The old way to win a Pentagon contract was to spend years in competitive bids, survive source selection, and then wait for the program office to figure out what it actually wanted. The new way, apparently, is to spend a year embedded with the 4th Infantry Division, prototype the full software stack in the field, and then get selected to lead one of the Army's top modernization priorities before the ink is dry on the evaluation.
That's what just happened with Anduril and the Army's Next Generation Command and Control program — and it's worth unpacking carefully, because it's not just a contract win. It's a demonstration of how the procurement relationship between the Pentagon and defense tech startups has fundamentally changed.
Prototyping as the New Sales Cycle
Breaking Defense reported on June 22 that the Army selected Anduril to lead the common data layer baseline for its NGC2 effort — the first major contract award as the program moves beyond prototype phase. The selection covers Anduril's Lattice platform combined with Palantir's Foundry, creating an edge-to-cloud data mesh with AI firm Raft providing data federation and transformation tools.
The number that matters most here isn't the contract value — no specific dollar amount was disclosed for this award. What matters is the vehicle it was awarded under: a 10-year, $20 billion ceiling enterprise licensing agreement that the Army entered into with Anduril, which consolidates multiple procurement agreements under a single umbrella. The Army can now pick and choose Anduril software and hardware across that vehicle without running a new competitive process each time.
That's a structurally different relationship than a traditional defense contract. It's closer to how a large enterprise buys software from a cloud vendor — a master agreement that enables rapid deployment decisions without the overhead of repeated source selections. For Anduril, it means the sales cycle for the next decade of Army software procurement is effectively already won.
But here's what made that possible: Anduril spent the last year prototyping the NGC2 full stack with the 4th Infantry Division at Fort Carson. The division just concluded its incremental sprint series in May and is heading to Project Convergence in July as the final test. That's not a traditional pilot program — it's a continuous delivery model where the software is being refined against real operational requirements in real time, with soldiers providing feedback that shapes the product.
Brig. Gen. Shane Taylor, the capability program executive for Command and Control Information Network, framed it explicitly as a speed-of-relevance problem: "This is a major step forward as NGC2 evolves into a phase of continuous delivery." That language — continuous delivery — is borrowed directly from commercial software development. It signals that the Army is trying to operate more like a tech company's enterprise customer than a traditional defense acquisition office.
What the Data Layer Actually Unlocks
The NGC2 common data layer is one of four layers in the full stack, alongside transport, integration, and application layers. It's the least glamorous of the four — nobody writes breathless coverage about data mesh architecture — but it's arguably the most strategically significant.
The data layer is what determines whether information generated at the tactical edge (a drone feed, a sensor reading, a unit position report) can be ingested, normalized, and made available to decision-makers at every level of the command hierarchy in real time. Without a coherent data layer, you have islands of capability that can't talk to each other. With one, you have the foundation for the kind of AI-enabled decision-making that the Army has been trying to build toward for years.
The combination of Anduril's Lattice and Palantir's Foundry is interesting precisely because it pairs two different philosophies. Lattice is built around real-time sensor fusion and autonomous systems management — it's designed to operate at the edge, in contested environments, with intermittent connectivity. Foundry is built around data integration and analytics at scale — it's designed to make large, complex datasets queryable and actionable. Together, they're trying to cover the full spectrum from the forward operating base to the enterprise data center.
The Raft Data Platform adds the federation layer — the ability to pull data from disparate sources without requiring everything to be in the same format or the same system. That matters enormously in an Army that has a "combination of legacy and modernized kit," as the Breaking Defense report notes. You can't rip and replace everything overnight. The data layer has to work with what exists while enabling what comes next.
This is the kind of architectural thinking that legacy prime contractors have historically struggled with. Lockheed, Raytheon, and their peers are extraordinarily good at building hardware systems to specification. They're less good at building software that evolves continuously, integrates heterogeneous data sources, and improves through operational feedback. That gap is exactly what Anduril and Palantir have been exploiting.
The Enterprise Agreement as Strategic Moat
I've written before about how the Pentagon's procurement reform efforts — OTA contracts, DIU awards, enterprise licensing agreements — are creating structural advantages for companies that can navigate them. The Anduril NGC2 award is the clearest illustration yet of how those advantages compound over time.
The 10-year, $20 billion ceiling enterprise licensing agreement isn't just a procurement vehicle. It's a commitment architecture. By consolidating multiple agreements under a single umbrella, the Army has effectively made Anduril a platform vendor rather than a point solution provider. That changes the competitive dynamics significantly.
Under a traditional contract structure, every new capability requirement triggers a new competition. Incumbents have to defend their position; challengers have a shot at displacing them. Under an enterprise licensing agreement, the incumbent has a structural advantage for every subsequent decision — not because competition is foreclosed, but because the transaction costs of switching are much higher. The Army has already done the security vetting, the integration work, the training. Switching to a different vendor means doing all of that again.
This is how enterprise software companies build durable businesses. Salesforce doesn't win every deal on merit alone — it wins because once an organization is running on Salesforce, the cost of switching is enormous. Anduril is trying to build the same kind of stickiness in defense software, and the NGC2 award suggests it's working.
The self-teaming model is also worth noting. The Army explicitly encouraged "industry self-teaming and collaboration" in the NGC2 award, and the Anduril-Palantir-Raft consortium is the result. This is a departure from the traditional prime-sub relationship, where a large prime contractor controls the team and parcels out work to smaller companies. Here, three companies with complementary capabilities are teaming as peers, with Anduril as lead. That structure is more agile — it allows each company to focus on what it does best — but it also requires a level of coordination and trust that's harder to maintain than a traditional hierarchy.
The Broader Signal: Speed Is Now a Capability
Zoom out from the NGC2 award and the pattern becomes clearer. The defense tech companies that are winning major contracts right now share a common characteristic: they're not just building better products, they're building faster. Faster to prototype, faster to iterate, faster to deploy.
The Financial Times reported that venture capital investment in defense tech has already surpassed 2025 totals this year — a reflection of investor confidence that this procurement transformation is durable, not cyclical. The money is flowing because the model is working.
Meanwhile, the contrast with legacy procurement is stark. Bloomberg reported that Lockheed Martin won a $35 billion deal to boost interceptor output — a massive contract, but one that's fundamentally about scaling existing production rather than introducing new capabilities. That's the legacy model: large, long-cycle programs that deliver known capabilities at scale. It's necessary, but it's not sufficient for the kind of adaptive, software-driven warfare the Army is trying to build toward.
The NGC2 award represents something different. It's a bet that the Army's most critical modernization priority — how units share and pass data across the battlefield — should be built by a company that's been iterating on the problem in the field for a year, not a company that won a competitive bid based on a proposal document.
That's a philosophical shift as much as a procurement one. It says that operational experience and continuous delivery matter more than traditional acquisition credentials. It says that the ability to show up at Fort Carson, embed with the 4th Infantry Division, and actually make the software work in the field is worth more than a polished source selection package.
For defense tech startups watching this, the lesson is direct: the path to major Pentagon contracts now runs through the field, not through the proposal office. Get your technology in front of operators. Iterate based on what they tell you. Build the relationship before the contract exists.
What Comes Next
The NGC2 program has a clear near-term milestone: Project Convergence in July, where the 4th Infantry Division will serve as the final test of the full stack. That exercise will be the first real stress test of the Anduril-Palantir-Raft architecture at scale, and its results will shape how quickly the Army moves to expand the program across the service.
The dual-architecture approach — running parallel systems to determine the right capability mix — means the Army isn't fully committed yet. There's still a competitive dynamic at play, and the July exercise will generate data that informs how the enterprise licensing agreement gets exercised going forward. Watch for post-Convergence reporting on which elements of the stack performed and which needed adjustment.
The deeper question is whether this model scales. The NGC2 award works partly because Anduril had the resources and the relationships to spend a year prototyping with a division. Not every defense tech startup can do that. The companies that can — the ones with enough capital to sustain extended field engagements, enough credibility to get access to operational units, and enough technical depth to actually deliver — are the ones that will build the next generation of defense platform businesses.
The Army just showed exactly what that looks like.
