What are the main business models in defense tech?

Why is defense tech such a big business story right now?

Defense tech is a much bigger business story today because governments are spending more, buying more urgently and opening procurement to commercial technology companies that would have struggled to enter the market a decade ago.

SIPRI's latest global dataset puts military spending at $2.887 trillion in 2025, the eleventh consecutive annual increase and 41% above the level ten years earlier. Europe was particularly aggressive: spending rose 14% in real terms in a single year. NATO has since moved beyond budget promises. At its 2026 Ankara Summit, the alliance announced more than $50 billion of new procurement deals covering areas such as air defense, long-range strike, uncrewed systems, intelligence and advanced technologies.

Capital is following those orders. PitchBook data reported by Defense News showed venture investment in defense technology jumping from $27.2 billion in 2024 to a record $49.1 billion in 2025. Anduril alone raised another $5 billion at a $61 billion valuation after reporting roughly $2.2 billion of 2025 revenue.

Procurement rules are changing too. The Pentagon has recently pushed commercial solutions higher in the acquisition hierarchy, while its Software Acquisition Pathway is now the preferred route for software components. The U.S. Army's new ten-year Anduril enterprise contract is a good example of how far this has gone. More than 120 separate procurement actions for Anduril products were consolidated into one vehicle with a ceiling of $20 billion.

The opportunity has widened considerably, but a government can still spend $1 billion on software, missiles, engineering hours or satellite capacity and create four completely different businesses.

What does a defense-tech company actually sell?

A defense-tech company can make money in at least eight genuinely different ways, and the useful distinction is what the customer keeps paying for after the first contract is signed.

At one extreme, governments pay a prime contractor to develop and deliver an entire weapons system. Lockheed Martin might be responsible for an aircraft or missile program that includes engineering, production, integration, upgrades and support.

A product company works differently. AeroVironment can manufacture systems such as Switchblade and sell additional units as governments increase inventories. A missile manufacturer gets another form of repeat demand because weapons are fired, used for training or kept in stockpiles that eventually need replenishment.

Software companies can charge for continuing access to operational platforms. Palantir does this with defense and intelligence customers. Shield AI is trying to go one layer deeper by putting reusable autonomy software inside aircraft built by several manufacturers.

BlackSky shows another model: a government can subscribe to satellite imagery and guaranteed collection capacity without owning the satellites. The same customer can later buy a sovereign satellite system while continuing to pay for operations, imagery and analytics.

Then there are services. Governments pay companies for engineering, integration, cyber work, R&D and people with specialist or classified expertise. Finally, once equipment has been deployed, maintenance, spare parts, upgrades, training and logistics can generate revenue for decades.

These categories often sit inside the same company, which is why simply calling a defense company "hardware" or "software" usually tells us very little about its economics.

Defense-tech business model	What the customer actually pays for	Typical examples	What drives repeat revenue
Prime contracting	A complete military system or major program	Fighter aircraft, missile defense, submarines	Follow-on lots, upgrades and new program phases
Productized hardware	Standardized units	Drones, sensors, counter-UAS systems	More units and wider deployment
Munitions and attritable systems	Equipment that gets used or lost	Missiles, interceptors, loitering munitions	Replenishment and stockpile expansion
Software platforms	Continuing access to mission software	Command-and-control, data and AI platforms	Renewals and expansion inside the customer
Embedded autonomy	Reusable software inside third-party hardware	AI pilots and autonomy stacks	More platforms, licenses and upgrades
Capability as a service	Guaranteed access to an operational capability	Satellite imagery and intelligence capacity	Annual or multi-year subscriptions
Engineering and mission services	Specialist people and project work	Cyber, R&D, integration, technical support	Contract extensions and additional labor
Sustainment	Keeping deployed equipment operational	Spares, maintenance, training, modernization	Installed base and equipment lifetime

Are the old defense primes still the biggest business model?

Traditional prime contracting is still the biggest defense business model by a huge margin, even though newer defense-tech companies are starting to compete for work that once almost automatically went to incumbents.

Lockheed Martin generated $75.0 billion of sales in 2025 and ended the year with $193.6 billion of backlog. That backlog represents more than two and a half years of annual sales. General Dynamics, Northrop Grumman, RTX and other large contractors operate at similar orders of magnitude that venture-backed defense companies have not yet reached.

The reason goes beyond incumbency. Governments need someone to be responsible for the whole system. A fighter aircraft, missile-defense network or submarine may combine thousands of suppliers, classified software, weapons, sensors, engines, communication systems and years of testing.

Large primes also mix several revenue streams. Lockheed's 2025 accounts show $62.7 billion of product sales alongside $12.4 billion of service sales. The same aircraft program can therefore produce development revenue, aircraft deliveries, spare parts, modernization and support.

What is changing is the list of companies allowed to play that role. Anduril is moving far beyond selling individual drones and sensors. The Army's enterprise procurement vehicle covers the company's software, hardware, data infrastructure, compute and support. Anduril also took over responsibility for future hardware and software development on the Army's soldier-headset program from Microsoft.

Prime contracting is still the center of gravity in defense revenue. The newer story is that a few younger companies are now trying to capture much larger pieces of that stack themselves.

Can defense hardware really scale like a product company?

Defense hardware can scale like a real product business when customers repeatedly buy a common design. The economics get much worse when every military unit wants its own version.

AeroVironment gives us a useful view of what productization looks like. Its product sales increased from about $586 million in fiscal 2024 to $693 million in fiscal 2025 and then $1.42 billion in fiscal 2026. Acquisitions helped drive the latest jump, but the underlying model remains clear: much of the company's growth comes from manufacturing and delivering physical systems rather than adding more billable engineers.

Anduril has pushed the same idea into a broader portfolio. It develops products before every customer requirement has been specified, invests private capital in factories and then tries to sell a common technical architecture across different programs. Its Army enterprise contract makes that easier because customers can order commercially available Anduril technology without creating a fresh contracting structure each time.

The attractive part is reuse. A company that has already designed the airframe, electronics, autonomy stack and manufacturing process does not start from zero when the next customer arrives. Additional volume can lower unit costs and spread engineering spending.

Customization is the trap. Military customers routinely want different radios, payloads, encryption, datalinks, environmental qualifications or software integrations. Enough of those requests can quietly turn a product company into an engineering contractor.

So the metric worth watching is fairly simple: how much of the product survives unchanged from one contract to the next?

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

Are missiles and cheap military drones basically repeat-purchase businesses?

Missiles, interceptors and attritable drones have unusually strong repeat-purchase economics because military customers must keep replacing equipment that gets fired, lost, tested or added to stockpiles.

Raytheon's current Tomahawk program shows the scale this can reach. RTX recently received a seven-year, $22.9 billion U.S. Navy contract that supports production rising above 1,000 Tomahawks a year. The company says it delivered three times as many Tomahawks in the first half of 2026 as in the same period a year earlier.

That increase sits inside a much broader production ramp. Earlier agreements committed Raytheon to at least 1,900 AMRAAM missiles a year and more than 500 SM-6 missiles, with production on several weapon families expected to rise two to four times from previous levels.

Low-cost autonomous weapons extend the same logic into a new price range. The Pentagon's Replicator program was built around fielding thousands of relatively cheap autonomous systems, and AeroVironment's Switchblade 600 was one of the first weapons selected. NATO has now launched its own Drone Edge initiative, with $40 billion planned for counter-uncrewed capabilities over five years.

The economic difference from ordinary equipment is simple. A radar can stay installed for twenty years. A loitering munition may disappear during its first combat mission.

Volume still requires heavy upfront investment. RTX spent $2.6 billion on capital expenditure in 2025 while expanding production capacity, which is why long-term procurement commitments matter so much for manufacturers.

Munitions are therefore a recurring defense business, with repeat demand coming from physical consumption rather than subscriptions.

Can defense software really make SaaS-like money?

Defense software can now produce genuine software economics, and Palantir has moved far enough beyond pilot projects to prove that the model works at multi-billion-dollar scale.

Palantir generated $4.48 billion of revenue in 2025 and an 82% gross margin. Government customers contributed $2.40 billion, up 53% in a year, including roughly $1.9 billion from the U.S. government.

The revenue structure looks much closer to enterprise software than traditional contracting. Customers pay for access to Palantir's software in cloud or customer-controlled environments, generally with continuing operations and maintenance. Revenue is then recognized across the contract period.

The customer-expansion numbers are particularly interesting. Average annual revenue from Palantir's top twenty customers increased from $64.6 million in 2024 to $93.9 million in 2025. That 45% increase shows what can happen once the software gets embedded inside a large organization: the same product finds more teams, missions and workloads.

The Pentagon is also making this model easier to scale. Its software acquisition policy now explicitly treats software-defined warfare as the current operating reality and tells defense organizations to use the Software Acquisition Pathway as their preferred route. Commercial Solutions Openings and Other Transactions are also being pushed as default approaches for qualifying software purchases.

Defense software still carries integration costs that ordinary SaaS companies can often avoid. Palantir's cost of revenue includes cloud infrastructure, subcontractors and field-service representatives because sensitive military deployments frequently need hands-on work.

Even with that burden, an 82% gross margin puts successful defense software in a completely different economic category from most hardware and services businesses.

Business type	Recent financial example	Approximate economics	What limits scaling
Defense software	Palantir: $4.48B revenue, 82% gross margin	Very high software margins	Deployment, procurement and customer concentration
Defense hardware	AeroVironment: $1.42B product sales	Roughly 32% product gross margin	Manufacturing, components and working capital
Intelligence subscriptions	BlackSky: $65M space-based intelligence and AI revenue	Direct costs were roughly one-quarter of revenue before D&A	Constellation investment and capacity
Government services	Booz Allen: $11.2B revenue, 11% adjusted EBITDA margin	Stable but labor-heavy	Hiring, utilization and billable headcount

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

Can autonomy software become a standard layer across military aircraft?

Military autonomy software has a credible path to becoming a reusable layer across many aircraft, and Shield AI's recent contracts are the clearest evidence that governments are starting to buy autonomy separately from the airframe.

The U.S. Air Force awarded Shield AI a production contract in 2026 to implement Hivemind autonomy software for its Collaborative Combat Aircraft program. The interesting part is how the Air Force structured the program: mission autonomy can be evaluated and upgraded independently from the aircraft itself.

Hivemind is already flying on Anduril's YFQ-44A as part of the program. Shield AI has also worked on integrations involving General Atomics, Kratos, Northrop Grumman, Airbus and other platforms. A recent L3Harris flight test combined Hivemind with an autonomous electronic-warfare aircraft.

That gives Shield AI a very different path from a drone manufacturer. If the software works across ten aircraft, the company can earn from ten platforms without building ten separate factories.

The main uncertainty is how reusable the software really becomes. Every aircraft has different sensors, flight controls, weapons, safety limits and mission requirements. Integration work can easily become expensive enough to eat into the software advantage.

Shield AI itself gives us a useful clue. When it announced an earlier integration on the Kratos BQM-177A, the company said it was getting faster at integrating Hivemind because the platform had been deliberately built for reuse across Defense Department hardware.

That is the number to watch more closely than the number of autonomy demonstrations. If each new integration needs substantially less engineering than the previous one, the economics become very attractive. If large parts of Hivemind still need rebuilding for every aircraft, the business starts looking more like an advanced engineering contractor.

The production award for Collaborative Combat Aircraft makes the first outcome more credible than it was a year ago, although the model still needs to prove itself across a much larger installed base.

Can governments subscribe to defense capability instead of buying the equipment?

Governments can already subscribe to operational defense capability, especially in commercial space, and BlackSky shows how a company can turn expensive physical infrastructure into recurring service revenue.

BlackSky's customers can pay for annual or multi-year access to satellite imagery, analytics and satellite tasking through its Spectra platform. Its higher-priority Assured contracts can effectively reserve collection capacity over specific areas on a take-or-pay basis.

One international defense customer signed a seven-year agreement worth more than $100 million that guarantees minimum annual capacity through 2032. More recently, BlackSky converted several smaller Gen-3 trials into larger recurring subscriptions. One customer moved from an initial pilot to a seven-figure quarterly run rate.

The economics become clearer in BlackSky's 2025 accounts. Space-based intelligence and AI services generated $65.1 million of revenue with $16.6 million of directly identified costs excluding depreciation and amortization. Mission Solutions, where BlackSky builds customized sovereign space infrastructure, had much heavier physical delivery costs.

Customers can now move between those models. BlackSky recently signed an eight-figure international agreement combining the sale of a Gen-3 satellite with recurring on-orbit operations and continuing imagery and analytics subscriptions.

The hybrid structure is especially interesting. A country that wants fast access can start by renting capacity. Later it can own dedicated infrastructure without abandoning the recurring software, operations and data relationship.

The same idea can extend beyond imagery. Communications, computing, cyber capabilities, logistics and some autonomous services can potentially be sold as continuing availability rather than one-off equipment.

For governments, that can avoid years of infrastructure development. For suppliers, it creates recurring revenue from an asset that can serve several customers. That's a pretty attractive trade.

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

How much of defense tech is still people doing services?

A surprisingly large part of defense technology is still people doing engineering, integration, cyber work and mission support, and those revenues behave very differently from software or standardized products.

Booz Allen is a useful reference point. The company generated $11.2 billion of revenue in its latest fiscal year, with $38.2 billion of backlog. Its adjusted EBITDA margin was about 11%.

The contract structure explains a lot. In a recent quarter, most Booz Allen revenue came from cost-reimbursable or time-and-materials work. In those contracts, governments are essentially paying for qualified people, hours, specialist knowledge and project execution.

There is nothing weak about that business. Clearances, mission expertise and deep government relationships can make these contracts sticky, and cost-reimbursable work transfers much of the cost risk back to the government.

Scaling is harder, though. A services company that doubles revenue usually needs a lot more employees. A software company can theoretically double usage without doubling the engineering team.

The distinction gets blurry because even product companies need service people. Palantir uses field-service representatives. Shield AI performs integration work. BlackSky runs advanced technology programs that adapt its platform to classified customer environments.

The useful test is whether those people support a reusable product or whether adding revenue keeps requiring another large engineering team.

Is sustainment where defense companies quietly make recurring money?

Sustainment is one of the most durable recurring business models in defense because every aircraft, engine, radar or vehicle already in service creates future demand for spare parts, maintenance, training and upgrades.

Lockheed Martin generated $12.4 billion of service revenue in 2025, even though products still represented most of the company. That service base sits on decades of previously delivered aircraft, helicopters, missiles and other systems.

RTX gave us an even cleaner example recently. Pratt & Whitney received a nearly $1.3 billion contract for F135 engine spare parts and support for the F-35 fleet. The F135 network now supports 42 bases and 13 ships, with more than 1,500 engines delivered across 20 allied nations.

Every new F-35 sold therefore expands the future maintenance network. More engines create more inspections, spare parts, depot work and modernization opportunities.

The installed base also makes displacement harder. A military that has trained pilots and mechanics, built depots, stocked spare parts and integrated software around one system faces real costs if it switches suppliers.

Newer defense-tech companies are already building the same logic into their models. Anduril's enterprise contract includes technical support alongside hardware and software. BlackSky can sell a sovereign satellite and then keep charging for operations and intelligence services. Autonomy vendors could potentially provide software upgrades for years after an aircraft enters service.

Once the installed base gets large, sustainment can keep the same customer paying for decades. Not glamorous, but very durable.

Who carries the risk when a defense product gets built?

Defense products can be funded by governments, private investors or the contractor itself, and the business can look completely different depending on who pays before the product is delivered and who absorbs the overruns.

Traditional contractors often develop technology through cost-reimbursable government contracts. The customer pays allowable development costs and the contractor earns an agreed fee. That approach makes sense when nobody knows exactly how expensive a new radar, aircraft or missile will be to develop.

Venture-backed defense companies increasingly take more of that risk themselves. Anduril has raised billions of dollars that can fund factories, R&D and products before every future customer order exists. Shield AI has funded some autonomy integration work using its own capital. Palantir sometimes pays for early customer pilots itself.

Private funding can make development much faster because engineers do not need to wait for a new government contract every time they change the design. It also creates a harder commercial test. If governments do not eventually buy the product, shareholders absorb much more of the loss.

The Anduril model shows the size of that bet. The company raised another $5 billion after reaching roughly $2.2 billion of annual revenue, with capital still being poured into manufacturing and new programs.

Large contract headlines create another layer of confusion. The $20 billion Anduril Army contract, for example, is the maximum value of a ten-year IDIQ vehicle. It makes future ordering easier, but it does not mean $20 billion has already been committed.

Contract type also changes who takes the pain when costs move. AeroVironment generated about $1.38 billion of its latest annual revenue from firm-fixed-price work, compared with roughly $454 million from cost-plus contracts. Fixed-price contracts can produce more profit when work is completed cheaply, but the contractor carries much more risk when costs run over.

Working capital adds another problem. A missile company may need to order components, hire employees and build a factory well before the government pays for finished units. RTX's recent long-term munitions agreements were deliberately structured with collaborative funding partly so the company could expand capacity without putting too much pressure on near-term free cash flow.

The strongest setup is usually one where development risk falls as the product matures: private or government money absorbs the early uncertainty, then repeatable production and longer-term orders take over once the design is proven.

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

Can a defense-tech company grow beyond one government customer?

Defense-tech companies can grow well beyond a single government today, but international sales and dual-use markets work best when the same core technology survives the move into each new customer.

The international opportunity has become too large to treat as secondary. The U.S. Defense Security Cooperation Agency reported $104.4 billion of implemented Foreign Military Sales transfers in fiscal 2025. NATO countries are simultaneously increasing budgets, and the EU's SAFE program makes up to €150 billion of loans available for defense procurement.

Lockheed Martin already gets 28% of its sales from international customers. For younger companies, the share can be even more dramatic. BlackSky reported during 2025 that roughly 91% of its backlog was tied to international contracts, while recent contracts have included customers buying subscriptions, sovereign satellite systems or combinations of both.

Selling abroad is easier when the core product remains common. The same imagery platform can serve several governments. The same autonomy software can sit inside aircraft from several countries. The same missile family can be ordered by a group of allied militaries.

Dual-use can widen the market further. Palantir generated 46% of its 2025 revenue from commercial customers despite its large government business. Commercial space infrastructure can serve both defense agencies and civilian users. Cybersecurity, logistics, AI and communications often have similar overlap.

But we should be skeptical when "dual-use" just means a defense product could theoretically have a civilian customer someday. A Tomahawk does not need a commercial market to be a great business. Neither does an interceptor.

Exports come with their own friction as well: government approvals, technology-transfer rules, local manufacturing requirements and political restrictions can all slow revenue.

Still, a company that can sell substantially the same technology to the U.S., several NATO allies and selected commercial customers has a much stronger growth ceiling than one that depends on a single program office.

Should a defense startup sell through Lockheed Martin or try to replace it?

A defense startup should usually sell through the primes when its real advantage is one reusable technology layer. Companies that gain something important from controlling the entire system have a stronger reason to become primes themselves.

Shield AI is a good example of the first route. Hivemind can potentially sit inside aircraft made by Anduril, General Atomics, Kratos, Northrop Grumman or other manufacturers. Shield AI does not need to win every airframe contract to benefit from the growth of autonomous aircraft.

A sensor company can make the same choice. So can a semiconductor company, communications supplier or specialist software vendor. Selling components into many platforms can create a broader installed base with much less manufacturing risk.

Anduril is moving the other way. Its advantage comes partly from controlling the relationship between sensors, hardware, autonomy and Lattice software. The company increasingly takes responsibility for integrated systems rather than supplying one module to an incumbent.

Owning the whole system increases the possible contract value. It also means owning far more of the problems: factories, supply chains, qualification, schedule delays, maintenance and system integration.

That trade-off explains why becoming "the next Lockheed Martin" should not automatically be the ambition of every defense startup. A company that can become the standard autonomy layer across five competing aircraft manufacturers may have a better business than one that spends billions trying to manufacture the sixth aircraft.

The right question is where the company's technology has the most leverage. If controlling the hardware makes the software meaningfully better, vertical integration can make sense. If the technology becomes more valuable as more manufacturers adopt it, staying one layer inside the ecosystem can be the smarter move.

Which defense-tech business models look strongest now?

The strongest defense-tech business models today combine reusable technology, repeat demand and enough contract visibility to justify long-term investment.

Software platforms have the cleanest economics. Palantir has already shown that a defense-heavy software company can produce gross margins above 80%, and customers can expand dramatically once the software becomes part of daily operations.

Embedded autonomy could eventually be almost as attractive. Shield AI now has a production award that separates its software from the aircraft, which is exactly the architecture the model needs. We still need more evidence that integration costs fall consistently as the number of platforms grows.

Capability subscriptions also look strong. BlackSky can sell guaranteed satellite capacity repeatedly while keeping the physical constellation under its own control. Its recent move from pilots into larger annual subscriptions suggests governments are increasingly comfortable buying intelligence this way.

Munitions deserve to rank near the top under current conditions. The margins are lower than software, but consumption naturally creates repeat orders. Multi-year contracts and very large production ramps now give manufacturers much better visibility than a series of isolated annual purchases.

Productized hardware is attractive when the design stays common. Sustainment becomes particularly powerful once the installed base is large. Services can be excellent businesses too, but they have less operating leverage because growth remains tied more closely to people.

Defense-tech model	Repeat revenue	Margin potential	Capital needed	How well it scales	Our view today
Software platforms	Very high	Very high	Low to moderate	Very high	Best pure economics
Embedded autonomy	Potentially very high	High	Moderate	Very high if integration gets cheaper	One of the most attractive emerging models
Intelligence / capacity subscriptions	High	High	High upfront	High once infrastructure exists	Strong recurring model
Munitions and attritable systems	Very high during replenishment cycles	Moderate	High	High with mass production	Exceptionally strong under current demand
Productized hardware	Moderate to high	Moderate	High	High if customization stays controlled	Strong but execution-heavy
Sustainment	High	Moderate	Moderate	Grows with installed base	Extremely durable
Engineering and services	High through renewals	Lower	Low	Limited by people	Reliable, but less scalable
Bespoke development	Irregular	Contract-dependent	Often shared with government	Low unless technology gets reused	Useful on the way to a product, weak as the end state

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

So what are the main business models in defense tech?

The main defense-tech business models are prime contracting, standardized hardware, consumable weapons, software platforms, embedded autonomy, capability subscriptions, engineering services and lifecycle sustainment, and the strongest companies increasingly combine several of them.

The traditional prime model still controls the largest pools of money. Lockheed Martin's $75 billion of annual sales and nearly $194 billion of backlog show how far even the biggest new entrants still have to go.

Yet the more interesting change is happening underneath those huge programs. Palantir has proved that defense software can have true software margins. Shield AI is getting closer to proving that autonomy can be sold independently from the aircraft. BlackSky is showing that governments will subscribe to operational intelligence capacity. Anduril is trying to combine software, standardized hardware and manufacturing into a new type of vertically integrated prime.

Meanwhile, physical weapons remain commercially powerful. The latest Tomahawk contract stretches across seven years and supports annual production above 1,000 missiles. When a product is consumed and governments need larger inventories at the same time, recurring demand can become enormous.

The weaker model is the defense-tech company that looks scalable from the outside but has to rebuild its product for every contract. A stream of bespoke government projects can generate plenty of revenue, yet the company never gets much easier to operate as it grows.

So we come back to one test across almost every defense-tech business: how many times can the company sell the same underlying technology?

The best businesses develop technology once, get it qualified, deploy it with an initial customer and then keep earning from the same technical base through more units, more users, more platforms, more countries, software upgrades, replenishment or sustainment.

That is currently the clearest dividing line between a defense contractor that can grow very large and a defense-tech company that can become genuinely scalable.

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.