There are many companies and government agencies that would love to know months in advance how much food will be needed to feed large populations. But it’s difficult to forecast crop yields across the U.S. and the world, and preparing for coming shortages isn’t something governments do well. That’s where Descartes Labs comes in, applying machine learning to giant satellite imagery data sets to analyze and predict crop yields. It works with clients in the finance, insurance, agribusiness, and environmental impact industries and is already profitable. Its ability to do better crop yield forecasts than anyone else comes from ongoing analysis of the three petabytes of data in its corpus—the equivalent of 60 million four-drawer filing cabinets filled with text, or almost 40 years of HD video. It also has 10 terabytes of new data coming online every day—the equivalent of 10,240 yards of books on a shelf.