In July 2020, a barnacle-covered, shipping-container-sized cube was pulled up from 117 feet under water off the coast of Scotland’s Orkney Islands. On its end was a logo you might not expect from a sea treasure: Microsoft. The tube was part of Project Natick, a project from the tech giant to test the feasibility of running data centers underwater, instead of on land. It had been underwater for nearly two years, and it was still functioning perfectly.
Data centers, which power our online world, get very hot, and keeping them cool involves a lot of energy—and a lot of water, which runs through the system to keep the temperatures low. By putting the data center in the water, Microsoft can make it more efficient. The fact that the tube was filled with nitrogen, instead of regular air, and that there were no humans to accidentally jostle or bump any components also seem to have made the data center much less prone to damage, which would save future repairs.
The team built a new robotic platform that can attract mosquitoes and other bugs; a small, cube-shaped device emits CO2 and light to lure insects inside, then automatically identifies species and captures selected insects so their genes can later be analyzed to find critical information about viruses and larger animals that might be infected. In a single night, one device can observe thousands of mosquitoes. A smaller number can be captured for genetic testing.
The devices use sensors to monitor how mosquitoes fly, and then uses artificial intelligence to identify the species based on the wing pattern. When a species known to carry dangerous disease is present, the system can predict where the mosquitoes will be spreading the next day so that the government can spray insecticide in precisely targeted areas. After that intervention, the system can also track how much the mosquito population drops. The data can also be used to make longer-term predictions about where hot spots will emerge, so that the government can proactively treat the environment.

Microsoft also worked on T-Detect, a project with Adaptive Biotechnologies to analyze the human immune response to COVID-19. Using Microsoft’s AI and machine learning, the study examined the blood of 4,500 patients and created an open data base of the results. One breakthrough from the data: it identified which of the immune system’s T cells were responding to COVID. The result was a COVID T-cell test that is 97% accurate, helping give a more accurate picture of how many people had infections missed by testing.