October 31, 2016
Foundation Capital
It’s been five years since IBM’s Watson claimed victory in the game show Jeopardy. Ten years ago, Artificial Intelligence and Machine Learning were mostly still in the realm of science fiction.
Today, we encounter this technology whenever we watch Netflix or talk to Siri. Even the ads we see on Google are powered by AI. And in the near future, AI will play an even more central role in our day-to-day lives, vaulting from prevalent to ubiquitous.
But there’s a problem in AI advancement that people aren’t talking about: computing power. AI still requires a massive amount of computing power. To train a powerful AI using neural network technology might take several hours, days, or even weeks, and use hundreds or even thousands of computer servers.
In response to computing demands, computer servers have quickly gotten more powerful—and shrunk in size. Now, it’s time for the same to happen with AI computers.
Which is why GraphCore just closed on $30 Million.
The market opportunity for AI software and systems is significant: the IDC estimates $70 billion in spending in this area by 2020. CTO Simon Knowles has created some exciting technology while “in stealth” for the past two years. And the team that Simon and CEO Nigel Toon have assembled to address this market is top of the line.
Right now, other than a few scientists and engineers, not many people are thinking about the implications and new applications if the available computer power for AI was 10 to 100 times faster than today’s computer servers. It would have applications in healthcare, financial services, and productivity, just to name a few. My prediction: this will all happen in a blink of the eye.
As a side note, GraphCore’s “seed” investment was done inside another company. While rare, this strategy can work when the right opportunity presents itself, but a traditionally venture-backed start-up wouldn’t be able to execute it for reasons of focus or capital requirements.
In this case, the “parent” company—XMOS—is focusing on another large market opportunity (Voice User Interface systems that “voice enable” homes, consumer devices, businesses, and autos). As a board, we decided it didn’t make sense to try to execute on both market opportunities within the same company, so we recently spun GraphCore out on its own.
As someone who firmly believes in the importance of focus, I think that decision will prove prescient.
So congratulations to the entire GraphCore team! You can read more about their announcement here.