Scale AI wins $500 million Pentagon contract, five times larger than its previous military deal

Scale AI Secures Massive Pentagon Expansion

Scale AI has landed a $500 million contract from the US Department of Defense, representing a five-fold increase from the company's previous $100 million agreement signed in September 20251

1

. The deal was inked with the Pentagon's Chief Digital and Artificial Intelligence Office, known as CDAO, which has emerged as the primary procurement vehicle driving military AI integration spending through 2025 and 20261

1

.

Dan Tadross, who leads Scale's public-sector business, told Bloomberg that the Pentagon was already "pushing the limits" of the original deal1

1

. The rapid expansion reflects the Defense Department's accelerating demand for AI tools that can support military decision-making and data-processing workflows. This marks one of the larger individual AI procurement awards the Pentagon has made to date.

Military AI Push Gains Momentum

The contract expansion aligns with Defense Secretary Pete Hegseth's broader strategy to accelerate AI adoption in military contexts. In a January strategy memo, Hegseth called for expanded use of AI tools and reducing bureaucratic hurdles that have historically slowed the deployment of new technologies2

2

. The Scale AI agreement utilizes a Production Other Transaction Authority (OTA), which enables any Pentagon component to channel funding through a centralized contracting authority without issuing separate competitive bids2

2

.

This flexibility has fueled rapid adoption since the contract's initial award. The vehicle now supports Project Agreements covering computer vision, generative AI decision support, and data operations—sidestepping the fragmented, years-long procurement cycles that have traditionally hampered innovation2

2

.

Data-Labelling Emerges as Critical Layer

Scale AI joins Microsoft, Amazon, and Google, which all signed major classified military-use AI agreements in May 20261

1

. However, Scale's role differs structurally from the hyperscaler contracts. While those deals focus on cloud-and-model infrastructure, Scale operates in the data-labelling and decision-support layer that sits on top1

1

.

The company has positioned itself around the data-quality bottleneck that has held back operational deployment of AI in military contexts. Models perform well in benchmarks but struggle when underlying training data is fragmented, mislabelled, or inconsistent with operational reality1

1

. Scale's defense business has been growing fastest precisely because it addresses this critical gap.

Meta Investment Fuels Public-Sector Growth

The Meta-backed designation reflects the social-media company's June 2024 investment that, alongside other strategic capital, valued Scale at over $14 billion1

1

. This capital base enables Scale to expand its public-sector business at the speed the new contract requires. Scale's chief executive, Alexandr Wang, has become one of the more visible faces of US AI policy advocacy, regularly testifying before Congress on military AI competitiveness1

1

.

The Pentagon has now approved eight firms' AI systems for use on classified networks, including major hyperscalers1

1

. The $500 million contract demonstrates that DoD AI procurement budgets are now structurally large enough to support multiple competing vendors at the half-billion-dollar single-contract level—a significant shift from the much smaller awards common as recently as 20241

1

.

Physical AI Applications on the Horizon

Scale's broader strategic positioning has shifted alongside its public-sector growth. The company has been moving into "physical AI" applications, where AI systems are deployed onto autonomous platforms, robotics, and uncrewed military hardware1

1

. The data-labelling foundation translates into the same value proposition for those platforms as it has for traditional military AI workflows. The new $500 million contract gives Scale roughly two to three years of runway to expand in both directions simultaneously, though questions remain about whether the data-labelling layer will remain a separate category as foundation models continue to improve1

1

.