Abstract

The hardware industry has experienced one of the largest ever waves of corporate consolidations in the last few years. Now only a few large companies are left that can design complex chips in advanced technology nodes fabricated by one of only four major semiconductor manufacturers. Innovation is very difficult in this risk-averse environment. Hardware startups are almost non-existent, and the big companies appear to have run out of ideas beyond cost cutting through mergers.

Tragically, the hardware industry’s death spiral comes just at the point where applications ranging from IoT to the cloud are demanding innovative customized hardware to enable new system capabilities. The traditional semiconductor business model cannot support thousands of startups developing custom chips, and startups cannot afford the cost, delay, and risk required to build chips in the traditional way.

Moore’s Law’s demise is largely irrelevant to the problems in the hardware industry, as except for a few high-volume products, cost per transistor is insignificant compared to upfront design costs. Reducing design time and cost are the key to reigniting innovation.

In this talk, I’ll describe our experiences in transitioning our UC Berkeley work on agile hardware development and the free and open RISC -V instruction set architecture into the real world through our startup, SiFive. SiFive is attempting to build a new semiconductor business model based on an agile, open-source approach to mass customization of semiconductors, dramatically reducing the cost to get new ideas into silicon.