Semiconductor Startup Taps Chip Industry, to Release AI Chip in 2018


Photo by: blickpixel via Pixabay


Semiconductor startup Groq reveals a plan to develop its first artificial intelligence chip next year, disclosing that the first set of chip products will run 400 trillion operations per second.

Focused on developing processor chips, the Palo Alto-based company will introduce its first machine learning product that is twice the capacity of Google’s tensor processing unit (TPU), an app-specific integrated circuit that was specially developed for machine learning.

Groq also shared that it is tapping the chip industry to create their very own custom server chips to be used for artificial intelligence. It also notes to compete against semiconductor company Nvidia, whose graphics chips are presently sold at its highest standard for its capacity to run with intense calculations needed for deep learning software.

The Groq team is composed of engineers who previously worked with Google and also behind the tech giant’s TPU. Eight out of 10 members who designed Google’s TPU are now working with the startup. Their project is funded with a $1.3 million budget from venture capitalist Chamath Palihapitiya.

The company also recently hired semiconductor firm Xilinx's VP of sales Rangasayee to serve as Groq’s chief operating officer.

While the company revealed in its website that they will share their first machine learning product in 2018, no statement was released yet if they will also begin the shipment of the product.