The $64B Sensor Gap the Pentagon Can't Close Fast Enough

In February 2024, a Houthi anti-ship ballistic missile streaked toward a commercial vessel in the Red Sea. The USS Carney fired two SM-2 interceptors. Each one costs roughly $2.1 million. The missile it was trying to destroy? Estimates put its cost below $50,000.

That math has haunted the Pentagon ever since. But the cost-per-shot problem is actually the second-order concern. The first-order problem is this: the electronics inside every interceptor, every radar array, every electronic warfare pod, and every sensor turret on the planet are running into a bottleneck that money alone cannot solve.

The bottleneck is processing. Specifically, the rugged, radiation-hardened, thermally managed compute modules that turn raw electromagnetic data into a targeting solution in microseconds. These aren't consumer chips. You can't order them from TSMC's standard menu. They're custom-built subsystems that fuse analog RF signals with digital processing, often operating in temperature ranges from minus 40 to plus 85 degrees Celsius, while being shaken, dropped, and exposed to directed energy.

During RTX Corporation's Q1 2026 call, CEO Chris Kellio mentioned a word that doesn't usually generate headlines: sensors. He called the company's sensor portfolio "underappreciated in the context of Golden Dome" (the administration's proposed missile defense architecture). That was a tell. Golden Dome isn't just about interceptors. It's about persistent, layered sensing from space to the ocean floor. And every layer needs processing hardware that can survive environments where commercial silicon would melt.

The defense electronics supply chain has been strained for three years. COVID-era chip shortages were the opening act. What followed was a structural shift: weapons programs that used to consume hundreds of processing modules per year now consume thousands. The Patriot system's AN/MPQ-65A radar uses digital beamforming that demands orders of magnitude more compute than its predecessor. The Navy's SPY-6 radar (built by Raytheon) uses modular building blocks, each packed with custom processing. The F-35's electronic warfare suite processes more data per second than most commercial data centers handled a decade ago.

And then there's the drone problem. Counter-UAS systems (think of what's deployed across the Red Sea, Ukraine, and increasingly along NATO's eastern flank) depend on fast, compact sensor fusion. A radar spots something. An electro-optical camera confirms it. An electronic warfare system jams it. All within seconds. The compute modules enabling that kill chain are the same class of product: ruggedized, integrated, purpose-built.

Teledyne Technologies' Q1 2026 results offered a window into this demand. CEO Robert Mehrabian reported a book-to-bill of 1.16 for the tenth consecutive quarter, with defense-related revenue growing at high single digits organically. He increased CapEx to approximately $150 million and bumped R&D spending by $10 million in Q1 alone. For a CEO known for patience and capital discipline over 25 years, that kind of acceleration says something specific: capacity is the constraint, not demand.

Boeing's defense backlog hit a record $86 billion. RTX's backlog reached $271 billion. Those numbers are abstract until you consider what fills them: aircraft, missiles, and ships that each contain dozens to hundreds of custom electronic subsystems. Every one of those subsystems needs someone to design, manufacture, and integrate the processing boards.

Here's the part most analysts miss. The big primes (Lockheed, RTX, Northrop) don't make most of these subsystems themselves. They buy them. The defense electronics tier sits below the primes but above the chip fabs, and it's occupied by a small group of companies with deep security clearances, long qualification cycles, and customer relationships that stretch back decades. Switching costs are enormous because every module must pass rigorous DO-254 or MIL-STD qualification. Once you're designed into a platform, you stay there for the program's 20- to 30-year lifespan.

One company in this tier has been operating since 1981. It started as a maker of signal processing computers for sonar applications, then expanded into radar, electronic warfare, and electro-optical systems. Over four decades, it built a portfolio spanning digital RF memory units (used in electronic warfare), signals intelligence payloads for small UAVs, and integrated subsystems that go inside the weapons made by every major prime contractor. Its products are, in a literal sense, the nervous system of modern defense platforms. And it spent the past two years working through a margin reset that left much of Wall Street looking the other way.