Read full article featuring portfolio company, TurbineOne.
---
American soldiers are starting to carry artificial intelligence in their pockets and rucksacks, the result of a $98.9 million contract between the U.S. Army and a San Francisco startup.
The contract with TurbineOne reflects twin realities of the modern battlefield: Drones and AI have accelerated the speed of combat to a blistering pace, and ubiquitous signal jamming makes it difficult to send and receive data at the front lines.
TurbineOne’s software runs on soldiers’ laptops, smartphones and drones, eliminating the need for a steady cloud connection. The AI application equips individual soldiers with the ability to quickly identify enemy threats, such as a drone-launch site or concealed troop position, and the context needed to decide how to respond without relying on analysts sitting miles away.
A four-year-old startup still in its infancy for defense work, run by a former Navy nuclear engineer turned tech executive and venture capitalist, TurbineOne isn’t the kind of company that has historically had much success winning business from the Pentagon. But a series of directives from Secretary of Defense Pete Hegseth has pushed the armed services to buy more commercial software and shed old and expensive weapons systems in favor of drones and AI. The Army in particular was tasked in April with transforming itself into a modern fighting force that uses fewer people and less-expensive but still lethal technology.
TurbineOne’s software is part of that transformation, according to Army senior executive Andrew Evans. It hints at the military’s appetite for technology from new, unproven startups that might help it prepare for future conflicts that scarcely resemble prior wars. The Army is rolling out TurbineOne to one unit at a time, starting with the infantry and cavalry, stress-testing the software and gathering more soldier feedback.
While embedded with the Army, TurbineOne made more than 200 software revisions in a week based on user feedback, said Chief Executive Ian Kalin. The TurbineOne deal marks one of the quickest paths—just shy of four years—for a software startup to secure a long-term contract with the Pentagon.
The modern battlefield is teeming with drones—in the sky, in the water and on the ground. That creates both unceasing surveillance and a deluge of data that soldiers must sort through, in seconds, to determine the most pressing threats and how to respond. Evans, the Army’s director of strategy and transformation, said the service’s goal is to process data 10 to 25 times faster than its adversaries, a benchmark it considers crucial to battlefield superiority.
“This is the most volatile, complex, uncertain and ambiguous environment that we have operated in,” Evans said. “Being able to hide, being able to obfuscate our signature, we are losing that advantage. Everyone is losing that advantage.”
TurbineOne’s AI examines large data sets of infrared imagery, radar, radio signals and other sources. Soldiers can ask it to detect general categories of threats, such as any aerial drone, as well as specific ones, such as a particular type of tank armed with particular weapons. A query returns information on the location of relevant threats and an assessment of the risk they pose, constantly pinging soldiers and their software-powered weapons, like drones, with updates as the target moves or changes.
TurbineOne compresses a task that might take a human 20 hours, such as sorting through images of hundreds of square miles of terrain, down to 20 seconds, Evans said. The software works with any language model; the Defense Department in July awarded contracts to model developers OpenAI, Anthropic, xAI and Google.
In opting for software that processes all data on-device, the Army is taking note of a lesson from Ukraine: Wars will be fought in a communications blackout, without radio links and GPS, which have been rendered null by the proliferation of jammers.
“What this does in war-fighting is it directly addresses the most strategic and significant threat that we face, which is being cut off from the rear and operating with only the kit that we carry,” Evans said.
This is a departure from the cloud-based AI systems the military has commonly used, but which aren’t only ineffective when jammed but potentially dangerous to the user, as any signal can be used to pinpoint location.
“In the Ukraine war it’s a bit of a dystopian reality, because if you turn on a cellphone, if you turn on a radio signal, you become a target,” said Kalin, who spent more than five years as a naval counterterrorism officer and nuclear engineer before launching a Silicon Valley career.
The Army is also exploring how to use AI to direct swarms of autonomous systems to attack targets in coordination—the latest front in drone warfare playing out in Ukraine. In these attacks, drones are turned loose with instructions for what to do when they encounter particular threats. TurbineOne software can deploy and coordinate drone swarms, a feature that has been used by other parts of the military separate from the Army contract.
The company has raised about $57 million from venture capitalists, including from the firm Kalin used to work for, and has a $300 million valuation. Its software has been used during U.S. military training exercises in Europe, the Pacific and on the northern U.S. border.