Scout AI raises $100 million for military autonomy

Scout AI said it has raised $100 million to train an AI model for battlefield logistics, running autonomous four-seat all-terrain vehicles on a US military base in central California, according to TechCrunch. The 2024 startup’s training program has been operating for six weeks, and the company says its model—called Fury—will learn to operate and command military assets.

The round, led by Align Ventures and Draper Associates, follows a $15 million seed raise in January 2025 and comes alongside $11 million in US defence development contracts from DARPA, the Army Applications Laboratory and other Pentagon customers, TechCrunch reports. Scout AI is one of roughly 20 autonomy companies whose products the US Army’s 1st Cavalry Division is using during its training cycle at Fort Hood in Texas, with the unit planning to deploy with “proven autonomous products” in 2027.

The pitch is that warzones punish brittle automation. Scout’s chief technology officer Collin Otis, a former executive at autonomous trucking firm Kodiak, told TechCrunch he started the company after concluding that road-tested self-driving systems were not intelligent enough for “unpredictable war zones”. The firm is building what it describes as a Vision-Language-Action model, a technique first published by Google DeepMind in 2023 that uses large language model-style representations to connect perception and instructions to physical control.

In practice, Scout’s training environment looks less like a lab than a private proving ground bolted onto public infrastructure. The company’s exclusive tour for TechCrunch took place on an unnamed base with hilly terrain, loose sand and disappearing tracks—conditions that make it hard to pre-map routes or rely on lane-like structure. Scout’s team includes former soldiers overseeing tests, and the company frames the model’s development as a kind of apprenticeship, comparing it to how troops are trained from age 18.

The business logic is that software can move faster than procurement. If a division expects to deploy autonomy in 2027, the vendors that can demonstrate reliability in training cycles can become embedded long before the political debate about lethal autonomy is settled. Scout says Fury’s first role will be logistics support, but it also describes a pathway toward autonomous weapons, a shift that would move decision-making from operators holding controllers to models interpreting intent.

That shift also changes who carries the risk. When systems fail in a commercial context, the costs show up as recalls, lawsuits and lost customers. In defence, a large share of the bill is paid through contracts, and the failures are more likely to be classified than litigated. The result is a market where capability is advertised in fundraising decks while accountability is negotiated later in procurement language.

Scout’s vehicles are already driving on military property, funded by venture capital and federal contracts, and being evaluated in an Army training pipeline. The company’s next milestone is not a public launch date but a deployment calendar set by a combat division.