Signals Inbox·July 21, 2026·Defense Tech

Anduril vs Shield AI: who will win?

Anduril is winning the broader defense-autonomy race today, but Shield AI still has a credible path to control the independent AI-pilot layer through Hivemind.

Q1Why do people keep comparing Anduril and Shield AI?

People compare Anduril and Shield AI because both companies want to supply the autonomous aircraft and software that will fight alongside human forces. They also compete more directly now than they did a few years ago.

The cleanest example comes from the U.S. Air Force’s Collaborative Combat Aircraft program. Anduril builds the FQ-44 autonomous fighter selected for production. Shield AI’s Hivemind and Anduril’s Lattice are also competing to become the aircraft’s primary mission-autonomy software.

The Air Force made that overlap unusually visible. Both autonomy systems have flown on Anduril’s aircraft, and both companies received places in the paid software competition announced in June 2026. The service expects to select a primary provider in 2027.

Their ambitions overlap, but their businesses have taken different shapes. Anduril sells aircraft, missiles, submarines, counter-drone systems, sensors, command software and the factories needed to produce them. Shield AI concentrates much more of its value around Hivemind, V-BAT and the future X-BAT aircraft.

So the comparison goes beyond two drone manufacturers. Anduril wants to become a modern defense prime. Shield AI wants its AI pilot to become a standard across many manufacturers’ vehicles.

Q2Why is the Anduril vs Shield AI race difficult to call?

The Anduril versus Shield AI race is difficult to call because Anduril leads on company scale, while Shield AI may lead on the narrowest definition of reusable aircraft autonomy.

Revenue gives Anduril a large advantage. Aircraft production gives it another. Military procurement channels, factories and available capital all point in the same direction.

Shield AI looks more competitive when we isolate mission-autonomy software. Hivemind has flown on aircraft from several manufacturers and across very different vehicle classes. That record supports Shield AI’s claim that its AI pilot can travel more easily than a tightly integrated aircraft-software stack.

The available measurements also differ. Anduril usually reports contracts, aircraft milestones, factory capacity and operational deployments. Shield AI often emphasizes integrations, autonomous behaviors, platform classes and the speed of first flight.

We therefore need two scoreboards. The first measures who is building the stronger defense company. The second measures who could supply the best independent AI pilot. Anduril leads the first race by a wide margin. Shield AI holds a narrower and less proven advantage in the second.

Q3How much bigger is Anduril than Shield AI today?

Anduril is currently operating at about 7.3 times Shield AI’s latest disclosed annual revenue.

Anduril CEO Brian Schimpf said the company more than doubled revenue to $2.2 billion in 2025. Shield AI executives told Fortune that their 2025 revenue base was roughly $300 million and that they expected growth above 80% in 2026.

The businesses are not identical. Anduril earns money from a much wider portfolio, including air defense, command software, missiles, surveillance, autonomous aircraft, underwater systems and engineering work. Shield AI remains more concentrated around V-BAT, Hivemind and related services.

That breadth is part of Anduril’s lead. The company has already turned autonomy into several revenue streams instead of depending on one aircraft family or one software product.

Shield AI would exceed $540 million in revenue if it reaches its current projection. That would represent excellent growth, but it would still equal less than one-quarter of Anduril’s 2025 revenue. Anduril may also grow during the same period, so the gap is unlikely to close soon.

Commercial scale comparison

Q4Is Shield AI growing faster than Anduril now?

Shield AI may grow faster in percentage terms now, but Anduril has recently produced much more growth in actual dollars.

Shield AI’s forecast implies at least $240 million of additional annual revenue, moving from roughly $300 million to more than $540 million. The company also said in a May 2026 employee update that fiscal-year orders and revenue had doubled. It did not disclose the starting values or enough accounting detail to compare that statement cleanly with Anduril.

Anduril added at least $1.1 billion when revenue more than doubled to $2.2 billion in 2025. Its annual increase alone was about 4.6 times Shield AI’s entire projected increase for 2026.

Percentage growth still deserves attention. Smaller companies can overtake larger ones by compounding faster for several years. Shield AI also appears to be entering a stronger commercial period, helped by international V-BAT orders, the Air Force autonomy program and the Aechelon acquisition.

We just do not see evidence of convergence yet. Shield AI is expanding quickly from a smaller base, while Anduril has been doubling from a base already measured in billions.

Q5Which company has built stronger momentum lately?

Anduril has built the denser run of production-grade wins lately, while Shield AI’s recent momentum remains more concentrated around integrations and smaller aircraft orders.

Anduril’s recent sequence includes the FQ-44 production selection, the Army’s decision to make Anduril the lead for its NGC2 common data baseline, a first NATO contract to evaluate Lattice and an agreement to establish Barracuda missile production in Poland. Those announcements cover aircraft, command software, alliance infrastructure and munitions manufacturing.

Shield AI has also had an active run. It received a paid CCA mission-autonomy award, completed the Aechelon acquisition, added V-BAT orders from Poland and Greece and flew an autonomous electronic-warfare mission with L3Harris. Hivemind detected an electromagnetic threat and rerouted follow-on aircraft without human intervention during the latest test.

Several Anduril programs have crossed from prototype into production, procurement infrastructure or military-wide deployment. Many Shield AI projects are still proving that Hivemind can integrate quickly and perform new missions.

Shield AI is generating good technical news. Anduril is producing more evidence that customers are reorganizing budgets and operations around its products.

Q6Is Shield AI ahead of Anduril in the AI pilot race?

Shield AI is narrowly ahead of Anduril in the independent AI pilot race, although the decisive Air Force competition remains open.

The Air Force placed six autonomy vendors inside its CCA software pool. Only Anduril, Shield AI and RTX Collins Aerospace received the first paid production options. Two six-month phases will lead toward the selection of a primary provider in 2027.

Shield AI has the better portability story. The company says Hivemind now pilots 30 classes of vehicles, from small drones to collaborative combat aircraft. It has integrated the software with platforms linked to Anduril, L3Harris, Northrop Grumman, Mitsubishi Heavy Industries and Destinus.

Anduril holds a stronger position around the software. It owns the FQ-44 airframe, builds Lattice into its wider command architecture and can connect the aircraft with sensors, weapons, operators and other Anduril products.

The Air Force has deliberately prevented either company from locking the other out. Its government-owned architecture separates aircraft from autonomy software and allows the service to replace providers as performance changes.

That structure currently favors Shield AI’s pitch. Hivemind can win without Shield AI manufacturing the aircraft. Anduril still has more ways to earn money from the same program.

Current position in the CCA race

Q7Does Shield AI actually have better autonomy technology than Anduril?

On pure mission-autonomy evidence, Shield AI has the stronger public record right now. We cannot yet prove that Hivemind performs better than Lattice in a common combat test.

Shield AI’s latest demonstrations show more than basic flight control. In the L3Harris electronic-warfare test, Hivemind used live threat data to reroute unmanned aircraft through a safe corridor. In a separate Destinus program, the teams integrated Hivemind with the Hornet aircraft in under two months, then tested coordinated flights, target updates, terrain following and route replanning.

Those examples support the same technical claim from different angles. Hivemind appears relatively easy to move, can react to changing mission data and can coordinate different aircraft while keeping the final engagement decision with a human operator.

Anduril’s strongest evidence comes from a broader system layer. Lattice has linked sensors, command posts, artillery, drones and data infrastructure during Army exercises. The Army later selected Anduril to lead the NGC2 common data baseline, suggesting that the software performed well enough to move beyond a limited prototype.

The evidence still falls short of a clean comparison. Neither company publishes common measures for mission success, operator workload, reliability under jamming, decision quality or performance against the same adversary.

We give Shield AI the technical edge with medium confidence. Hivemind has shown more visible portability and autonomous mission behavior. Anduril has shown more operational integration at military scale.

Q8Which company is solving the hardest defense-autonomy bottleneck?

Anduril is better positioned against the bottleneck military customers care about today: producing dependable autonomous systems in useful quantities.

Autonomy demonstrations have become common. The harder work comes afterward. Militaries need aircraft that can be built repeatedly, repaired in the field, connected to existing weapons and replaced quickly during a prolonged conflict.

Anduril has organized much of the company around that problem. Arsenal-1 in Ohio is planned as a five-million-square-foot manufacturing campus. Australia already has a dedicated Ghost Shark production facility. The company is expanding rocket-motor capacity and recently said it delivered more than one month of normal Pulsar production in a single week for two government customers.

The FQ-44 has also moved into production. That gives Anduril a live test of whether its manufacturing model can handle a sophisticated autonomous aircraft rather than smaller drones or sensors.

Shield AI has a 107,000-square-foot V-BAT facility in Texas and says it is preparing for higher demand. X-BAT could eventually give the company a much larger manufacturing business, but the aircraft remains earlier in development. Shield AI’s advantage becomes more valuable when another manufacturer carries most of the factory burden.

The market currently rewards companies that can deliver a complete capability. Anduril has taken on the harder industrial task and made more progress against it.

Q9Does Anduril or Shield AI have the better customers and distribution?

Anduril has deeper customer dependence today, while Shield AI has built the more neutral route into other manufacturers’ aircraft.

The U.S. Army offers the best example of Anduril’s customer depth. Before signing its new enterprise agreement, the government managed more than 120 separate procurement actions involving Anduril products. The contract now gives Army organizations pre-negotiated access to Lattice, hardware, compute infrastructure and support.

NGC2 goes further. The Army selected Anduril to lead the common data baseline after exercises involving the 4th and 25th Infantry Divisions. Once a military organization uses the same data layer across sensors, command posts and weapons, replacing that layer becomes difficult.

Shield AI’s customer base is narrower but increasingly international. V-BAT is used or ordered for maritime missions by the U.S. Coast Guard, Japan, Poland, Greece and the Netherlands. Ukraine supplies a demanding operational environment and a constant stream of field feedback.

Shield AI also works with manufacturers that may prefer buying autonomy rather than developing it internally. Recent partners include L3Harris, Destinus, Mitsubishi Heavy Industries and UAV Navigation-Grupo Oesía.

That partner model could scale well. It also creates less customer lock-in until integrations become paid production licenses. Anduril currently earns more from each customer relationship because it supplies a larger part of the operational system.

Q10Which company has more future demand already under contract?

Anduril has more visible contracted demand, although its largest headline number requires careful interpretation.

The Army’s ten-year Anduril agreement carries a maximum estimated value of $20 billion. The Army explicitly describes that amount as a ceiling rather than committed spending. Calling it a $20 billion backlog would be wrong.

The agreement still gives Anduril a major advantage. It consolidates more than 120 previous procurement actions, fixes terms in advance and shortens the process for buying additional products. The useful information lies in the purchasing mechanism, not the ceiling alone.

Other Anduril programs provide firmer visibility. Australia awarded an A$1.7 billion Ghost Shark program of record. The Marine Corps has a $642 million counter-drone contract ceiling. The Air Force’s FQ-44 agreement creates a path to additional production lots, although the initial quantity and value remain undisclosed.

Shield AI’s largest disclosed standalone award is the Coast Guard’s $198.1 million V-BAT services contract. Poland and Greece have placed additional V-BAT orders, but Shield AI has not published their values. Its Air Force autonomy award could become much larger if Hivemind wins the primary software role.

Anduril therefore enters the next few years with better revenue visibility. Shield AI’s biggest upside depends on converting current competitions and integrations into recurring software purchases.

Q11Which company has better real-world military evidence?

Shield AI currently has the better fresh battlefield evidence because V-BAT is helping Ukraine find valuable targets under real wartime conditions.

CBS News recently documented a V-BAT mission over the Black Sea that identified Russian personnel and equipment on an offshore platform. Ukrainian forces then used the intelligence to coordinate an attack. Ukrainian operators also said V-BAT had helped locate expensive air-defense systems and guide longer-range strikes.

That reporting is more useful than a company demonstration. The aircraft is operating against an adversary that jams communications, changes electronic-warfare frequencies and actively tries to destroy surveillance systems. Shield AI also keeps personnel in Ukraine, allowing software and operational procedures to change with battlefield conditions.

Anduril has extensive operational exposure across counter-drone defense, border surveillance, Army exercises and several combat theaters. Its Ghost aircraft accumulated more than 1,200 hours across 13 Army units during one reported year. Lattice has also been tested across increasingly large command-and-control exercises.

We still see less public evidence linking an Anduril system to a specific combat outcome with the detail available for V-BAT. Some Anduril programs may be classified, so the public record can understate actual use.

Based on what can be verified openly, Shield AI wins this question. Its Ukraine presence gives the company a feedback loop that laboratory testing cannot reproduce.

Q12Are Anduril and Shield AI becoming different kinds of companies?

Anduril is becoming a full defense prime, while Shield AI is building an AI-pilot company supported by selected aircraft and simulation products.

Anduril now operates across fighter aircraft, cruise missiles, underwater vehicles, counter-drone defense, tactical drones, electronic warfare, command software, rocket motors, sensors and soldier equipment. Lattice connects much of that portfolio.

Several acquisitions helped create this breadth. Area-I brought the Altius aircraft family. Dive Technologies added large underwater vehicles. Adranos added solid rocket motors. These capabilities now appear inside visible production plans rather than sitting as disconnected subsidiaries.

Shield AI has followed a more focused acquisition path. Heron Systems formed the foundation of its high-end autonomy work, while Martin UAV brought V-BAT. The completed Aechelon acquisition adds simulation, synthetic environments and sensor modeling used by military customers.

Aechelon could become especially important. AI pilots need huge volumes of difficult, dangerous and rare training scenarios. Shield AI can now connect simulated missions, live flights and software updates inside one development loop.

X-BAT complicates that strategy a little. Building its own autonomous fighter could create a major new business, but aircraft manufacturers may become less comfortable treating Hivemind as a neutral software layer when Shield AI also competes with them.

Anduril’s model has more moving parts and more execution risk. It also gives the company control over a much larger share of each defense program.

Q13Which company has the stronger moat today?

Anduril has the stronger overall moat today because its advantages reinforce one another across software, hardware, procurement and manufacturing.

A customer can begin with one Anduril product and gradually adopt Lattice, sensors, interceptors, aircraft, edge computing or support services. The Army enterprise agreement and the NGC2 data baseline make that expansion easier. New products then benefit from software and customer relationships Anduril has already established.

Factories add another layer. Defense manufacturing requires specialized suppliers, testing infrastructure, security processes, trained workers and government approvals. Competitors can copy a drone concept much faster than they can reproduce the production system around it.

Shield AI’s moat is more concentrated. Hivemind gains value from repeated integrations, flight experience, mission data, simulation and engineering tools that shorten the move from one aircraft to another. Aechelon should deepen that advantage by generating more training and validation data.

Open military architectures place limits on both companies. The Air Force wants autonomy providers to remain interchangeable, and the Army describes its systems as open rather than proprietary silos. Governments are actively designing procurement around reduced vendor lock-in.

Shield AI could still build the more valuable software standard. It first needs several manufacturers to buy Hivemind repeatedly at meaningful scale. Anduril’s moat already appears in contracts, deployments and production infrastructure.

Q14Which company has more money to keep competing?

Anduril has much more financial room than Shield AI, even after Shield AI’s large new financing.

Anduril raised $5 billion at a $61 billion valuation. Shield AI raised $1.5 billion in new equity and $500 million in fixed-return preferred financing at a $12.7 billion post-money valuation. Blackstone also committed an additional $250 million delayed-draw facility.

The headline difference is large. Anduril’s latest round brought in 2.5 times more capital, and its valuation is about 4.8 times higher.

Revenue makes the valuation comparison more revealing. Anduril’s latest valuation equals roughly 28 times its 2025 revenue. Shield AI’s equals about 42 times its estimated 2025 revenue, or around 24 times the revenue it hopes to generate in 2026.

Investors are asking Shield AI to grow into a more demanding valuation. The company must scale V-BAT, integrate Aechelon, fund Hivemind and develop X-BAT at the same time.

Anduril faces an even larger capital bill across factories, missiles, aircraft, submarines and software. Its stronger revenue base and larger cash injection give it more room to absorb delays or failed programs.

Q15Which company has the more dangerous execution problems?

Shield AI currently carries the sharper single-product risk, although Anduril’s rapid expansion has created serious operational risks of its own.

A Reuters investigation found that more than 50 of roughly 200 upgraded V-BATs in Shield AI’s internal fleet had been destroyed during testing or training over 18 months. Reuters also reported two separate propeller incidents that caused severe finger injuries to military personnel.

Shield AI disputes the broader picture created by those reports. The company says V-BAT has completed more than 13,000 missions, maintains a lifetime mission-readiness rate above 98% and operates across six continents. It also says many recorded mishaps involve repairable damage rather than destroyed aircraft.

Both sets of information can be true. A heavily used aircraft will suffer accidents, especially during development and training. Losing roughly one-quarter of an internal upgraded fleet, if Reuters’ figure is accurate, still represents a major reliability and cost problem.

Anduril’s risks are spread across a much wider portfolio. A recent WIRED investigation described safety lapses, production strain and management problems around the company’s rocket-motor expansion. Earlier reporting also documented failures during autonomous maritime and aircraft tests.

Concentration changes the stakes. Problems with one Anduril product can be absorbed by revenue and progress elsewhere. V-BAT remains central to Shield AI’s current revenue, customer relationships and credibility as an aircraft manufacturer.

Q16Who is winning between Anduril and Shield AI right now?

Anduril is clearly winning the overall company race right now. Shield AI leads one important technical contest, but that advantage has not yet produced comparable commercial power.

Three factors decide the current verdict.

First, Anduril operates at roughly seven times Shield AI’s latest annual revenue. Second, it has moved more programs into production or military-wide infrastructure. Third, it controls more of the system surrounding autonomy, including aircraft, weapons, command software, factories and procurement channels.

Shield AI’s case remains serious. Hivemind has the better public portability record, V-BAT provides unusually relevant battlefield feedback and the CCA software competition could validate Shield AI’s independent AI-pilot model. The Aechelon acquisition also gives the company a stronger simulation and training engine.

Shield AI can narrow the gap by winning the primary CCA autonomy role, turning several OEM integrations into recurring licenses, improving V-BAT reliability and getting X-BAT into credible flight testing.

Anduril must prove that its recent contract wins can survive large-scale production. Arsenal-1 needs to deliver reliable aircraft and weapons on schedule. Lattice must continue working as deployments expand from individual units to larger military networks. The company also needs enough management discipline to handle a portfolio that increasingly resembles a traditional prime contractor.

The long-term race remains open, but the current answer requires little hedging. Shield AI may become the leading independent AI pilot. Anduril is already closer to becoming the company that supplies the autonomous military around it.

Anduril vs Shield AI: current scorecard