Carnegie Mellon University
November 30, 2017

Franchetti quoted on AI chip startups

Recently, a growing belief has emerged among some investors that AI could be a unique opportunity to create significant new semiconductor companies. As companies continue to invest heavily in hardware to run deep-learning systems, the limitations of existing chips, such as Nvidia’s graphic chips, are being exposed. Despite having been tweaked to adapt, they soak up a lot of energy when working in parallel. CMU has had to ask its researchers to throttle back their chip use due to the strain they placed on the university’s power system. Franz Franchetti says that CMU is looking for an alternative power source; companies like Graphcore, Mythic, Wave Computing, intend to answer precisely to that need, developing newer chips tailored for AI applications. Read more.