Defense used to be slow by design. Long cycles, cost-plus comfort, and programs that ran so long you forgot why they started. That world still exists, but it’s getting crowded out by faster vendors that ship software, autonomy, and security capabilities on timelines that feel… normal.

And yes, the money is following it.

The spending backdrop (now verified)

The NATO numbers in the prior draft needed a correction. NATO’s latest Secretary General annual report (released in late March 2026) puts NATO Europe and Canada defense spending at USD 574 billion in 2025, which corresponds to 2.33% of GDP (as presented in the report). That’s still a clear step-up versus recent years, even if it’s not the 2.6% figure we had before.

On the U.S. side, the specific “$58B for advanced tech/autonomy/cyber” figure wasn’t something I could substantiate in primary budget docs. What I can support from the FY2026 DoD materials is that autonomy is explicitly called out and discussed as a priority area, with a stated $13.4B level for autonomy and autonomous systems in the FY2026 budget briefing. Space and cyber are also highlighted as major funding areas, but they’re discussed across multiple lines and accounts rather than one clean combined bucket.

Slight tangent, but it matters: when a theme is real, you don’t need perfect “one number” marketing slides to prove it. The signal is usually spread across a dozen line items and a bunch of unglamorous PDF tables.

Where attention is clustering

Late May 2026, the “popular” side of defense tech still leans hard toward software and autonomy. But a couple of the stats in the earlier version were off, so here’s the cleaned up version.

• Palantir (PLTR): In its Q1 2026 release/filings, Palantir reported U.S. government revenue of $687M for the quarter versus $373M in the prior-year quarter (those are U.S. government numbers, not total government segment). The earlier “government segment $373M” line was mixing these figures. If you’re tracking the stock, the debate hasn’t changed: strong growth and big contracts, but valuation still asks you to believe the good times last.
• Shield AI (private): Still private as of this writing. Because it’s private, the “options activity” angle needs to be treated carefully; you can’t have options on Shield AI itself. The more defensible way to frame it is simpler: any credible IPO chatter tends to pull attention toward the closest public peers in autonomy and defense software.
• Kratos Defense (KTOS): The prior draft’s revenue/backlog numbers didn’t match the most reliable sources I could find. In the company’s recent commentary around Q1 2026, backlog is described as record $2.0B. (The earlier $1.3B figure looks stale.) Kratos remains one of the cleaner “production plus autonomy” exposures for investors who want something more tangible than pure software.

One thing I’m watching: the market’s tolerance for anything that looks like “platform + national security” is higher than it was a couple years ago. Not infinitely high. Just higher.

The part people skip

The biggest difference versus past defense upcycles isn’t just spending levels. It’s who’s building. Venture-backed teams are showing up with product velocity and hiring plans that look like mainstream tech, then meeting defense buyers where they are. Old primes can respond, but they respond the way big organizations respond: slowly, and with a lot of process.

The obvious risk is policy. Program timing can shift. Priorities can change fast. And the same investors who love “growth” on the way up can decide they hate it on a random Wednesday.

If you want, I can turn this into a tighter watchlist with the exact links to the NATO report section and the relevant Q1 filings we’re pulling these numbers from. Worth a look if you’re building a real coverage list for 2026.

–