> ## Content Index
> Fetch the complete content index at: https://nexi.fund/llms.txt
> Use this file to discover other available public pages before exploring further.

# The Navy Bought Data Plumbing, Not Robots
- URL: https://nexi.fund/applied-intuition-navy-data-engine/
- Published: 2026-10-30T14:00:26.000Z
- Updated: 2026-10-30T14:00:26.000Z
- Description: A $25 million production contract puts Applied Intuition's perception system on five Arleigh Burke-class destroyers. The real deliverable is an over-the-air retraining loop the Navy has wanted for a decade.
- Author: Nexi.fund Labs
- Tags: Defence & Robotics, #mode-6, #hook-paradox, #track-E, #layout-signal-grid

The US Navy's first fleet-scale AI perception system is not going onto a new ship.

It is going onto five Arleigh Burke-class destroyers, hulls wired and commissioned years before anyone thought a camera feed needed a software supply chain. Applied Intuition announced the production contract on 22 October. The first capability under it is an Intelligent Perception System that fuses shipboard cameras and infrared sensors into one picture for the bridge.

The contract is reported at $25 million. The ships are not named. The install date is not public. None of that is the interesting part.

The interesting part is what the Navy is actually buying: a retraining loop that happens to ship with a perception model attached.

## Two pieces of software, one of them invisible

IPS runs on Applied Intuition's Acuity perception platform. It ingests camera and infrared feeds, many of which were never designed to talk to each other, and performs real-time detection, tracking and classification of surface contacts. Operators get visual overlays, predicted trajectories and automated alerts when a vessel is on a dangerous course. The system also replaces handwriting in the bridge logbook with machine-readable contact records that feed pattern-of-life analysis and fleet intelligence sharing.

Underneath IPS sits the Data Engine. It records passively from navigation sensors, curates the useful clips, pushes them ashore by satellite when bandwidth and security conditions allow, retrains models on land, then sends updated software back to the ship over the air. One vessel's contact becomes every vessel's model.

Applied is supplying the hardware as well: cameras, infrared sensors, onboard compute. That matters for a fleet that cannot wait for a mid-life modernization to add a new sensor mast.

$25M initial contract value 5 destroyers in the first tranche 

#### Navy production award to Applied Intuition

Announced 22 October 2026 under the DPRM Robotic and Autonomous Systems and PAE Mission Systems initiative. The company's release does not state a figure; $25 million comes from Breaking Defense. · *Breaking Defense, 2026*

## What five hulls buy the rest of the fleet

Five is the smallest number in this story. The award is a production contract, which in Navy language means the service is buying a repeatable configuration rather than another demonstration. Applied's pitch rests on four reusable layers: onboard compute, sensor integration, data pipelines and over-the-air update mechanisms. Any future perception or detection model can ride the same rails.

That changes the unit economics of a software upgrade. A new detection model stops being an integration project with a shipyard period attached and becomes a release. The Navy has wanted that sentence to be true for a decade. Whether it is true depends on validation records, configuration control and the willingness of a bridge team to trust a screen.

The difference between the prototype and the product is visible in scope:

| Parameter       | DECK prototype                         | IPS production                               |
| --------------- | -------------------------------------- | -------------------------------------------- |
| **Delivered**   | March 2026                             | Announced 22 October 2026                    |
| **Integration** | ✗ Standalone, kept out of ship systems | ◐ Wired into ship systems, hardware included |
| **Output**      | Sensor data for model training         | ✔ Live fused picture plus training data      |
| **Scope**       | At least one test ship                 | Five Arleigh Burke-class destroyers          |

Program history, 2026\. Sources: Applied Intuition, Breaking Defense, IN Defence

## The integration bill nobody has priced

A destroyer is a hostile place for a camera. Salt film, spray, stack gas and vibration degrade optics faster than any lab test predicts, and every new sensor competes for power, cooling and deck space that were sized decades ago.

The contract covers new cameras, infrared sensors and onboard compute, which makes this a hardware programme wearing software clothes. Applied has not said how much of the $25 million buys iron. The Navy has not said which hulls get modified first, or how long a yard period each installation needs.

Bandwidth is the second constraint. The Data Engine can only send ashore what the satellite link and security conditions allow, so curation happens on the ship before transmission. That puts a premium on deciding what is worth keeping while the ship is underway, with no engineer in the loop.

💡

**The retrofit test**  
A perception capability that needs a new mast, a new cooling loop and a dedicated operator is a new system. One that rides existing sensors and existing watch bills is an upgrade. The five destroyers will show which one the Navy bought. 

## The prototype that had to stay unintegrated

The Data Edge Collection Kit, DECK, was delivered to the Navy in March 2026 under the service's Portfolio Acquisition Executive for Robotic and Autonomous Systems. It collected thousands of hours of shipboard sensor data for model training, tested on at least one vessel.

Applied's head of product for defense, Alex Campbell, was blunt about how it got there. To move quickly, DECK was deliberately not integrated with other ship systems. The Navy then asked for the next level: the same pipeline, wired into the ship, while also giving a watchstander something useful on day one.

> Today's surface warfare picture depends on brittle manual processes, and data is siloed across networks and paper logs.— Peter Ludwig, co-founder and chief technology officer, Applied Intuition

The retrofit pattern is familiar from the ground side. As we wrote in October, Overland AI became the first ground-autonomy company to prime a production contract for the US military by bolting autonomy onto a vehicle chassis troops already maintain. [The same question applies here](https://nexi.fund/overland-ai-ibex-agv-contract-2026/): whether a legacy platform can absorb a new software layer without turning every fault into a maintenance bill.

## The same week, a different robotics contract quadrupled

Two days before the Applied announcement, the Navy and the General Services Administration raised the ceiling on Gecko Robotics' inspection contract from $71 million to $300 million. That is a 4.2× expansion six months after the initial award, and it comes with a number the service can defend: up to $8 million in avoided repair cost per ship.

Gecko's robots scan hulls, decks and welds, including through non-skid deck coatings that used to require removal. The company claims inspection work runs up to 50 times faster than conventional methods. The Chief of Naval Operations has set an 80% fleet readiness target, which gives the program a scoreboard. Applied Intuition has no equivalent scoreboard yet.

$300M contract ceiling ↑ 4.2× vs initial $71M award 

#### Gecko Robotics Navy inspection IDIQ

Ceiling raised by the Navy and GSA on 20 October 2026, six months after the initial award. The figure is maximum potential value, not guaranteed spending. · *Gecko Robotics, 2026*

The gap between the two deals is evidence quality. One contract bought robots that produce a thickness map a repair team can act on. The other bought a perception layer whose accuracy, false-alarm rate and behaviour in fog, spray or glare have not been published.

That asymmetry is not a scandal. It is what early software procurement looks like. As we wrote in October, Gecko's robots were qualifying submarine welds for Anduril two years before that shipyard opens; inspection data is easy to audit because the artefact is a number. [Perception is harder to audit](https://nexi.fund/gecko-anduril-submarine-inspection/) because the artefact is a judgement.

## What the announcement leaves out

The five destroyers are unnamed. No installation schedule has been published. The contract vehicle, period of performance and option structure are undisclosed. Applied's release does not attach a dollar figure; the $25 million comes from trade press, not the company.

The operational record is thinner still. No classification accuracy figures. No false-alarm rates. No test report from the Director, Operational Test and Evaluation or the Navy's Operational Test and Evaluation Force. Until one exists, five bridge teams are the test programme.

There is a human factors problem hiding in a single fused screen. A low-confidence classification can look as settled as a contact confirmed by radar and a lookout. If the watch team offloads its cross-checking to the display, the failure mode arrives quietly. Neither the company nor the Navy has said whether IPS surfaces per-contact confidence.

⚠️

**The boundary that matters**  
IPS is described as decision support for watchstanders. Published sources do not connect it to Aegis, fire control, weapons or ship control. Any claim that the system identifies lawful targets or authorizes engagement goes beyond what the contract announcement supports. 

## Why the data engine is the actual product

Jason Brown, Applied's general manager, compared the Navy's problem to the driverless car problem: the same volume of messy sensor data, the same need to perceive an environment before planning a path. The company built data engines for commercial autonomy first, then pointed the same approach at naval operations.

The Navy has already said the quiet part out loud. At AFCEA West on 12 February 2026, Secretary of the Navy John Phelan argued that a data engine is the precondition for an AI-enabled force, and that DECK turned ships into learning systems rather than static platforms. A service that believes that sentence will fund the loop before it funds the model.

Applied Intuition was founded in 2017 and describes itself as valued at $15 billion, with defense as one of six verticals. Its commercial base is the reason the Navy can buy a perception capability as a configuration item instead of a development programme. It is also the reason to read the contract carefully: a company optimizing for reusable infrastructure will ship the plumbing first and the performance claims later.

The strategic logic is easy to state. A fleet that retrains faster than a threat evolves holds an advantage that compounds with every deployment. A fleet that retrains slowly owns a collection of one-off systems with expiry dates.

Two things decide which fleet the Navy gets. The first is whether five hulls produce a validation record honest enough to expand from. The second is whether a bridge team, in bad weather, learns to trust a fused picture without forgetting how to check it.

## What to watch next

📊

**Four signals to track**  
  
Named hulls and an installation schedule — the first public evidence that the retrofit has left the slide deck.  
  
Any accuracy or false-alarm figure, or the first DOT&E report covering IPS.  
  
The contract vehicle and period of performance, which reveal whether this is a pathfinder or a programme of record.  
  
A second tranche beyond five ships, which would confirm the reusable-layers thesis. 

The Navy did not buy a robot that sees. It bought the pipeline that lets every ship learn what one ship saw, and it bought it at a price that keeps the option open.

The $25 million is a down payment on that loop. The five destroyers are where the loop either produces evidence or produces anecdotes.

[ Navy awards Applied Intuition $25M to deploy AI sensor fusion at sea The trade report that established the contract figure and the two-job structure of IPS: fuse the picture now, collect the data for better models later. Breaking Defense ](https://breakingdefense.com/2026/10/navy-awards-applied-intuition-25m-to-deploy-ai-sensor-fusion-at-sea/?ref=nexi.fund) 

The only source in the set that puts a dollar value on the award.

[ US Navy selects Applied Intuition for shipboard AI perception A careful read of what the contract does and does not authorize, including the distinction between operator support and autonomous engagement. IN Defence ](https://indefencemag.com/us-navy-selects-applied-intuition-for-shipboard-ai-perception/?ref=nexi.fund) 

Best source on the limits of the claim, and on what a repeatable pipeline has to record.

[ Navy picks AI contact-alert system for five unnamed destroyers Documents the open questions: unnamed hulls, no install date, no accuracy figures, no operational test report, and the risk of a watch team over-trusting one screen. SOFREP ](https://sofrep.com/news/navy-picks-ai-contact-alert-system-for-five-unnamed-destroyers?ref=nexi.fund) 

The sceptical read that keeps the rest of the coverage honest.

[ Gecko Robotics contract more than quadruples to $300 million The same-week comparison: a Navy robotics contract with a measurable return, used here to show what evidence a perception deal still lacks. Gecko Robotics ](https://www.geckorobotics.com/news/navy-idiq-expansion?ref=nexi.fund) 

Company release; figures are the contractor's own, including claimed savings per ship.