Apple buys edge-based AI startup Xnor.ai for a reported $200M

Xnor.ai, spun off in 2017 from the nonprofit Allen Institute for AI (AI2), has been acquired by Apple for about $200 million. A source close to the company corroborated a report this morning from GeekWire to that effect.

Apple confirmed the reports with its standard statement for this sort of quiet acquisition: “Apple buys smaller technology companies from time to time and we generally do not discuss our purpose or plans.” (I’ve asked for clarification just in case.)

Xnor.ai began as a process for making machine learning algorithms highly efficient — so efficient that they could run on even the lowest tier of hardware out there, things like embedded electronics in security cameras that use only a modicum of power. Yet using Xnor’s algorithms they could accomplish tasks like object recognition, which in other circumstances might require a powerful processor or connection to the cloud.

CEO Ali Farhadi and his founding team put the company together at AI2 and spun it out just before the organization formally launched its incubator program. It raised $2.7M in early 2017 and $12M in 2018, both rounds led by Seattle’s Madrona Venture Group, and has steadily grown its local operations and areas of business.

The $200M acquisition price is only approximate, the source indicated, but even if the final number were less by half that would be a big return for Madrona and other investors.

The company will likely move to Apple’s Seattle offices; GeekWire, visiting the Xnor.ai offices (in inclement weather, no less), reported that a move was clearly underway. AI2 confirmed that Farhadi is no longer working there, but he will retain his faculty position at the University of Washington.

An acquisition by Apple makes perfect sense when one thinks of how that company has been directing its efforts towards edge computing. With a chip dedicated to executing machine learning workflows in a variety of situations, Apple clearly intends for its devices to operate independent of the cloud for such tasks as facial recognition, natural language processing, and augmented reality. It’s as much for performance as privacy purposes.

Its camera software especially makes extensive use of machine learning algorithms for both capturing and processing images, a compute-heavy task that could potentially be made much lighter with the inclusion of Xnor’s economizing techniques. The future of photography is code, after all — so the more of it you can execute, and the less time and power it takes to do so, the better.

It could also indicate new forays in the smart home, toward which with HomePod Apple has made some tentative steps. But Xnor’s technology is highly adaptable and as such rather difficult to predict as far as what it enables for such a vast company as Apple.