×

Google worked with Intel on a custom A.I. chip for its Pixel phones

  • Google says it worked with Intel on the Pixel Visual Core chip.
  • The chip is meant to boost artificial intelligence and image processing on the Google phones.
Brian Krzanich, CEO of Intel.
Justin Solomon | CNBC
Brian Krzanich, CEO of Intel.

A special-purpose chip for artificial intelligence and other tasks in Google's new Pixel smartphones draws on Intel technology.

In addition to a Qualcomm Snapdragon 835 chip, the Google Pixel 2 and Pixel 2 XL both feature a new custom Pixel Visual Core co-processor, which is meant to improve speed and battery life when shooting photos with Google's HDR+ technology, and better handle AI workloads in apps, Google has said. But the company didn't disclose details about its partners on the chip.

Then, last week, device repair website iFixit published a teardown of the Pixel 2 XL that showed what the Pixel Visual Core chip actually looks like. The serial number on the chip in a photograph begins with "SR3," like some Intel chips.

In an email to CNBC, a Google spokesperson confirmed that the company worked with Intel on the Pixel Visual Core and noted that no existing chip had exactly what Google wanted for the new Pixel devices.

The Pixel 2 and Pixel 2 XL may not sell well — the first generation Pixels, released in 2016, barely made a dent. Even so, the collaboration with Google on the Pixel Visual Core chip is a win for Intel. Last year Intel canceled some chips for smartphones and tablets, although certain commercially available phones and tablets still have Intel chips or modems inside.

It's also some evidence that Intel is making some progress in AI, where Nvidia's graphics processing units are popular among researchers. Last week the company said it's collaborating with Facebook to develop AI chips for data centers.

Google has a mixed relationship with Intel when it comes to AI. Waymo, Google's sister company working on self-driving cars, has relied on Intel chips for years, but Google has done a lot of work on its own to develop its tensor processing unit chips for accelerating AI workloads in its data centers.

Apple, for its part, designed its own "neural engine" chip for its upcoming top-of-the-line iPhone X.

Intel declined to comment.

(Clarification: A quote from Google was mistakenly included in a previous version of this story.)