Google has made articles load faster on Google search, Google News, and other properties on mobile devices using its Accelerated Mobile Pages (AMP) technology. The next logical step is to speed up the delivery of ads, which often appear on the same pages as articles but load more slowly.

People collaborating on the open source AMP project are “in the process of setting up a live experiment,” Malte Ubl, tech lead for the project, wrote today in a blog post. There is early code for an implementation of AMP-powered ads from Google’s AdSense ad network, and the plan is to also add support for Google’s DoubleClick ad exchange.

Indeed, Paul Muret, vice president of display, video ads, and analytics at Google, is talking about AMP today at the Doubleclick Leadership Summit.

But the idea is to eventually support all ad networks.

VB TRansform 2020: The AI event for business leaders. San Francisco July 15 - 16

That doesn’t mean that one day soon regular ads won’t be shown on AMP pages. AMP “will continue to support non-AMP ad creatives to allow for a gradual transition of the wider ecosystem,” Ubl wrote.

Just as Google has been ramping up the AMP project, Facebook has its own program, Instant Articles, for quickly showing articles on mobile devices. But there isn’t a big open source project where Facebook can collaborate with lots of third parties on new code for the project. Ads are at the very core of Google and its parent company Alphabet, so this is an important day.

Google has a few principles in mind for launching AMP for ads. For one thing, CPU usage should only happen for screens that are on screen at any given time to prevent unnecessary battery drain. Also, Google is splitting apart the process of ad requests and ad rendering. “By separating the two, A4A achieves much faster ad rendering at no additional CPU and memory cost,” Ubl wrote.

Here’s a look at what’s possible with AMP for ads alongside an AMP article:

An example of AMP for ads as implemented in an AMP article.

Above: An example of AMP for ads as implemented in an AMP article.

Image Credit: Google