Access and Feeds

AI Chips: Special-Purpose Chips Accelerate New Algorithms

By Dick Weisinger

Artificial Intelligence is being used everywhere, but so far the technology has been built on general purpose CPU and GPU computing chips. The next generation of AI will be based on special-purpose, much more efficient computer chips for AI.

Today’s GPU chips have advantages over standard CPU chips that include large number of simple core processors with dedicated VRAM memory. GPUs have worked well in handling progressive learning applications that require a large amount of parallel processing.

But just the name ‘GPU’ is an indicator that the chip originally wasn’t designed with AI in mind. The ‘G’ of ‘GPU’ refers to ‘Graphics’. The market is moving away frmo general-purpose GPUs as companies like FaceBook, Amazon, Google, Tesla and others are all attempting to create special-purpose highly-optimized AI chips.

But the profile of the AI chips being sold in the market is expected to change in the coming years. McKinsey forecasts that AI chipsets will begin to take 50 percent of the AI chip market share by the mid 2020s.

A report by PwC found that “AI creates an unprecedented opportunity for semiconductor vendors due to its applicability across virtually every industry vertical, the strong forecast for the sheer number of chips needed both in the cloud and at the edge, and the growing need for specialized computing requirements to accelerate new algorithms.”

MarketAndMarkets forecasts that the AI chipset market will grow to $57.8 billion by 2026. Overall, the industry is currently seeing 40 percent annual growth.

Digg This
Reddit This
Stumble Now!
Buzz This
Vote on DZone
Share on Facebook
Bookmark this on Delicious
Kick It on DotNetKicks.com
Shout it
Share on LinkedIn
Bookmark this on Technorati
Post on Twitter
Google Buzz (aka. Google Reader)

Leave a Reply

Your email address will not be published. Required fields are marked *

*

thirteen − two =