The Realities of 21st Century Defense Contracting

2022 June 1 Twitter See all posts


Atoms are more important than bits when killing people. This post could loosely be a book review of "The Kill Chain" by Christian Brose.

Defining the Kill Chain

The "Kill Chain" is:

  1. Identifying a target
  2. Transmitting relevant information to an appropriate weapon system
  3. Executing the kill

A chain can be within a single system, such as an anti-aircraft gun with its own radar. It can also span across many platforms. A drone might identify an enemy troop concentration and transmit the information to its operator in Nevada. That operator sends the data to a battalion command post in the theater that relays firing coordinates to a platoon fire control center for several artillery pieces. It can finally reach an actual weapon from there.

Militaries are more effective the faster and simpler their kill chains are.

Realities of Defense Spending

DOD Will Always Be a Bad Customer

Given politics, changing priorities, budgeting, and the fickleness of war, Dept. of Defense (DOD) procurement is often messy. Programs change halfway through. Big budgets can evaporate overnight. It might take years for the military to make decisions.

A recent example is the Pentagon wants to cut spending on JDAMs (smart bomb kits) by 88% from 2021 to 2022. That could put manufacturers and their workers in a tight spot. Abrupt changes like this are common.

You might wonder why companies bother. The military is aware that it's a poor customer and doles out much more generous contract terms than a commercial customer would offer. Defense contractors often get large amounts of money for R+D or work off cost-plus contracts that eliminate their risk. The military pays way above commercial rates to compensate for extra requirements. And DOD usually pays companies to give them a demo or pilot of their product.

The Tyranny of Manned Systems

The US military is an all-volunteer force. We expect it to fight in an expeditionary role far from home to protect US interests and global commerce.

Many Americans would lay down their lives to fight an invasion on US soil. Fewer people are willing to die for the right to import affordably priced TVs and oil. It is critical to military recruiting that potential enlistees are more likely to kill an enemy than die themselves. No one is signing up to be cannon fodder. Besides, training a soldier costs a lot of money, and soldiers aren't easy to replace.

Building the most capable and survivable weapon systems is incredibly expensive. The military can only afford a limited number of designs in relatively small numbers. It is challenging to meet a wide range of needs with only a few weapon systems.

A death spiral follows where military planners can't afford to build a new weapon for a specific need. Instead, they add a requirement to an existing program. Increasing requirements escalate costs meaning the military can't buy as many systems. Fixed cost desorption further increases unit costs.

Cost overruns in manned systems are hard to escape.

Silicon Valley's Struggle To Break In Defense Tech

Investors and startups are often critical of the DOD.

"Palantir Is the Only Success."

Startups and venture capitalists want to win lucrative government contracts. There have been some successes. SpaceX, which gets a sizeable portion of its business from the government, is worth more than Boeing and nearly as much as Lockheed Martin.

But success is limited for software-only companies. Investors and startups whine that only Palantir, which does data analysis, has had a positive exit.

Shands Pickett and Katherine Boyle both complain that defense-oriented software startups are withering on the vine and DOD's practice of paying for demos and pilots is unfairly leading on these companies. I couldn't find a single company with a physical product on a list that Pickett references.

I find it entertaining that these companies and investors criticize the military for being incompetent at procurement and technology when they only provide a sliver of a kill chain solution. The DOD has no hope of integrating software-only technologies if the criticism is correct. It is making the right decision not to award production contracts.

Where Success Lives

There are successful and innovative defense contractors.

Examining Successful Companies

SpaceX, AeroVironment, Anduril, Kratos, and AEVEX are examples of companies winning billions in contracts while providing incredible capabilities for the US.

Successful defense contractors have several features:

  1. They are good at winning contracts.

    Selling to the government is a skill that is impossible to overstate. A driver for industry consolidation is that the companies that can win contracts buy talent and technology from those that can't. Most of these companies are decades old and have honed their craft.

  2. Their revenue is diversified.

    Government contracts often have long sales cycles and can be volatile. Long-lived companies have a wide range of contracts while mixing exports and commercial business. Most of AEVEX's job openings relate to contract intelligence analysis, while they also won contracts for the Ghost Phoenix loitering munition. Anduril's original focus was on flying drones but has built static camera systems and recently acquired an unmanned submarine maker.

  3. Innovative companies are usually medium size.

    SpaceX, Kratos, AeroVironment, and Anduril all have 1000+ employees. The companies need to be big enough to have diverse contracts but not so big as to lose their dynamic nature. Even big firms can have their moments. Boeing's Orca Extra-Large Unmanned Underwater Vehicle (XLUUV) provides as much capability per dollar as many nimbler companies' products.

  4. Companies create usable products.

    The contractors do most of the integration and provide a ready-to-use product or service. Raven and Switchblade drones span the entire kill chain. Kratos's Valkyrie drone offers incredible capabilities for a few million dollars.

Make What the Military Needs

It is hard to kill a human with a bit. And the military's business is killing. So it is no wonder that software-only startups struggle to fit in while companies that create physical, software-enhanced products rake in the contracts.

When companies create weapons with product-market-fit, like AEVEX's Phoenix Ghost or Anduril's kinetic drone defense, they win contracts in record time.

Most of these successes are unmanned systems. Avoiding the bloat, politics, and complexity of manned systems is critical in providing a good product at a reasonable cost.

Shore Up the Supply Chain

Pickett is correct about the opportunity for pure-software companies to be subcontractors for large defense contractors. The F-35 program lists over 1900 companies as suppliers. There is no shortage of business to win.

Additionally, the physical technology for manned systems is mature. Systems like the M-1 Abrams tank, B-52 bomber, or F-22 fighter have been around for decades. Designs changed rapidly in the 1950s and 1960s, but underlying inputs like energy and materials technologies haven't advanced enough to justify new designs. The military upgrades older systems with new electronics and software instead. Legacy contractors need help doing this!

Defense-related manufacturing companies rarely automate because of rapid changes in demand and low production volumes. There is an opportunity for startups to compete with Tier-I and Tier-II suppliers by using more flexible automation in manufacturing. Hadrian is an example as they try to make lower-cost aerospace components at faster turnaround times. SpaceX produces ~85% of the parts they need not because they want to but because no one else can.

Is VC the Right Model?

Since sales cycles are long and the government provides generous R+D contracts, one might wonder if venture capital makes sense as a funding model.

Founders will often be much better off winning contracts for development. Elon Musk owns a much larger share of SpaceX than Tesla because of business from NASA and the military.

Once you have the skill to win contracts, you can pick up some lower-margin contract work to lower your burn rate instead of raising more money and diluting ownership more.

Short contract lengths and rapid budget changes limit the multiples investors pay for exits.

VCs whining isn't necessarily a bad sign for America. It is a sign the system is working if software-only companies aren't winning contracts and innovative defense contractors can raise money in other ways.

It is safe to conclude that VC funding won't be as dominant as in the commercial software industry, but it still has a place.

Rethinking Military Software

Why can't I program a bomb like I use my iPhone?

Setting Realistic Expectations

Writing good software is difficult, expensive, and time-consuming. As consumers, we are spoiled by how easy our phones are to use. Critics expect the military to have software as capable as our phones.

If you examine the numbers, it quickly becomes apparent that the military can't afford iPhone-level software. Apple, Google, Microsoft, and Facebook had combined operating expenses of over $600 billion in 2021. The military's total budget is around $750 billion.

The mass of all the physical products these companies sell is probably less than one Ford-class aircraft carrier, and the number of SKUs is relatively limited. And remember, a defining feature of the software business is that marginal cost is near zero. It costs about the same to design high-quality software for 100 F-35s as for 200 million copies of the plane.

Yikes.

Implications of Software Developer Productivity

The hits keep coming. Software security is mission-critical for the military.

It costs extra money to do security audits and write more secure code in the first place. Supply chain attacks limit what code libraries developers can use.

The sheer number of platforms and contractors that work for DOD means that coordination between companies and specific weapons (some of which can be 80 years old) is expensive. The military does not have the luxury of vertical integration like Apple does.

We must reconfigure our expectations on what software can do for warfighting and focus on the highest value uses.

Internet of Things vs. Crypto Networks

The DOD's software and networks already resemble the "Internet of Things." Devices don't interact as advertised, security is often poor, and no one knows how to get all the features to work. IoT is the wrong framework for the military's software architecture.

The US military doesn't need a blockchain, but it faces many of the same challenges crypto networks do. Architectural decisions will be similar. Some major issues are:

  1. It is expensive to coordinate software and communications within its ranks.
  2. Security is paramount.
  3. Downtime is not acceptable.
  4. Enemy jamming or loss of equipment can limit throughput.
  5. Highly centralized nodes make the network vulnerable.

Crypto networks have several features to address these challenges:

  1. They use a simple base layer protocol.

    The simplicity makes coordination between nodes possible, and security is tight - less code means less potential for bugs.

  2. Only the minimum amount of data goes on-chain.

    On-chain storage and computing are scarce, like communications bandwidth in a war zone. It is logical to transmit the bare minimum and have the recipients do more computing.

  3. Nodes interact in limited, pre-prescribed ways.

    Smart contracts, nodes, and other actors have strict rules on what kind of transactions can happen. They operate like an application programming interface (API) that defines how other participants should interact with them. These features improve security by limiting attack surfaces. They also add "composability." Composability makes it easier to create higher-level programs by combining smaller code functions.

Similar architectures provide features desired by the military while economizing on how much code an effective network needs. The military already has an analog protocol called Link-16. It is compatible with most aircraft and ships. An ideal solution might look something like a digital Link-16 that can hop across more frequencies and is cost-effective to integrate into the smallest platforms.

Ditch the User Interface

In systems, it matters not only how much total complexity there is but where it lies. You can have an incredibly complex weapon that is simple to use. Or you can have a relatively simple weapon that requires a lot of training to use or integrate with other systems.

The British NLAW anti-tank missile is an example of a complex weapon with simple operation. The operator points it at a target, fires, and the launcher and missile do the rest.

You can also have a relatively simple weapon like a howitzer that requires dozens of selections on drop-down menus to adjust settings or plan a shot.

It takes a significant amount of code and effort to create user interfaces. The howitzer likely has less "machine" code that aims the barrel, but it could have more code overall due to bloat in the user interface. Additionally, soldiers need copious amounts of training to use these complex menus.

A better approach is to ditch the user customization. You can give soldiers something akin to a "big red button" requiring fewer inputs using modern software (like neural nets) and information derived from the protocol to calculate or fill in the rest. Specialists can still make changes using universal interfaces like the command line and command line code editors. And no self-respecting military would deploy something like an artillery piece without manual controls and troops trained to do basic ballistics calculations.

The sparse UI approach pushes the complexity into the "machine code," reduces training requirements, reduces human error, and speeds up the kill chain. Massive spillovers from commercial technology make more complex backends possible. Their cost declines every day while complicated user interfaces get more expensive. The military already has enough equipment they can't fully use or maintain.

The Fallback Strategy

Even a minimal protocol would be challenging for DOD to employ across its entire organization. It is never easy to get the six branches to agree.

An alternative is the vertical integration of the kill chain. A classic example is the Navy's Phalanx CIWS Gatling gun, which targets and destroys incoming missiles with integrated radar, software, and Gattling gun. AeroVironment makes the Raven scouting drone that sends targeting information to the Switchblade loitering munition.

The vertically integrated systems avoid coordination problems by keeping the entire kill chain within one firm. Under this paradigm, BAE Systems, which manufactures US howitzers and precision-guided artillery shells, would build a cheap artillery spotting drone.

Vertically integrating kill chains may make sense even if the military succeeds in creating a base-level protocol. It can support higher-level integration faster and cheaper than multi-firm development.

Closing the Kill Chain

Rely on Decentralized Decision Making

The undisputed advantage of the US military is the superior training of its lower-level officers and non-commissioned officers (NCOs). These leaders allow our fighters to react faster to the tumultuous battlefield and seize an opportunity.

Part of the weakness of many existing digital systems is that their kill chain runs through higher-level command posts. Often a staff officer creates firing solutions on their computer in a tent instead of a pinned-down platoon sergeant calling fire directly. Centralized solutions are slower and vulnerable to decapitation strikes from near-peer foes.

Any system has to take advantage of the quality of the individual soldier and let them interact directly with the firepower.

Future is Relentless Focus on Kill Chain

A good heuristic for success is if a sergeant can kill a destroyer within seconds of knowing its position and without talking to HQ.

Decisions must happen close to the fight without being looped through extensive command and control infrastructure. Any company building products that shorten and decentralize the kill chain will likely do very well. We shouldn't expect the military to integrate advanced software products, meaning defense contractors need to do the heavy lifting by using limited inputs or vertical integration to close the chain.