Microsoft Corp. is building a global network of antennas co-located with its cloud data centers that will enable satellite operators to download data from orbit more easily.

The antennas, or ground stations, will become available on-demand via a new service called Azure Orbital, which debuted today at Microsoft’s virtual Ignite event. The service puts it on a more even playing field with Amazon Web Services Inc., whose rival public cloud includes a similar service known as AWS Ground Station.

AWS Ground Station and Azure Orbital both focus on the same basic challenge: Satellite operators need ground-based antennas to communicate with their satellites, but building such antennas can be prohibitively expensive. The cloud providers’ solution is to give customers the ability to rent antennas on-demand just as they would rent a virtual machine or database instance.

Though it may seem like a somewhat narrow niche, demand is expected to balloon as a result of a steady decline in the cost of launching satellites. “Five years back, this cost used to be about $50,000 to send one kilogram to space,” Ashish Jain, a principal project manager with the Azure Networking unit, explained in an Ignite session today. “The cost has now come down 20 times to about $2,000 to $3,000 per kilogram.”

Jain said estimates call for more than 25,000 satellites to be launched in the next two to years years alone. “As satellite operators send more satellites to space, they need more ground stations to communicate to their satellites,” he said.

Azure Orbital will provide access to antennas operated by both Microsoft and the company’s partners. The Microsoft-operated ground stations will feature roughly 20-foot-tall antennas that will be deployed either near the company’s cloud data centers or directly on the premises. This proximity will allow for short data transfer times, which should provide the benefit of allowing customers to process satellite data with fairly low latency.

Processing satellite signals after they’ve been downloaded is a fairly complicated process. Notably, companies must use specialized modems to extract the information encoded into the raw signals, a task with which Microsoft promises to help as well. The company will offer access to “software modems” for decoding downloaded data and, for projects with more advanced requirements, will also give customers the option to bring their own custom modems.

“Once data is delivered to the end point in customers’ virtual network,  customers can then use other Azure services like AI or ML, or even Azure Cognitive Services for that matter, to process the downlink data,” Jain said.

The intense competition between the top cloud operators previously led to the commoditization of traditional infrastructure-as-a-service resources such as storage and compute. Now that both AWS  and Microsoft are offering ground station services, the same may happen with satellite infrastructure. Consequently, the data processing features that cloud providers offer on top of their satellite infrastructure will likely take on a central role in how they differentiate their platforms for space sector clients.

Microsoft regularly expands its data analytics and machine learning portfolios. Just this morning, alongside the Azure Orbital announcement, the company said it will exclusively license OpenAI’s powerful GPT-3 language processing model to build new solutions for users.

Microsoft sees customers applying Azure Orbital in a broad range of fields, including agriculture, meteorology, oceanography and geology. The service is currently in preview.

Image: Unsplash