercial arm and retains 41 per cent of the spun-off manufacturing company today. The income that the charity has received over the years has enabled it to put 2.3mn young people through coding clubs and train more than 250,000 teachers. But the value of the endowment is still growing strongly as Raspberry Pi’s shares have risen 40.5 per cent since floating on the London Stock Exchange in 2024.

“As you’re probably going to learn over the next hour or two, I’m a massive geek,” Upton says at the start of our lunch. “I’m an omni-geek. I can be a geek about anything.”

He is not wrong. As the FT’s technology columnist, I am well versed in geekspeak, yet I struggle to follow Upton’s deepest dives into subjects as varied as the pronunciation of vowels, silicon chip design and AI’s asymptotic trajectory.

Upton dates his interest in geekery to his earliest childhood in Papua New Guinea, where his father taught English for a time. He remembers being transfixed by a repairman fiddling with a control box on the wall to fix a ceiling fan. “My first ever memory was of electrical engineering,” he laughs. But Upton has a far broader range of interests than most geeks, with musical tastes spanning Biffy Clyro to Johann Sebastian Bach. He also has the advantage of being able to speak human.

Upton has chosen to meet at the Roots restaurant in the Llanerch Vineyard Hotel in south Wales, not far from his mother’s hometown, where he was born. It is evident that this part of Britain, at least, has been blessed with some rain over the scorching summer. The famous Welsh valleys are refreshingly green.

It is drizzling when I arrive at the winery, which means we cannot take advantage of the terrace overlooking the grape-heavy vines. But the airy restaurant is welcoming enough.

Upton arrives late, thanks to the vagaries of the British rail network, dressed casually in a green jacket and a white knitted T-shirt. He is a bundle of good humour and restless energy, his voice booming across the half-empty restaurant.

It does not take him long to recount how he fell in love with computing as a child, graduating from Lego to a BBC Micro computer on which he built games, like so many kids of his generation. He was mostly self-taught, devouring Usborne’s famous computer and coding books. But he also attended a computer club every other Friday night in his then hometown of Ilkley, near the University of Leeds, where his father had moved to take up an academic post. “You had all this kind of anarchic infrastructure that generated an enormous cohort of people who had some experience of computer programming,” he says.

I did the software that generated the rough cut of the dictionary. And then my father did the practical implementation

I did the software that generated the rough cut of the dictionary. And then my father did the practical implementation

Upton’s father Clive, now an emeritus professor, is an expert in dialectology, reminiscent of Professor Henry Higgins in My Fair Lady. Like his fictional counterpart, Upton’s father would meticulously record how English was spoken in everyday life. He would then update the phonetic transcription of words in dictionaries for the benefit of BBC news presenters, among others. So, for example, “het” became “hat” some time in the 1980s. “My father invented the modern notion of received pronunciation,” he says.

Combining their respective interests, Upton father and son co-wrote the Oxford Rhyming Dictionary, published in 2004, containing more than 85,000 words. “I did the software that generated the rough cut of the dictionary. And then my father did the practical implementation to turn it into something you could actually use,” he says.

Upton admires the ways in which some poets have played with words, even when the rhyming is a little questionable. He recites Rupert Brooke:

“Ah God! to see the branches stir
Across the moon at Grantchester!”

This poetic interlude ends when our waitress asks what we want to eat. Upton goes for the carbonara arancini and Welsh lamb rump with Pembrokeshire baby potatoes. I opt for the Roscoff onion tarte tatin with smoked Caerphilly cheese and the seared fillet of hake in a massaman curry sauce. Upton sticks with sparkling water. But I cannot resist sampling one of the vineyard’s wines and choose a flight of their Solaris white, with hints of Granny Smith apples and lime zest.

Upton entered the University of Cambridge in 1996 to study physics, regarding computing as an enjoyable side hobby. But he later switched to engineering and completed a PhD, becoming director of studies for computer science at St John’s College. In that role, Upton was dismayed by the sharp decline in the number of applicants to study the subject since his own hobbyist days.

Whereas he and his cohort of friends had first learnt to programme on simple home computers such as the BBC Micro, the next generation could use only comparatively “terrible platforms”. From that realisation, the seeds of Raspberry Pi were formed: meeting the need for cheap and accessible programmable hardware. Fruits have become something of a naming tradition in the tech world (Apple, BlackBerry) although, as Upton points out, it is unusual to have a raspberry pie.

With his wife Liz, a Cambridge-educated lawyer, and several other co-founders, Upton launched the Raspberry Pi Foundation to “create a copy of the 1980s” computer culture. The availability of credit-card-sized, single-board Raspberry Pi computers, initially selling for as little as £25, helped spark a near cult-like revival of interest in programmable computing among schoolchildren. Last year, computer science was the most competitive undergraduate course at Cambridge, with only 7 per cent of the 1,739 applicants being accepted. Mission accomplished, it would seem.

However, Upton and his wife realised that the foundation’s initial approach was mostly helping a narrow, self-selecting set of students from relatively privileged backgrounds. “It’s very exclusionary. You just end up again with only 10 per cent of software engineers being women,” he says. “Without some sort of uniformity of provision, you get this weird demographic skew. Broadly, it’s the people who play computer games.”

Under the direction of Philip Colligan, who took over as chief executive in 2015, the foundation set about addressing that challenge. Its focus switched to teaching a far broader cohort of students to code at school. That has left Upton to focus mostly on the commercial business.

The foundation, which was endowed with £136mn when Raspberry Pi went public, now devotes most of its resources to running coding clubs, providing supporting materials for the schools curriculum, teacher training and research. The foundation is also expanding internationally, especially in India. “We can, should and must play everywhere to have a global impact,” he says.

As the table is cleared, Upton tells the waitress that his arancini were “wonderful”. The noise level in the restaurant rises as more guests arrive for lunch. My delicious mini glass of Welsh wine does indeed have an appley taste and perfectly complements my hake.

As a provocation, I make the argument that it is a waste of time for children to study coding in the age of AI. Nowadays, anyone can “vibe code” their way to success: you can type in plain text instructions and an AI model automatically generates the code. Upton, of course, kicks back strongly. Citing a report the foundation published last year, he argues that coding is the “modern form of literacy”.

Upton also draws an important distinction between coding, which instructs computers what to do, and programming, which formulates a problem in a computable form. Coding enables users to understand that software is an artificial and fallible construct that can be manipulated. As they say in the software industry: “program, or be programmed.” Unless you can understand how AI works, you may be unwittingly manipulated by unscrupulous companies or governments, he says.

AI is fantastic news for anybody who makes cost-effective hardware for the edge of the network

AI is fantastic news for anybody who makes cost-effective hardware for the edge of the network

Now we are on to AI, I ask Upton what he makes of our latest technological hype cycle. It quickly becomes clear that his views diverge wildly from the Silicon Valley consensus. The people who run the big AI companies in the US, he says, appear to have a near-religious belief that their models will quickly and inevitably become superhuman. “Most of them have this kind of slightly messianic feel to them,” he says. But Upton is more of a technological agnostic.

As he tucks into his lamb, he acknowledges the impressive capabilities of generative AI models. He suggests that AI may boost trend GDP growth by 1 percentage point over a couple of decades, which would be a big deal. But he is also unusually forthright in highlighting the AI models’ limitations. He doubts that they are going to “scream away to infinity” or achieve God-like powers. The “perimeters of productivity”, as he puts it, may be more bounded than many believe.

As a “mulish engineer”, he questions how well AI models are able to interact with the real world. “They are simultaneously amazing and disappointing. They are brutally limited in terms of the quality of their output,” he says, when it comes to coding.

A fan of S-shaped sigmoid curves, Upton speculates that AI models will plateau and may never achieve the superintelligence that has become such an article of faith on the US West Coast. In that sense, human intelligence may well prove to be an asymptote, a fixed line that is never breached by the AI curve.

He appears deeply sceptical of the Silicon Valley evangelists, intent on creating a machine God. “It can be a little bit hard to tell whether they’re talking this way because that’s how you raise money or whether they’re talking that way because they really believe it. But I think there are a lot of them for whom it’s both.”

Still, Upton says it is important to remember that it is less than 1,500 days since the release of the ChatGPT model by OpenAI. People are still flailing around trying to understand the most useful applications of these chatbots.

What is more certain, Upton argues, is that technology’s “centrifugal forces” will massively increase the need for AI at the edge of computer networks. That should eventually be good news for Raspberry Pi, whose devices are used for all kinds of everyday purposes. As simple AI models become far more efficient, they may no longer need to use the super-expensive graphics processing units (GPUs) designed by the US chipmaker Nvidia and may be able to rely on the old-fashioned central processing units (CPUs) now manufactured by Raspberry Pi. Whereas he had previously forecast that this evolution would take several years, he now suggests it may happen in months.

The company’s adaptable computer cards are already being used for countless industrial applications, including running many of the giant video screens at Heathrow airport. They are also being developed by doctors to spot skin cancer and by farmers, who can mount image-recognition software on drones to identify the driest parts of their fields. “AI is fantastic news for anybody who makes cost-effective hardware for the edge of the network,” he says.

Britain, during the Industrial Revolution, ‘was the only place to have got a durable decoupling of money and power’

Britain, during the Industrial Revolution, ‘was the only place to have got a durable decoupling of money and power’

Upton says that Raspberry Pi has been gradually shedding its image as a “toy company”, only popular among students and computer hobbyists. Indeed, some investors are even describing Raspberry Pi as a mini Nvidia, the world’s leading chip designer and most valuable company. But that seems a little premature. Upton winces when I note that Nvidia’s current stock market value is more than 4,000 times bigger than Raspberry Pi’s. But he jokingly counters with an alternative metric: Raspberry Pi boasts more followers than Nvidia on the Reddit online forum, suggesting it has a bigger fan base of users.

Upton draws inspiration from Britain’s long history of technological innovation. Britain’s gift to the world was to show how it was possible to separate commerce from politics during the Industrial Revolution, boosting living standards. “It was the only place to have got a durable decoupling of money and power,” he says.

His biggest worry today is that money and power are re-coupling as technology captures politics. On the West Coast, gloomy tech types talk about their p(doom) scores, the estimated probability of a runaway AI exterminating humanity and turning us all into paper clips, to use the philosopher Nick Bostrom’s famous analogy. Those fears have only heightened since the former Anthropic researcher Jacob Coxon this month warned that there was more than a 10 per cent chance that AI could wipe out humanity within a decade. But Upton fears that the political threat of tech-enabled authoritarianism is greater than that of a rogue superintelligence. “The p(doom) is probably not a paper clip maximiser. The p(doom) is probably Russia,” he says.

On a happier note, Upton suggests that we see how Raspberry Pi computers are made. I readily agree. I love a good factory visit. So, we end our lunch early, jump into a car and head to the nearby Sony UK Technology Centre, which doubles up as Raspberry Pi’s manufacturing plant.

Officially opened by Queen Elizabeth II in 1993, the plant once manufactured television sets, but no more. The factory still produces high-end Sony video cameras but it has now turned over much of its capacity to Raspberry Pi. In these parts, Upton is regarded as a local hero for having preserved so many jobs.

That said, the vast and spotless factory floor is dominated by robotic production lines. Rather neatly, one of them is set to manufacture a computer card every 3.142 seconds, a number that is the decimal approximation of Pi.

Upton says that Raspberry Pi began manufacturing its first computers in China but reshored production, significantly improving the interplay between design and manufacture. “People have the wrong idea about China. People think you cannot compete with them in high volume. But with robotics it does not matter where you are,” he says.

Recommended

Throughout his career, Upton has succeeded by proving people wrong. Not only has he achieved the rare feat of building a flourishing technology company listed in London with a manufacturing plant in Wales, he has also shown that some tech bosses care about far more than profits.

One other difference with China and the US also becomes apparent at the end of our factory tour. The Pencoed plant serves up a mean Welsh cake.

John Thornhill is the FT’s innovation editor

Find out about our latest stories first — follow FT Weekend on Instagram, Bluesky and X, and to receive the FT Weekend newsletter every Saturday morning