Tomorrow.io unveils DeepSky: constellation of large satellites and instruments

SAN FRANCISCO – Weather intelligence startup Tomorrow.io unveiled DeepSky, a satellite constellation designed to refine atmospheric forecasts by gathering the vast quantities of data needed to feed artificial intelligence models.

DeepSky, a proliferated low-Earth orbit constellation, will be “a quantum leap in all dimensions,” Rei Goffer, Tomorrow.io co-founder and chief strategy officer, told SpaceNews.

Individual satellites in the DeepSky constellation, announced Jan. 20, will be significantly larger than the six-unit cubesats that populate Tomorrow.io’s initial constellation, Gen1. And DeepSky satellite will be equipped with “instruments of a completely different caliber,” Goffer said. “We are not yet sharing what these sensors are, but we are saying that our satellites will carry multiple very-high-impact, co-located sensors.”

News of DeepSky follows the Jan. 12 announcement that Tomorrow.io’s Gen1 constellation achieved a global 60-minute revisit rate for atmospheric observations with the Jan. 11 launch and speedy commissioning of its 10^th and 11^th microwave sounder satellites. Tomorrow.io’s Gen1 constellation attracted hundreds of government and enterprise customers including Ford, Uber, BNSF Railway, JetBlue and the U.S. Air Force.

“When we interact with users, from a railway operator to a civil aviation authority, a ministry of agriculture or a disaster management authority, we understand their different problems and the gaps that need to be filled,” Goffer said.

“Operational resilience now depends on treating atmospheric data with the same rigor as any other mission-critical infrastructure,” said Nikhil Ahuja, Amazon senior director, planning and supply chain. “The advancement in sensing and rapid refresh frequency DeepSky enables creates a new
class of AI-driven decision systems that are more adaptive and localized. This evolution will define the future of the world’s largest-scale operations.”

Data-Hungry Models

AI promises dramatic improvements in weather forecasting accuracy. But the models require extensive data.

DeepSky is designed to provide the extensive, high-resolution data Tomorrow.io needs to feed a new generation of proprietary weather models. In the past, weather models were often limited by available compute power and physical understanding of the atmosphere. Now, data is the bottleneck.

“AI systems depend on dense, high-frequency, and diverse observations —coverage that today’s satellite infrastructure cannot consistently provide,” according to the news release.

“Modern supply chains can no longer rely on static planning or historical averages,” said Matt Garland, BNSF chief technology officer, said in a statement. “True resilience comes from continuously sensing operating conditions and translating that intelligence into network-wide decisions. DeepSky represents a meaningful step toward defining a new category of what is possible as agentic AI becomes a critical part of planning
and creating a unified operational picture.”

DeepSky will complement existing satellite systems, including the geostationary and low-Earth-orbit constellations that feed public weather models, by offering “improved prediction of rapidly evolving and extreme weather events,” the news release said.

Without providing details, Tomorrow.io announced that DeepSky satellites will feature “new types of sensors, including capabilities that historically have been limited to one-off science missions due to cost and revisit constraints.”

DeepSky “is intended to address and significantly exceed the baseline observational requirements of major government and international customers, while remaining adaptable to evolving mission priorities and emerging applications,” the news release added.