Everyone wants to be at the top of the stack. The flashy product. The headline number. The robotaxi. Nobody wants to be the pipe. And yet, every week, if you look at where real, durable value is actually accumulating, it's in the boring layer underneath the thing everyone is talking about.
We're watching this pattern play out across four completely different industries right now, and it's telling founders and operators something they need to hear.
The floor, not the ceiling
Anthropic's annualized revenue just cleared $65 billion, adding $18 billion in two months. That's a staggering number. But notice what's actually being sold. Not the AI feature your product uses. Not the chatbot. The underlying model. The foundation. The layer that every shiny demo up the stack depends on. Anthropic is the concrete under everyone else's skyscraper. Boring? No. Foundational? Yes. And foundational compounds.
Meanwhile, autonomous freight trucks are quietly scaling across Europe while everyone's fixated on robotaxis. Think about that. Long-haul trucking routes are more predictable, more structured, and easier to automate than chaotic urban roads. Yet robotaxis got 90% of the attention and investment. The real adoption is happening on the highway, moving pallets, out of sight. Infrastructure. Again.
And then there's DSM-Firmenich scaling yeast protein as a replacement for fishmeal and whey, quietly turning India's underused ethanol capacity into a new protein supply chain. Nobody's writing thinkpieces about this. But fishmeal feeds fish farms and livestock. Take fishmeal out of the equation and you've restructured how food gets made globally. That's not a consumer product. That's substrate. That's the invisible input that everything downstream relies on.
What the fluid theorists figured out
Here's where it gets interesting. Physicists just redefined how we understand the behavior of fluids from the ground up, replacing a framework that stood essentially unchanged since the 1800s. The breakthrough wasn't a new application built on top of the old theory. It was a new understanding of the base layer itself. They went deeper, not higher. And that change ripples up through everything built on those assumptions.
That's the pattern. In science, in logistics, in protein supply chains, in AI: the people who control the underlying model, the fundamental layer, the unsexy substrate, end up capturing most of the value. Everyone else is just renting from them.
And then there's the vulnerability
Here's the part that stings. A flaw in a Snowflake GitHub Actions workflow let a crafted issue inject commands and expose internal credentials. Not a breach through some exotic zero-day. Through a workflow file, sitting quietly in a public repository, doing something routine. The attack surface wasn't the product. It was the plumbing between the tools. The infrastructure connecting things.
This is the dual-edged reality of owning the infrastructure layer. It's where the value lives. It's also where the exposure lives. The pipe carries the water and, if you're not careful, the poison.
What this means if you're building something
Most founders we talk to are building at the application layer. That's fine. That's where products live. But the ones who get stuck, the ones who hit a ceiling at $1M or $2M and can't figure out why they're working harder for the same output, usually have one thing in common: they're completely dependent on infrastructure they don't understand and don't control.
Their customer data lives in a SaaS platform that can change pricing tomorrow. Their automation runs inside a workflow tool that can break, get acquired, or sunset a feature without notice. Their integrations are duct-taped together between APIs they've never actually read the terms of service for. They're not owners. They're tenants. And tenants are always one lease renewal away from chaos.
We're not saying you need to rebuild the foundation from scratch. That's the wrong takeaway. What we're saying is: know which layer you're on, know what's under you, and have a plan if it shifts.
Three things to audit this week
- What would break if one vendor disappeared tomorrow? Not "would it be annoying," would it break you. If the answer is yes, that dependency needs either a fallback, a contract, or a migration plan.
- Where is your workflow sitting in public? The Snowflake flaw was in a public repo. If your CI/CD, your automation, your integrations have any surface area exposed, someone on your team needs to audit it. This week. Not Q3 planning.
- What substrate is your product built on that you're not paying for? Open-source libraries, free tiers, deprecated APIs. These are borrowed infrastructure. Great for moving fast. Dangerous if you forget they can be revoked.
The companies pulling away right now, in AI, in logistics, in alt-protein, in everywhere else, are not the ones with the most elegant front-end. They're the ones who went one level deeper than everyone else and stayed there. The boring layer is where the compounding happens.
You don't have to own the foundation. But you'd better know exactly who does.