Graphcore – A Bristol client, designing chips for an AI-first world
The world now relies on computers, through AI, to solve complex problems that would be impossible for humans to even approach. Once the world has returned to the ‘new normal’, it is this technology that will play a major role in the recovery and return to long-term growth. But machine learning, the basis of any current AI system, requires the processing of huge amounts of data; in turn necessitating computing power which, when Nigel and Simon first came up with the idea for Graphcore eight years ago, simply wasn’t available.
How it started
So Graphcore stepped in. Beginning the Graphcore ‘project’ in stealth mode back in 2013, Nigel and Simon built the best team capable of rethinking processors and have now created the first chip specifically designed for AI and the workloads required to make machine learning work.
What Nigel and the team at Graphcore call their Intelligence Processing Unit, or IPU for short, is completely different from any CPU or GPU you might find in even a high-end PC and will allow innovators to make the next big breakthroughs in machine intelligence.
Nigel Toon and Simon Knowles of Graphcore – a Bristol business which brings Brunel spirit to a modern problem.
For those of you who understand the technology involved, Nigel explains that machine learning “requires chips to excel in a number of areas, including low latency, and the ability to handle sparsity, meaning you might have large amounts of data points but not all of them need to be computed. Our IPU processor is designed to do all of those things really well – and as a result users are seeing great gains in performance and efficiency”.
This breakthrough in processing technology has drawn the attention of some serious players, too. Graphcore has gained investment from household names such as Dell, Bosch, and Samsung, as well as having seen their chips implemented in the Microsoft Azure cloud for innovators to use in their AI projects (Microsoft being another investor in the technology).
As awareness of the sector grew, so did their interest… In the meantime, we were able to build a team of around 40 of the best silicon and software engineers in the world
Graphcore’s technology wasn’t always this well-known, but that made finding the right people all the more important. “In the very early days, many investors were sceptical about the value of an AI hardware business”, Nigel says, “but as awareness of the sector grew, so did their interest… In the meantime, we were able to build a team of around 40 of the best silicon and software engineers in the world”.
“We hire exceptional people with different skills, backgrounds and experiences, many at the very top of their field, and try to find a balance of freedom and responsibility for our team”, Nigel says when asked about his business’s culture, which has helped them weather the current crisis well. “We think that exceptional people deserve the best we can provide. So Graphcore employees have the freedom to make choices that work for them in terms of flexible working arrangements, which has made the shift to remote working much easier”.
The current coronavirus crisis aside, Nigel still has big plans for 2020. “We are continuing to nurture our relationships with customers, partners and academia in order to build a wider ecosystem of AI innovators with IPU access. There will be many more partnership announcements from Graphcore this year, as global expansion and market growth will really be a key focus for us”.
In a new world of remote working where data and computing are more important than ever before, Nigel’s drive to help innovators who are held back by hardware will surely place Graphcore’s technology at the heart of a revolution in how the world works.
Graphcore and Royds Withy King, together in the heart of Bristol city
Read about Graphcore in Ahead of the Curve magazine