The winning bidder has to deliver 2,500 systems within two years that can perform a set of basic capabilities. The contract could ultimately lead to a purchase of over 100,000 highly advanced head-mounted systems for offensive and defense use in operations.

“Augmented reality technology will provide troops with more and better information to make decisions,” a Microsoft spokesperson said. “This new work extends our longstanding, trusted relationship with the Department of Defense to this new area.”

This award comes as Internet giants like Amazon (AMZN), Google (GOOG), and Microsoft face questions, protests, and open letters demanding change from their employees about military and government law-enforcement contracts in the U.S. and elsewhere, and work that tailors to the needs of censorship.

Google employees resigned in May over a contract with the Pentagon for AI-related work. In June, a group of Amazon employees asked its CEO to cancel a contract for its Rekognition facial-recognition software for firms that work with U.S. immigration authorities, and to reconsider sales to police departments and governments. An open letter appeared in October that claimed to represent Microsoft employees urged the company to not bid on a military cloud-services contract worth $10 billion.

A number of companies entered the bidding, including Magic Leap (according to Bloomberg News) and several military contractors.

Microsoft didn’t immediately respond to a request for comment from Fortune.

The Integrated Visual Augmentation System has a series of stages through which the Army wants a company to proceed, starting with this first award to Microsoft that specifies a minimal set of capabilities could be achieved with commercially available technology today. However, the subsequent stages quickly stepping up to hardware that outstrips anything now available publicly or known to exist in government hands.

The initial request for proposals said the ideal weight of the gear would be no more than four pounds between head-mounted equipment—which must not require a helmet, but must also work with standard helmets—and any pack worn on the body. But the Army’s ideal is just over two pounds.

The system will need machine learning to identify objects and targets, and use artificial intelligence algorithms to suggest routes, detect threats (like tripwires), and alert a soldier to anti-electronic measures, like GPS jamming. It would also monitor health conditions and coordinate activities among large groups of soldiers.

The devices will both provide a form of artificial night vision that optimally would detect with 90% accuracy a human-sized object 150 meters away with illumination just by starlight, and provide augmented overlays that will ultimately have to be readable across all lighting conditions.