Big Budget, Small Market: Why Defense Software Startups Never Hit Escape Velocity

When Teamworks raised its Series A 10 years ago, every investor meeting eventually arrived at the same question: how big can this get?

It was a fair question. Sports is a huge industry, but our slice of it at the time, communication and operations software for elite athletic programs, was decidedly finite. There are only so many Power programs and pro franchises. We could build the best product in the category and still run out of customers before we ran out of runway.

We had two ways out. The first was to take our operations software into other industries. Hospitals and construction companies came up more than once. Our reaction to that was less than enthusiastic. Our relationships were in sport, our knowledge was in sport, and our CEO came from sport. Selling the same product to nursing managers and construction site superintendents certainly would have created financial opportunity, but the mission of empowering truly rare human talent at the upper echelons of sport was just too compelling to water down by splitting focus that early in our journey.

The second option was to do everything in sport: NIL, recruiting, nutrition, strength and conditioning, academics, data and analytics… every software category that touches the athlete, we had to do it. Build a market big enough to matter by delivering the best product for every category in the vertical instead of one sliver of it. That path needed a lot of capital, which is exactly why it worked as an answer. Investors fund ambitions, and this one was big enough to justify the check.

It also worked because sport had a buyer for it. Selling operations software put us in front of leadership, the athletic directors and general managers responsible for everything that happens around the athlete. When we brought a new product to the market, we already knew the person who could say yes to it. 

I think about that decision a lot now, because I watch defense software startups face the same TAM question every day, and almost none of them have an answer.

Capital is insatiable

Smart capital wants growth above everything else. A software company that reaches escape velocity follows a familiar curve: a few years of tripling, a few years of doubling, then a long glide path where growth slowly decays but keeps compounding for a decade or more. Investors fund that curve because the end of it is a company worth many multiples of what they put in. Unfortunately, for most defense software, the arithmetic doesn't work. The market isn't big enough to carry that curve.

That sounds absurd when the FY2027 defense request is $1.45 trillion, a 44% jump over FY2026 enacted, built from $1.1 trillion in discretionary funding and $350 billion in proposed reconciliation money (link; link). (The White House rounds it to $1.5 trillion. Nobody in Washington is going to quibble over $50 billion.) Every pitch deck I've seen in this space has a slide where the TAM is the entire defense budget. I've built a few optimistic TAM slides myself, so I say this with love: look at where the money actually goes, not what you hope an investor is naive enough to believe.

A single Ford-class carrier, the future USS Enterprise (CVN-80), carries an estimated total procurement cost of $14.2 billion in the Navy's FY2026 budget submission, before aircraft or a single year of operations (link). The F-47 fighter, still in development, carries a request of more than $5 billion in FY2027 research and development funding alone (link). Palantir's 2025 Army enterprise agreement, the Army's headline software deal, has a ceiling of up to $10 billion over ten years (link). And that number is a ceiling. The Army is committed only to a minimum guarantee it hasn't disclosed. One aircraft carrier costs more than the entire ten-year ceiling of that deal.

The Pentagon is very good at deploying capital into hardware. Software that rides on that hardware, or that becomes the operational backbone everything else plugs into, can reach defense-scale dollars. Everything else competes for slices of O&Ma and IT budgets that never get carrier-sized. Human performance, readiness, family programs and most of the workflow software that keeps units running address needs that matter, but they’re doing it in categories that will never be multibillion-dollar markets.

Investors know this. Defense tech venture funding hit a record $49.1 billion in 2025, up from $27.2 billion in 2024, according to PitchBook data shared with Defense News, and exits hit a record $54.4 billion (link). PitchBook's Q4 2025 Defense Tech VC Trends report found autonomous systems led trailing-twelve-month deal value at $12.1 billion, followed by sensing, connectivity and security at $8.4 billion, with advanced computing and software at $6.2 billion. (PitchBook's First Look counts Five Eyes defense tech, dual-use companies included, across 861 deals; its later Trends report revised the total to $49.9 billion across 966 deals, and narrower definitions produce much smaller totals (link).) Capital is flowing into defense, but it's definitely flowing around the categories that can't return it.

And it costs more to get there

A small ceiling would be manageable if the climb were cheap, but it isn't. Nearly every structural feature of selling software to the DoW raises the capital required to reach whatever ceiling exists.

Compliance comes first, and it comes expensive: FedRAMP, IL4 and IL5, an ATO for each environment, continuous monitoring. When GAO reviewed FedRAMP in 2024, agencies and vendors estimated authorization costs "anywhere from tens of thousands to millions of dollars," and DoW impact levels pile on from there (link). Budget a year or more of work, all paid before the first dollar of revenue. Commercial-first companies can spread that cost across a much larger business with committed long-term revenue. A DoW-only startup pays for it out of its seed round with no guarantee that long-term sustainable revenue ever materializes.

Once you’re past that, the timeline stretches. The budget process, CRS notes, "typically begins more than two years before the expected year of budget execution," prototype money rarely converts cleanly into sustainment money, and continuing resolutions freeze new starts, which is a polite way of saying nothing new begins until Congress agrees on something (not something you can easily count on these days) (link). SBIR, the program built specifically to help small companies get going, lapsed on September 30, 2025 and wasn't reauthorized until April 13, 2026 (link; link). Six months, or 195 days if you were a founder counting. Every month of waiting is a month of runway burned.

Margins thin out when a prime integrator sits between the product and the customer, taking a share of every dollar and often controlling access to the actual end user. Acquisition costs climb because champions rotate. The leader who brought you in will likely PCS within about two years (GAO found the average time between PCS moves was "about two years," in an older but still-cited 2001 report), and the successor may have different priorities (link). Winning a unit once doesn't mean keeping it.

Finally, investors discount revenue from a single customer who can end a contract on a hard date and recompete you out. Even the revenue you earn is worth less.

So you get higher upfront cost, a longer road to revenue, thinner margins, more expensive growth and a discounted valuation, all against a smaller ceiling. Any one of these is survivable. Together, they put escape velocity out of reach for almost anyone who depends on the DoW alone.

Why the sport playbook doesn't transfer

The obvious answer is to do what we did in sport: stop selling one product and build the whole platform. Recruiting, talent evaluation, performance, family readiness. These problems rhyme across sport and the military, and an integrated solution would do soldiers a lot of good.

The trouble is ownership. In sport, the athlete-centered operation has an owner. In the military, the soldier-centered operation has a dozen.

Recruiting, talent management, performance and family programs sit with different commands and different funding lines. A brigade commander owns the outcome for their soldiers but not most of the budgets that drive it. Even within a single program, the pieces get bought separately. In the Army's Holistic Health and Fitness program (H2F), the software platform was competed on its own, while the coaches and cognitive performance specialists sit on a separate services contract worth up to $247 million (link; link). Adjacent needs live in budgets the program's dollars can't touch at all.

So the bundle has no buyer. And if no one can buy the bundle, no investor will fund a DoW-only company to build it. The same structure that fragments the military's problems makes the integrated solution unfundable.

The exceptions prove the rule

Of course, Palantir would be the company everyone points to as the counter-example. Its U.S. government revenue grew 90% year over year to $809 million in the second quarter of 2026, its most recent reported quarter (link). But Palantir plays in one of the few software categories big enough to allow that, and it's doing it during a historic budget surge. Meanwhile its U.S. commercial business grew 149% to $764 million and is closing in on the government side. Management raised full-year guidance to 82% growth, and CEO Alex Karp told CNBC the run "looks like this is going to go on for at least another 18 months."(link) The company best positioned in defense software still sees its fastest growth outside defense.

Palantir can grow on government dollars today, but it probably can't do it forever. Growth that depends on the customer's budget request jumping 44% in a year has an obvious ceiling, especially when Congress's markups haven't included the $350 billion in reconciliation money that makes up much of the jump (link). When budgets flatten, the only way to keep growing is to get into a metaphorical knife fight and take share from someone else.

The autonomy companies drawing the biggest checks teach the same lesson. Anduril raised $5 billion at a $61 billion valuation in May 2026, and Shield AI raised $1.5 billion at a $12.7 billion valuation in March (link; link). While the magic of their products is most assuredly in the software part of the stack, both build or bundle their own hardware, which lets them ride procurement dollars directly. Software attached to a platform someone else builds usually accrues to that platform's prime.

Why this is the buyer's problem

All of this could just read like a complaint about venture capital, but I don’t really care if investors have an opportunity to make money selling to the DoW or if new software startups want to focus on selling to the DoW.  The part I’m concerned about is the problem on the other side of this, which is inherited by the Army: vendor viability.

A software company that can't grow can't fund the next authorization, can't keep its roadmap current and can't invest in what the mission needs next year. Eventually it stalls out or gets absorbed by an integrator. Either way, the unit is left with stranded data and a product that stopped improving years ago. Soldiers don't care what a vendor's cap table looks like. They care whether the tech they rely on still works, and still gets better, five years from now.

The way out

The military's needs aren't shrinking, and its problems will keep rhyming with the ones elite sport has been solving for years. The open question is who funds the solution.

For most categories, the answer is companies whose commercial business has already built and paid for the platform. Teamworks spent years building the athlete-centered operation in sport, funded by a market that could buy it. When we brought it to the military, the Army didn't have to underwrite the decade-plus of R&D that we’d already put into our systems. The platform already existed, built for the same practitioners. Strength coaches, dietitians and athletic trainers move between sport and the military constantly, and the problems they solve don't change when the uniform does.

It also lets a company our size land one unit or one category at a time without betting the company on the next appropriation. The bundle is already built. Each piece can stand on its own.

There is some encouraging news on the buying side. When the Army announced the Palantir agreement, CIO Leonel Garciga said it would be "one of many enterprise licensing agreements" and that others were already "teed up." (link)  The Army has since signed one with Appian (link). If that model spreads beyond data platforms, the military may finally develop a way to buy the bundle. That would change the math for every company serious about this mission, and it would give soldiers the integrated, soldier-centered operation they deserve. I’m obviously biased in seeing this as a positive, because I know my company is sitting on a platform that’s already been proven commercially at the highest levels of elite operations and human performance. But until then, founders building for the DoW alone one silo/department/problem at a time (and buyers looking at solutions from these companies) should ask themselves the question our investors asked us: How big can this get? If the honest answer is "not big enough," the mission deserves a better plan than hoping the market grows to meet them.