Exploring Growth Themes in Security, Rapid-App Development, and Scalability

Most of the conversation around AI investing focuses on the obvious: models, chips, and applications. But the more I look at how this cycle is playing out, the more convinced I become that the durable opportunity sits a layer beneath — in the infrastructure that lets everything else run securely, reliably, and at scale.

Three themes keep showing up in my research, and they're more interconnected than they first appear.

Security Is Becoming a Platform Problem

As compute scales, so do attack surfaces — both digital and physical.

Data centers now house model weights, proprietary datasets, and inference systems representing billions of dollars in intellectual property. Yet physical security at many facilities still relies on fragmented tooling: standalone badge readers, siloed camera systems, and outsourced monitoring that doesn't talk to the network layer.

The opportunity is in unification. Platforms that connect physical access events with network authentication and anomaly detection in real time turn security from a cost center into an integrated layer of infrastructure. As hyperscale buildouts accelerate, the companies providing that unified security fabric become structurally important — not just to AI companies, but to anyone operating at scale.

This isn't about selling cameras or firewalls. It's about owning the software layer that ties physical and digital security into one view.

Storage Is AI's Long-Term Memory

There's a useful analogy between how AI systems handle data and how human memory works.

High-performance SSDs are like working memory — fast, expensive, and used for the computations happening right now. But every AI system also needs the equivalent of long-term memory: somewhere to store checkpoints, training archives, logs, and the accumulated context that compounds over time. That role falls to high-capacity HDDs, which act as the persistent storage layer for everything that isn't needed in the current inference cycle but remains structurally critical.

As models grow larger and training runs produce more artifacts worth preserving, this expanding memory layer becomes increasingly important. It's not glamorous infrastructure, but it's the kind of thing that quietly becomes a bottleneck if it isn't there.

Lower Barriers to Building Mean More Demand for Infrastructure

This is the theme that ties the other two together.

AI is dramatically lowering the cost and complexity of building software. Tools that let individuals and small teams ship full-stack applications in days instead of months are creating a new wave of software — and every one of those applications needs deployment, security, and global distribution.

The interesting dynamic is that as it gets easier to build software, the infrastructure required to run it doesn't get simpler. If anything, the explosion in the number of applications increases demand for edge computing, DDoS protection, DNS, CDN, and the other services that sit between code and users. The companies providing that layer benefit from a structural tailwind: more builders means more customers, regardless of which specific applications win.

What This Means

These three themes — unified security, scalable storage, and infrastructure for a growing application layer — share a common characteristic: they're not bets on which AI model wins. They're bets on what every AI system, and increasingly every software application, requires to function.

The picks-and-shovels metaphor is overused, but the logic holds. When an entire industry is scaling rapidly, the companies supplying the structural necessities tend to compound quietly while the spotlight stays on whatever is newest and most exciting. The best infrastructure investments are the ones where demand grows regardless of which specific technology or application captures the market's imagination.

That's where I'm spending my research time.