The Race to Build Microchips to Power Artificial Intelligence
MIT Technology Review says AI may be evolving faster than chipmakers.
Venture capitalists have long been wary of silicon, semiconductor chips, even though they gave Silicon Valley its name. They are more expensive to develop than software and there has been little room to distinguish new versions, writes MIT Technology Review. Most new companies don’t survive, and the ones that do have wafer-thin profit margins. Intel and Nvidia are formidable competitors both regarding knowledge and funds.
But recently there has been a change. MIT Technology Review reports that venture capitalists have invested $113 million in AI-focused chip start-ups this year, which is almost three times as much as in all of 2015. There is now a belief among some investors that AI could be a place for semiconductor companies to grow. There is a strong belief that “tailor-made processors for AI applications can beat less specialized chips at a wide range of machine-learning tasks,” according to MIT Technology Review.
However, there are limitations to using semiconductor chips in AI. For one, when large numbers of graphic chips work in parallel, they soak up a lot of energy, reports MIT Technology Review. A leading AI research center, Carnegie Mellon University, asked researchers to throttle back their use of the chips temporarily because it was affecting the university’s power system.
Startups are also going to have to face big chip companies, who have already started making their own made-for-AI chips to compete with the startups. Plus, it can take years to get a chip to market, and AI might be evolving faster than that.
This article was featured in the InsideHook newsletter. Sign up now.
Suggested for you