Defense AI: which startup is ahead?

Which defense AI startups actually belong in this comparison?

Anduril, Shield AI, Helsing, Saronic, Govini, Vannevar Labs and Epirus are the seven startups we can compare seriously today.

We chose companies whose core product directly affects military decisions, autonomy, sensing, weapons or force operations, and whose commercial progress can be checked through contracts, deployments, revenue or production milestones. That keeps the field broad and avoids turning the article into another drone ranking. Anduril spans command software, aircraft, surveillance, counter-drone systems, underwater vehicles and weapons. Shield AI focuses on autonomous pilots and aircraft. Helsing is building a European portfolio across drones, electronic warfare, combat software and autonomous platforms. Saronic concentrates on autonomous vessels. Govini and Vannevar Labs sell operational software, while Epirus applies software-defined electronics to counter-drone defense.

We excluded Palantir because it is a large public company. Scale AI also sits outside the group after its ownership structure changed and because defense remains part of a much broader AI infrastructure business. Skydio, Quantum Systems, Castelion and several other companies are important adjacent players, though each would pull the comparison toward drones, missiles or another narrower category. Primer and Rebellion Defense remain relevant, but their public commercial numbers are too thin for a fair line-by-line comparison.

Private funding totals vary by database because some sources include debt, preferred financing or delayed facilities. The figures below are rounded and should be read as rough scale comparisons.

Startup	Main defense AI position	Approximate cumulative funding
Anduril	Full-stack autonomy, command software and defense hardware	More than $11B
Shield AI	Hivemind autonomy software, V-BAT and X-BAT aircraft	About $3.6B, including preferred and credit facilities
Helsing	European defense AI, strike drones, combat software and autonomous platforms	About $3.3B
Saronic	Autonomous surface vessels and maritime command software	About $2.6B
Epirus	Software-defined high-power microwave counter-drone systems	More than $550M
Govini	AI software for defense acquisition, logistics and supply chains	About $170M
Vannevar Labs	AI for intelligence, foreign-language data and operational planning	About $92M

Is Anduril already running away with defense AI?

Anduril is already running away with the overall defense AI startup race.

The gap shows up across several measures at once. Anduril reported $2.2 billion of revenue for 2025 after more than doubling in a year. Its latest round valued the company at $61 billion. Helsing now carries an $18 billion valuation, Shield AI $12.7 billion and Saronic $9.25 billion. Anduril is therefore worth about 3.4 times Helsing, 4.8 times Shield AI and 6.6 times Saronic.

Valuation alone would be weak evidence. Here it lines up with revenue, deployed products and entry into major production programs. Anduril has more than 400 Sentry towers deployed globally, recently won a $363 million order for more than 200 additional extended-range towers, and moved its FQ-44 collaborative combat aircraft into the Air Force production phase. None of the challengers currently combines that breadth with comparable commercial scale.

The wider defense market still belongs overwhelmingly to traditional contractors. The Wall Street Journal recently found that Pentagon spending on 15 leading defense technology startups had tripled since 2022 while remaining below 1% of total defense contracts. Anduril dominates the startup race, although it remains small beside the full defense industrial base.

If you want more recent data on this point, please see our latest defense tech market report.

Who has real revenue, and who is growing fastest?

Anduril has the biggest real revenue base. Shield AI is the fastest credible chaser.

Anduril’s $2.2 billion revenue base places it in a different commercial class. Shield AI expects at least $540 million of revenue this year, up more than 80% from roughly $300 million. Even after that jump, Shield AI would reach only about one-quarter of Anduril’s previous-year revenue. The percentage growth is impressive; the absolute gap is still huge.

Govini has built the best recurring software business in the group. It passed $100 million in annual recurring revenue. That is a smaller company than Anduril or Shield AI, yet the recurring model shows that defense buyers are using Ark as an ongoing workflow rather than paying for occasional experiments.

Helsing’s investor materials forecast €441 million of revenue this year. That would put it near Shield AI in scale, but we give the figure less weight because it is a forecast rather than reported revenue. Saronic has a $392 million Navy production contract and nearly $200 million was reportedly committed at award, but the company has not published a reliable annual revenue figure. Epirus and Vannevar Labs also disclose contracts more often than current revenue.

Anduril is adding the most dollars, Shield AI is expanding at the clearest high rate, Govini has the best recurring software economics, and Helsing or Saronic could still surprise once their large orders flow through reported accounts.

Who has defense AI products that are ready for real operations?

Anduril has the most battle-ready portfolio; Shield AI has the cleanest proof in autonomous aircraft.

Anduril already has several products in real operations. Its Sentry family has passed 400 deployed towers, and the new order for more than 200 units extends a seven-year relationship with U.S. Customs and Border Protection. Anduril also delivered on its first international program of record for the Royal Australian Navy. Surveillance and undersea autonomy now provide separate proof that the company can move products beyond trials.

Shield AI’s V-BAT has passed Coast Guard operational evaluation and is flying intelligence and surveillance missions with U.S. and allied users. Hivemind has also entered an Air Force production contract for collaborative combat aircraft mission autonomy. Shield AI already sells an aircraft and is now getting its autonomy software accepted on a separate high-end platform.

Saronic has crossed into Navy production with Corsair, which carries more weight than another maritime demonstration. Its larger Marauder vessel and the planned Port Alpha shipyard could expand that position, though the big-ship strategy has far less operating history than Corsair.

Helsing says thousands of HX-2 systems are deployed in Ukraine and announced a U.S. factory designed to produce more than 2,000 drones each month. That gives Helsing meaningful volume. Independent data on reliability, mission success and replacement rates remain limited, so we rank its maturity below Anduril and Shield AI for now.

Who has the strongest defense autonomy technology?

Shield AI currently has the best standalone autonomy stack, but Anduril owns the stronger end-to-end system.

The Air Force has given us an unusually useful comparison. Anduril’s FQ-44 flew with both Lattice mission autonomy and Shield AI’s Hivemind during the same flight campaign. The service then selected both companies for the next phase of collaborative combat aircraft autonomy. Each stack is credible at the highest level of the market.

Shield AI gets the edge when autonomy has to move across different vehicles. Hivemind is now piloting 30 classes of vehicles, including aircraft, helicopters, boats and ground systems. The completed Aechelon acquisition adds high-fidelity simulation used by the U.S. military, giving Shield AI a stronger training and validation environment before live flights. If Hivemind keeps moving across manufacturers, it could become a standard AI pilot layer rather than software tied to one vehicle.

Anduril’s advantage comes from owning more of the surrounding system. Lattice connects sensors, operators, autonomous vehicles and weapons, while Anduril also builds many of the products feeding data into that network. Customers can buy the software, sensors and vehicles from one company, then add more pieces to the same network.

Helsing is the main European autonomy contender, with products across strike drones, electronic warfare, underwater systems and the planned CA-1 Europa aircraft. Its portfolio is broadening quickly. The public evidence is still younger than the Air Force-validated work from Shield AI and Anduril.

If you want more recent data on this point, please see our latest defense tech market report.

Who can manufacture defense AI systems at scale?

Anduril is the only startup already manufacturing across several defense categories at meaningful scale.

Anduril says it more than doubled the number of developmental systems moving into scaled production during 2025. The company is already producing surveillance systems, Roadrunner interceptors, small drones and autonomous underwater vehicles. Recent reporting also identified a network of roughly ten factories and deliveries spread across several hardware categories.

That expansion has exposed real weaknesses. A WIRED investigation based on interviews with 37 current and former employees described safety lapses, production delays and uneven processes at several sites. We still rank Anduril first because it has delivered more hardware across more categories than its peers, although its factories now need to prove that speed can coexist with reliability.

Helsing is building a more focused mass-production model around HX-2. Its newly announced West Virginia facility is designed for more than 2,000 drones per month, and the company already reports thousands deployed in Ukraine. The monthly capacity remains a target until the plant is fully operating, but it gives Helsing a credible route into high-volume precision weapons.

Saronic has the largest industrial plan. Port Alpha in Brownsville is a proposed $3.2 billion shipyard on an initial 835-acre site, with operations planned for 2028 and capacity for vessels up to 850 feet. Saronic already owns smaller facilities in Texas and Louisiana. The new yard could become a deep competitive advantage, though its scale, workforce and schedule remain unproven.

Who has won the defense contracts that actually change a company?

Anduril has won the broadest set of company-changing defense contracts.

Anduril’s contract lead comes from several programs working together. The Army created an enterprise route for buying its commercial technologies and the Air Force moved the FQ-44 into production. U.S. Customs and Border Protection also ordered more than 200 additional towers for $363 million. The widely quoted Army figure of up to $20 billion describes a ten-year ceiling, so actual spending may finish far below it. The real prize is easier repeat purchasing across several products.

Saronic’s $392 million Navy award is unusually important for a company founded only a few years ago. Nearly $200 million was reportedly placed on contract at the start, which makes the award more concrete than a large empty ceiling. It also establishes Corsair as a production system rather than a prototype seeking a buyer.

Helsing has won a €268 million German order for HX-2 drones and a separate €223 million combat-cloud contract. Those two programs give it the largest disclosed European contract base among startups. Shield AI’s $198 million Coast Guard vehicle supports V-BAT services, and its Air Force mission-autonomy production award could become more strategically valuable if Hivemind spreads across several aircraft.

Govini has access to large government purchasing vehicles, including a $919 million supply-chain risk contract, alongside a $50 million award and a five-year Army arrangement for Ark. Epirus received $43.5 million for two second-generation Leonidas systems and related support. These awards are smaller, but they show that both companies have moved into repeatable government buying channels.

Which defense AI startup gives customers the best value?

Govini gives customers the clearest financial value today.

Govini’s value comes from recurring use across acquisition, logistics, production and supply-chain decisions. Those activities return every budget cycle, giving Ark a natural renewal path. The company avoids the enormous factories and inventory required by aircraft or shipbuilders, which lets more of each invested dollar support software and deployment.

Anduril proves its value through repeat use. More than 400 Sentry towers are already deployed across several variants, and customers have moved from tests into network expansion. The towers automate wide-area surveillance and reduce the number of people needed to watch remote terrain, although Anduril does not publish a transparent cost-per-area or payback calculation.

Shield AI uses a contractor-owned, contractor-operated model for the Coast Guard. The customer buys surveillance as a service while Shield AI handles the aircraft and operations. That can shorten adoption because the buyer does not need to build a complete pilot, maintenance and support organization before using V-BAT.

Epirus may eventually deliver the biggest cost advantage in counter-drone defense. Leonidas is designed to affect several electronic targets with one high-power microwave system, avoiding a separate missile for every cheap drone. The purchasing logic is strong, although public cost-per-engagement and battlefield reliability data remain too thin for us to rank Epirus first on value.

If you want more recent data on this point, please see our latest defense tech market report.

Who is using defense AI funding most efficiently?

Govini is far more capital-efficient than the hardware-heavy leaders.

The comparison is rough by design. Annual recurring revenue, annual revenue, contract value and forecasts are different measures. Software companies also need far less capital than aircraft and ship manufacturers. Even with those limits, the table shows a clear split between businesses already producing substantial real-world output and companies still financing future capacity.

Startup	Best available commercial evidence	Approximate funding	What the comparison tells us
Govini	More than $100M ARR	~$170M	Best disclosed capital efficiency
Vannevar Labs	$25M revenue and profitability reported in 2023	~$92M	Efficient early software business, though the figure is old
Anduril	Revenue equal to roughly 0.2x cumulative funding	More than $11B	Strong output for a highly capital-intensive manufacturer
Shield AI	At least $540M forecast revenue	~$3.6B	Fast growth, with valuation still pricing in major future scale
Helsing	€441M forecast revenue	~$3.3B	Commercial promise is substantial, but forecast-heavy
Saronic	Navy production award equal to about 15% of cumulative funding	~$2.6B	Capital is funding shipbuilding capacity ahead of disclosed revenue
Epirus	$43.5M recent Army contract	More than $550M	Product has entered procurement, while commercial scale stays modest

What can each defense AI startup do that competitors cannot easily copy?

Anduril’s full-stack position is the hardest to copy.

Anduril’s edge comes from the combination of Lattice, a growing hardware catalogue, government access, manufacturing sites and years of integration work. A competitor may match one aircraft or one software feature. Recreating the full network would require capital, approvals, factories, field data and working relationships across several military branches. The same breadth also creates execution risk because Anduril must manage many complex programs at once.

Shield AI owns the clearest vehicle-independent autonomy position. Every aircraft or vehicle integrated with Hivemind can expand its training data and certification experience. Aechelon’s simulation tools deepen that advantage. Shield AI does not need to manufacture every platform if militaries choose its software as the shared pilot across several suppliers.

Helsing benefits from Europe’s push for sovereign defense technology. Its German, British, Swedish and Ukrainian relationships give it political and industrial access that an American startup cannot quickly recreate. The new U.S. factory also shows that Helsing wants to compete outside Europe rather than remain a protected regional player.

Saronic has built a clear startup lead in autonomous surface vessels. If Port Alpha opens on schedule, the company would combine autonomy software, small-vessel production and a huge modern shipyard. Govini’s moat sits in defense acquisition data and daily workflows. Epirus owns specialized high-power microwave engineering. Vannevar Labs has developed tools around foreign-language and hard-to-reach intelligence data. Each company has a real niche, though none currently matches Anduril across the full category.

If you want more recent data on this point, please see our latest defense tech market report.

Which defense AI startup has the strongest momentum right now?

Anduril has the strongest momentum now, although Shield AI is closing the gap fastest.

Anduril has stacked several major advances within a short period: a $5 billion funding round, revenue that more than doubled, delivery of its first international program of record and a production agreement for the surface-launched Barracuda-500M. Each development supports the next one. Revenue funds development, production wins justify factories, and a larger installed base makes Lattice more useful.

Shield AI has recently moved Hivemind into Air Force production, completed the Aechelon acquisition, doubled orders and revenue in its latest fiscal year, and pushed X-BAT through engine and thrust-vectoring milestones. V-BAT also received a European operational authorization after a Frontex maritime pilot. The company is converting a strong autonomy story into several independent routes to market.

Helsing has raised $1.8 billion at an $18 billion valuation, selected West Virginia for its first U.S. factory and set a target of more than 2,000 HX-2 drones per month. Those moves give it the fastest geographic expansion. They also raise the burden of proof: Financial Times reporting put Helsing’s valuation at roughly 32 times forecast revenue, well above the multiples reported for Anduril and Shield AI.

Saronic raised $1.75 billion, secured a $9.25 billion valuation and chose Texas for the $3.2 billion Port Alpha yard. Its momentum now comes from industrial expansion more than fundraising. The key test is whether construction and hiring stay on schedule while Corsair production expands.

How much of the defense AI evidence can we really trust?

The ranking is reliable, but many headline defense AI numbers are much softer than they look.

We give the most weight to government production selections, obligated funds, delivered units and reported revenue. We trust company forecasts, factory plans and maximum contract ceilings less. A ten-year contract worth “up to” $20 billion can produce a small fraction of that amount. A $198 million IDIQ gives Shield AI the right to receive orders within a ceiling; it does not book the whole amount immediately. Saronic’s $3.2 billion shipyard describes planned investment and capacity that should begin operating in 2028.

The same caution applies to performance. Anduril and Shield AI can point to Air Force flight campaigns and production decisions, which are stronger than marketing demonstrations. Helsing reports thousands of HX-2 systems deployed, while independent mission-level performance data remain scarce. Epirus has shown impressive counter-swarm results in controlled events, yet public battlefield data and transparent engagement costs are still thin.

We also separate forecasts from completed years. Helsing’s €441 million and Shield AI’s $540 million are forward figures. Anduril has reported a completed revenue year, and Govini has disclosed recurring revenue. That difference is one reason we remain more confident in Anduril’s first place than in the exact ordering of Shield AI and Helsing over the next few years.

Which defense AI startups are actually ahead?

Anduril is the convincing overall leader today.

Anduril wins because it leads on the dimensions that count most at this stage of the market: reported revenue, product breadth, operational deployments, major production programs and manufacturing depth. Its lead is several times larger than the valuations alone suggest because the company has already connected software, hardware and repeat procurement.

Shield AI ranks second because Hivemind could become the shared autonomy layer across aircraft made by several companies. Helsing takes third through its European contract base, broad product ambition and rapid manufacturing expansion. Saronic ranks fourth because it owns the clearest maritime position, although its largest industrial advantage will remain prospective until Port Alpha starts producing ships.

The ranking could change. Shield AI would close the gap if Hivemind becomes standard across several production aircraft and revenue reaches the billion-dollar level. Helsing could move into second if its factories hit their targets and independent operating evidence catches up with its valuation. Saronic could rise quickly if Corsair expands and Port Alpha opens on schedule. Anduril would lose ground if safety, delays or factory complexity prevent its growing contract book from becoming reliable output.

Anduril’s $2.2 billion revenue is the clearest numerical separator. The harder advantage to copy is the complete operating system around it: Lattice, deployed sensors, autonomous platforms, production programs and repeat customers.

Rank	Startup	Why it ranks here
1	Anduril	Clear leader in reported revenue, product breadth, deployment and major production programs. Manufacturing execution is now its main risk.
2	Shield AI	Best independent autonomy stack, credible aircraft deployments and rapid revenue growth. Still far smaller than Anduril.
3	Helsing	Europe’s leading defense AI startup, with large contracts, thousands of reported systems and aggressive factory expansion. More of its case depends on forecasts and company-reported performance.
4	Saronic	Leading maritime autonomy startup, backed by a major Navy production contract and the Port Alpha plan. Large-vessel manufacturing remains early.
5	Govini	Best capital efficiency and a strong recurring software business. Its acquisition and logistics focus is narrower than full-spectrum operational autonomy.
6	Epirus	Clear specialist in high-power microwave counter-drone systems, with Army procurement and a difficult technology to copy. Deployment and revenue remain limited.
7	Vannevar Labs	Efficient, useful intelligence software with real military adoption. Current public revenue and contract data are too limited to place it higher.

If you want more recent data on this point, please see our latest defense tech market report.

Who is the author of this content?

NEW MARKET PITCH TEAM

We track new markets so founders and investors can move faster

We build living "market pitch" documents for emerging markets: AI, synthetic biology, new proteins, and more. Instead of outdated PDFs or hallucinated LLM answers, our clients get a clean, visual, always-updated view of what's really happening: key players, deals, regulations, and signals that matter. Learn more about us.