Graphcore, a Bristol-based machine learning startup, has raised $30 million (£23 million) as it prepares to ship its first artificial intelligence (AI)-focused computer chips later this year. Total investment in the company now stands at $62 million (£48 million)
Graphcore's chips, known as intelligence processing units (IPUs), are designed to help computer scientists and programmers create highly-intelligent computer systems that can learn for themselves when fed large amounts of data. Such systems will be vital to power technologies such as autonomous cars and AI-powered cancer detectors in the future.
The company claims that its IPUs will allow researchers to develop new forms of AI quicker and more efficiently than today's graphics processing units (GPUs) and central processing units (CPUs), which have been around for decades and struggle to process the large quantities of data used in AI development.
"We're developing a chip but also building that into a system that goes into servers and cloud infrastructure as well," CEO Nigel Toon told Business Insider on Thursday. "It is a system to accelerate machine learning."
The funding round was led London-based by venture capital firm Atomico, which was set up by Skype billionaire Niklas Zennstrom in 2006.
Siraj Khaliq, the Atomico partner who will join the Graphcore board of directors, said in a blog post that "software relies on hardware to deliver its potential. And when it comes to AI, current hardware is proving a poor fit for the task."
A number of high-profile angels also participated in the round including: Google DeepMind cofounder Demis Hassabis; Uber chief scientist Zoubin Ghahramani, who is also a professor at the University of Cambridge; and the cofounders of Elon Musk's AI research firm, OpenAI. The company was also backed by corporates like Samsung, Dell, and Bosch in a series A funding round last year.
"Having them on boards and supporting us is a very strong validation," said Toon. "They have that insight into not only what is happening in machine learning today, but they're the innovators on what is going to come next. It gives us the insight into where the technology is going and how our IPU technology can help."
Toon said the money will be used to help Graphcore scale up its team from 60 people today to 120 people by the end of next year, and to help it plan its next generation of products.
"Building systems capable of general artificial intelligence means developing algorithms that can learn from raw data and generalise this learning across a wide range of tasks," said Hassabis in a statement. "This requires a lot of processing power, and the innovative architecture underpinning Graphcore's processors holds a huge amount of promise."
Greg Brockman, cofounder and CTO of Open AI, added: "Training machine intelligence models in minutes rather than days or weeks will profoundly transform how developers work, how they experiment and the results they will see. Being able to experiment across a much broader front, at a much faster pace will create new breakthroughs and will allow us to combine many machine intelligence techniques to jumpstart progress."