Microsoft is buying SwiftKey in a bid to beef up its artificial intelligence push, according to a Financial Times report filed late Tuesday. A source close to the deal confirmed the basics to Fortune, but said the reported $250 million price is a tad high. The deal is expected to be announced Wednesday.
SwiftKey’s technology sifts through millions of typed-in keyboard combinations and collects that data in hopes of correctly guessing what you’re entering even before you finish typing. It’s like the autofill feature used by many software programs, but much, much smarter.
Microsoft had no comment.
Update: Here’s the official confirmation. Terms were not disclosed.
SwiftKey’s predictive search technology relies on software installed on Android and Apple—but not Windows—smartphones that log users’ keyboard activity to detect patterns about what users search for or want to know.
For more on artificial intelligence, check out the following Fortune video:
There’s an arms race going on in artificial intelligence, or cognitive computing, the goal of which is to enable computers to learn from their experiences in order to deliver better results. Microsoft (msft), Google (goog), Apple (aapl), IBM (ibm), and Amazon (amzn) Web Services, are all scrambling for an advantage by buying innovative startups and building their own technology.
Google, as the Times pointed out, bought DeepMind, another London-based AI company for about $400 million in early 2014. In October, Apple bought Perceptio, a startup specializing in image recognition for smartphones.
AI is already enabling computers to not only recognize people’ faces from digital images, but ascertain their age, and even their moods based on facial expressions. It also automates human language translation.
These are big, complex problems with lots of real-world uses.
Get Data Sheet, Fortune’s daily newsletter about the business of technology.
Swiftkey has built a trove of user-input queries and basically “scrapes” all the phrases entered on the public Internet to construct its data models, as Gigaom reported two years ago. It also parses the phrasing and language of individual users to better predict what the current user is about to ask.
The continuous interplay between what the user is typing now and its back-end model means it has a good shot of figuring out what you want, even before you’re done trying to express it. The more data it aggregates, the better its predictive results become.
No one’s saying much, but a reasonable person might assume that SwiftKey’s treasure trove of models and data would find a welcome home in Harry Shum’s Microsoft Technology and Research Group.
Per his blog post confirming the deal, Shum wrote that the buy aligns with Microsoft’s goal of bringing software and services to all platforms.
This story was updated at 7:17 a.m ET. on February 3, with Microsoft’s confirmation of the deal and with Mr. Shum’s quote.