Why it may be the opening move in a production race, and what it means for your role
At a glance
- What: DIU's SWAP-USV Sprint 2 prize challenge, for unmanned boats armed to shoot down drones and helicopters more than 3 nautical miles away (DIU PROJ00721).
- Money: Up to $40M in Sprint 2 prizes, plus $200M the government has budgeted for follow-on prototyping and production across the three-sprint challenge (DIU PROJ00721).
- Deadline: November 30, 2026, 11:59:59 PM Eastern (8:59 PM Pacific), per DIU. Some aggregators list Nov 29; DIU's own page is the authority.
- The catch: Entries must be complete boat-plus-weapon systems. Few companies build both, so most will need a partner.
- Who should care: CEOs, engineering leads, researchers, and investors in defense, manufacturing, and AI. Find your square in the matrix at the end.
- Next step: SOTAS matchmaking session for hull builders, weapon integrators, and autonomy teams Date TBD; register with the form at the bottom of this page.
A cheap threat, $40M in prizes, $200M behind it, and a Nov 30 deadline
The threat: Drones that cost about as much as a car can now threaten warships. The Navy has often answered them with missiles that cost up to 150 times more (an SM-6 against a $35K Shahed-class drone; see the table below), and every one it fires is one it can't use against a cruise or ballistic missile.
The opportunity: DIU is putting up to $40M in prizes, with $200M budgeted for follow-on production, behind a cheaper answer: unmanned boats that go out ahead of the fleet and shoot drones down. Each entry must be one integrated system, with boat, launcher, and weapon together.
The implication: Most coverage calls this a story about expensive missiles and cheap drones. That may be the smallest version of it. The bigger shift might be this: when precision gets cheap, naval defense starts to look like a race to build and coordinate more machines than the other side. For CEOs, that means choosing partners now. For engineering leads, it means autonomy that survives jamming. For researchers, it means hard problems that are still open. For investors, it means the factories and software this race may need.
The rest of this piece explains why, in four parts.
1. Why the threat keeps growing: precision got cheap
For most of history, hitting a moving ship from far away took an expensive weapon. That's changing. A Shahed-class attack drone has a planning cost of about $35K (CSIS, 2025), roughly the price of a car.
Ukraine showed what that can mean at sea. With almost no navy, it used sea drones costing roughly $200K apiece (TIME), alongside cruise and anti-ship missiles, to help damage or destroy around two dozen Russian warships, up to a third of the Black Sea Fleet. UK Defence Intelligence assessed in July 2024 that 26 Russian Navy vessels had been damaged or destroyed in the Black Sea region between February 2022 and June 2024 (Defense Express). That count covers every weapon, not just sea drones; the flagship Moskva, for one, was sunk by land-based anti-ship missiles (Center for Maritime Strategy). By October 2023, UK Defence Intelligence assessed, Russia had moved many of its most valuable ships, including cruise-missile carriers and submarines, from Sevastopol east to bases such as Novorossiysk (Forces News).
Cheap precision means the attacker can afford to keep shooting. And the trend suggests drones will keep getting cheaper.
Why the threat will keep getting cheaper
How much damage a drone does depends on three things:
- Energy: how much the explosive packs per kilogram. This has barely moved in 80 years. The main military explosives date to World War II or earlier, and even CL-20, the most powerful one developed since, releases only about 1.4 times TNT's energy by a standard Lawrence Livermore calculation (LLNL). Other measures put it somewhat higher, but nowhere near an order of magnitude. Energy sets the floor.
- Design: how the warhead directs that energy. The same explosive does far more damage when it's shaped. A shaped charge focuses its blast into a jet that can punch through steel several times its own width. Ukraine's cheap FPV drones have often carried shaped-charge warheads taken from RPG-7 rounds, some made decades ago; one Ukrainian video showed a PG-7 warhead manufactured in 1968 (Forbes).
- Precision: where it hits. This appears to be improving fastest, because guidance, cameras, chips, and autonomy ride the consumer-electronics cost curve. Hit exactly the right spot, like a radar, a launcher, or the waterline, and you need far less explosive.
Better design and better precision mean a smaller warhead can do the same job, so the drone carrying it can be smaller and cheaper. The explosive barely changes, but everything that aims and focuses it keeps getting better and cheaper. The likely result is cheaper, smaller, and more numerous threats.
You don't have to sink a ship to take it out of the war
Cheap drones don't need to sink a warship. A destroyer's radar faces, antennas, and launchers are expensive and exposed. Hit enough of them, even with small warheads, and the ship has to leave for repairs. A ship stuck in a repair yard for months is out of the fight just as surely as one on the seafloor, at least in the short to medium term. Repair capacity is scarce too. GAO found that private yards finished only 36% of non-nuclear surface ship repair periods on time in fiscal 2022, and that destroyer repairs ran a combined nearly 10,000 days longer than planned from 2015 to 2022 (GAO-25-106286). That puts repair capacity into the production race as well.
That tilts the math against the defender. The attacker may need only a few hits; the defender has to stop nearly everything.
2. The magazine: why the destroyer's missiles matter more than their price
A destroyer carries a fixed number of vertical launch cells: 90 on early Arleigh Burkes and 96 on Flight IIA and later ships (USNI Proceedings). Reloading them at sea isn't routine yet. The Navy first reloaded a launch cell underway in October 2024, using a device called TRAM to lower an empty canister into the cruiser USS Chosin, and said it expected to field the capability in two to three years (U.S. Navy; follow-on tests in 2025, USNI News). For now, ships "deploy with the missiles they were loaded at port," tailored to "the most stressing threats" (CSIS, 2024).
So every high-end interceptor fired at a drone is one fewer for a cruise or ballistic missile. When the cells run empty, the ship leaves the fight to reload. The real cost of a shot is its sticker price plus what it costs you to no longer have that missile.
| Weapon (Navy unit cost, FY2026 budget books) | vs. $35K Shahed-class drone |
|---|---|
| APKWS guided rocket (~$30.2K) | ~0.86× |
| JAGM (~$249.5K) | ~7× |
| SM-2 Block IIIC (~$1.76M) | ~50× |
| SM-6 (~$5.3M) | ~151× |
Assumes one shot per kill, the best case for the defender. APKWS is guidance kit plus motor plus warhead; SM-2 Block IIIC is the FY2025 unit cost, the latest year the Navy bought complete rounds. Costs are from the Navy's FY2026 budget justification books (Weapons Procurement, Navy; Ammunition, Navy and Marine Corps). We dropped the viral "$2,000 drone" comparison because CSIS calls that figure misleading.
Cost per effect, corrected
DIU asks bidders for unit cost, cost per operating hour, and cost per effect. A fuller version of cost per effect:
cost per effect ≈ (weapon cost ÷ Pk) + (boat cost per hour × hours on station ÷ kills) + rearm and logistics + the value of the destroyer's interceptors and safety you preserved
Pk is the probability that a shot kills its target. Without that last term, the math can make firing an SM-2 from a ship already on station look "cheaper." With it, the case for the boat gets much stronger.
Lasers change the magazine math
High-power lasers are on DIU's list of acceptable weapons, and they change the equation. A laser shot costs mostly electricity, so the limit on how many shots you can take shifts from how many missiles you carry to how much power and cooling you have. That goes straight at the cost problem.
Lasers have their own limits, though. A laser has to hold its beam on a target for seconds, it engages one target at a time, and fog, rain, and sea spray weaken it. A big enough swarm can still overwhelm one by sheer numbers. And a small boat may not have the power to run a useful laser at all. The Navy's leading shipboard example, the 60-plus-kilowatt HELIOS on the destroyer USS Preble, shot down four drones in an October 2025 test (Defense Daily). It is still a system under test on one large warship, not a weapon spread across the fleet.
So lasers are unlikely to end the production race. They mostly change what you're racing to build: power, cooling, and beam control instead of missiles.
Closer is cheaper: the kill is an AI problem
The same physics that makes attack drones cheaper works for the defender too. How an interceptor kills a drone matters:
- Blast: the shock wave weakens quickly with distance. Under standard blast-scaling rules (Hopkinson–Cranz "cube-root" scaling, used in the Pentagon's blast design manual, UFC 3-340-02), the explosive needed for the same effect grows with the cube of the distance. Get 10 times closer and you need about 1,000 times less explosive.
- Fragments: a warhead that throws metal shards spreads them over a widening sphere, so their density drops with the square of the distance. Ten times closer means about 100 times more hits per square meter.
- Direct hit: strike the drone itself and you barely need a warhead at all.
Exploding 10 meters away, exploding 1 meter away, and hitting the drone outright are three different weapons. The closer you get, the smaller and cheaper the interceptor can be, and the more of them a boat can carry. It's the magazine problem again, solved by precision instead of volume.
But getting close is hard. It takes speed, maneuverability, and very good AI: sensors and guidance that can track a small, maneuvering target and correct in the last fraction of a second, often while being jammed. Ukraine already uses cheap interceptor drones to chase down Russian Shahed-type drones; maker Wild Hornets says its Sting interceptors downed more than 1,000 Shaheds and similar drones in their first four months of serial production (Oboronka, company-reported). That makes the kill itself largely an AI problem. Whoever's guidance gets closest can use the smallest warhead, and that can compound into lower cost, deeper magazines, and an edge in the production race.
3. Why China could change the problem: a budget problem becomes a production problem
Against Iran-backed groups in the Red Sea, the cost ratio is a budget problem: painful, but survivable. Against China, it could become a production problem.
- A U.S. Office of Naval Intelligence slide, which the Navy confirmed as authentic, put China's shipbuilding capacity at about 23.2 million tons versus under 100,000 for the U.S., about 232 times greater (Business Insider; Fox News). The Navy cautioned that the slide shows context and trends, not a deep dive into China's mostly commercial shipbuilding industry.
- DJI alone has roughly 70% of the commercial and consumer drone market, and Morgan Stanley notes that China "reportedly can produce millions of drones per year" (Investing.com). China also dominates the parts inside them: it made 94% of the world's sintered permanent magnets (the kind in drone motors) and refined 91% of magnet rare earths in 2024 (IEA), and it holds nearly 85% of global battery cell production capacity (IEA).
In a long fight, the bigger question may not be whether a shot costs $30K or $2M. It may be whether we can build shooters as fast as the other side builds targets. That makes this a manufacturing and supply-chain story as much as a military one. Cheap counter-drone boats matter most if they can be built at scale, with parts that don't depend on the adversary.
Similar logic may apply elsewhere. In the Baltic and North Sea, against Russia, many of the Black Sea lessons could carry over to coastal waters and chokepoints. NATO is already testing that idea: in June 2025 its Task Force X-Baltic fielded more than 70 air, surface, and underwater drones for chokepoint monitoring and undersea-cable patrols, at about one-third the cost of comparable coverage with crewed frigates (NATO ACT). In the Pacific, distances are vast. A boat built to DIU's minimum of 200 nautical miles at 10 knots can't escort a carrier group across an ocean, so basing, refueling, and carrying these boats on larger ships become central open questions.
4. The nesting doll: coordination is the next battlefield
When everything is cheap and numerous, the hard problem may shift from building platforms to coordinating them.
Early signs are already visible. Ukraine's Magura sea drones now carry and launch FPV attack drones, and in late July 2026 used them to hit two Russian radar-related targets in Crimea (UNITED24 Media). Ukrainian sea drones have also launched FPVs guided by fiber-optic cable, which radio jamming can't cut, since September 2025 (UNITED24 Media). A newer Magura, the MV11, is a purpose-built mothership. Its maker UFORCE lists it at up to 12 meters long, with up to 2,200 kg of payload, up to seven days at sea, and an air-defense version carrying several dozen aerial interceptors (UFORCE, company-reported). DIU's own Sprint 1 asked for boats that launch drones (GovCon Wire).
Picture the stack: a fleet of ships, unmanned boats around it, drones launched from those boats, and maybe smaller drones launched from those. Ships enter and leave the area. Then add an enemy jamming communications and GPS (DIU's Little Creek test includes navigation under active signal jamming).
The nesting doll works both ways. An attacker can send one carrier drone that splits into many smaller ones near the target. Kill it before it splits and you've spent one shot. Wait, and you may have to stop every one of its children. Since even small hits can force a ship into repairs, letting a few through may be enough to lose. That's another argument for boats that fight far forward: they get a chance to kill the carrier before it releases its payload.
A human likely can't approve every move at that scale and speed. Autonomy starts to look like the weapon itself. The teams most likely to win are those whose systems can share targets, avoid shooting each other, hand off engagements, and keep working when the network goes down.
Push the fight forward: boats as pickets and charging stations
The earlier you stop a drone, the more you save, and with nesting dolls the savings multiply. Kill a carrier that would have split into ten drones, each of which might release more, and one shot could do the work of dozens. Every layer you intercept earlier multiplies the shots, money, and risk you avoid. And the farther out the fight happens, the fewer sailors are in danger.
That argues for pushing these boats as far forward as they can safely go, and it may change what they're for. Staying power may matter more than speed. A boat that can loiter for days at the edge of a threat area becomes a picket that watches and shoots first. It can also become a floating base for other drones, launching, recovering, and recharging them so short-lived aerial drones can stay in the fight. DIU's Sprint 1 already favored boats that could carry more drones and generate more sorties, and invited mixes of reusable, attritable, and expendable drones (Govly copy of the Sprint 1 solicitation). Ukraine's MV11 mothership is designed to stay at sea for up to a week, per UFORCE. The same idea could extend to ground robots on remote coasts and islands.
Seen this way, the nesting doll is also an energy network. Larger ships or shore bases refuel the boats, and the boats recharge the drones. Whoever builds that energy chain best could stay forward longest.
There are limits. The farther forward a boat sits, the more exposed it is, the harder it is to communicate with, and the more it has to decide on its own. Intercepting near an adversary's coast also raises hard questions about rules of engagement and escalation.
The so-what for LA
The biggest opportunity may not be building hulls. It may be what makes many cheap machines work as one system, and the ability to build them fast. What that means depends on your industry and your role, so find your row and column:
| Your role | Defense | Manufacturing | AI research |
|---|---|---|---|
| CEO | DIU only takes complete boat-plus-weapon systems, and few companies build both. Unless you're one of them, lock in partners now (proposals are due Nov 30). Pitch cost per effect, not unit price. | Production capacity, more than design, looks like the scarce resource. Show you can build at volume without Chinese parts; DIU says it may favor affordable, high-rate production. | Coordination under jamming is still a largely open problem. Position your company as the autonomy layer every boat-and-weapon team needs. |
| VP of Engineering | Design to DIU's tests: GPS jamming, fielding within 120 days, and open, modular interfaces (MOSA) that plug into Project AVO. Integration, more than component specs, is likely where teams get stuck. | Design for manufacture, fast rearming, and recharging drones at sea from day one. A boat that takes days to reload will struggle on cost per effect. | Build for broken networks: machines that share targets, hand off engagements, and avoid hitting each other when the link drops. Prove it in simulation and hardware-in-the-loop testing. |
| Researcher | We found no published hit rates for cheap weapons against drones from a moving boat. That data would be valuable to DIU and every bidder. | Cheap, fast production of motors, batteries, magnets, and munitions without adversary supply chains is a materials and process problem, not just a robotics one. | Terminal guidance that gets within a meter of a maneuvering drone, multi-agent coordination over unreliable communications, and how humans keep real control at machine speed. |
| Investor | Back teams with a path to production contracts, not demos. Watch who wins Sprint 2 and the follow-on awards. | The bottleneck may be the factory. Domestic drone components and fast-build boatyards could be the picks and shovels of this race. | Coordination software rides on every platform, so it may have the most leverage. But it needs defense-grade proof: jamming tests and Navy integration. |
Two terms from the table:
- MOSA (Modular Open Systems Approach) is the Pentagon's required design approach of building systems from modules with published, standard interfaces, so parts from different vendors can be added, swapped, or upgraded over the system's life (10 U.S.C. § 4401).
- Project AVO (Autonomous Vehicle Orchestrator) is the command-and-control layer DIU names as the primary orchestrator for SWAP-USV entries, handling boat maneuvering, target cueing, sensor sharing, and remote weapon release under human supervision (DIU PROJ00721). In January 2026, DIU, the Navy, and the Defense Autonomous Warfare Group opened a separate prize challenge of up to $100M for this kind of orchestrator: software that turns a commander's plain-language intent into coordinated action across many autonomous vehicles (DIU).
Local players include Epirus (Torrance), which has tested its Leonidas high-power microwave against boat motors in a Navy exercise coordinated by NSWC Port Hueneme (Epirus); high-power microwave isn't one of DIU's named weapon categories, so it would have to qualify as an alternate system. Saronic runs a West Coast supply hub in San Diego (Defense One). Anduril (Costa Mesa) is building autonomous surface vessels with HD Hyundai and Edison Chouest Offshore (Anduril), and Shield AI (San Diego) has run its Hivemind autonomy software on teams of unmanned boats (Shield AI). Whether any of them is bidding on SWAP-USV is not public. Port Hueneme and San Diego give teams open water to develop and practice on. That's a practical advantage over landlocked competitors, even though DIU's official trials are at Little Creek, Virginia, in early 2027 (DIU PROJ00721).
Call to action: DIU only accepts complete boat-plus-weapon systems, and few companies build both, so most bidders will need partners. Proposals are due November 30. To help, we're planning a SOTAS matchmaking session for hull builders, weapon integrators, and autonomy teams. The date is TBD; register with the form at the bottom of this page.
What we still don't know
- Whether a system must handle every target type (Group 1–3 drones plus helicopters) or can score well covering only some.
- The minimum number of shots per boat. DIU leaves "magazine depth" to bidders.
- How these boats deploy: based near shore, carried by larger ships, or traveling with the fleet.
- Hit probability and shots per kill for surface-launched APKWS against small drones.
- Cost per shot for lasers and guns.
Register for the SOTAS matchmaking session
The session is for hull builders, weapon integrators, and autonomy teams looking for partners before DIU's November 30 deadline. Date TBD. Register below and we'll send details as soon as the date is set.
Editor's note: Navy unit costs are from the FY2026 budget justification books, verified in the LARN research brief (§3a). Figures from Wild Hornets and UFORCE are company-reported.