The Infrastructure-Distribution-Data-Flywheel
The IDD Flywheel has always existed, and it will persist. AI does not replace the flywheel – it actually reinforces the flywheel’s core elements and makes it more important to understand.
I’m taking a slightly different approach this week, for the holidays. As opposed to a data-driven macro analysis that ties back in with current tech trends, this week’s post will be a bit more editorial in style. I’ll discuss what I refer to as the Infrastructure-Distribution-Data Flywheel – a timeless model for building companies that I believe is worth clearly identifying, especially in the age of AI. Hopefully, it gives current and future entrepreneurs a different lens to think through company building as they enter 2026.
The Flywheel, Start-Ups, and Galilean Motion
Almost every successful start-up in history has created an Infrastructure-Distribution-Data Flywheel, whether they realize it or not. A company builds a “wedge product” (infrastructure) and sells it to initial customers (distribution) – over time, when a company has enough customers, the company can gather enough data points to further inform how they build future infrastructure. The IDD Flywheel begins.
A flywheel is a physical entity that follows the laws of Galilean Motion. A company follows the same principles. Like a flywheel, companies start by creating a mass (infrastructure) – with enough force (force = smart, tenacious founders/employees), the flywheel starts to spin with a certain velocity (distribution), and if force is properly applied, we get acceleration (data) that increases the flywheel’s momentum. Momentum grows quicker as the flywheel spins faster – more mass, more velocity, and greater acceleration. To support growing momentum, companies need to apply more force (hire more people) – but if they don’t apply enough force or the force is pushing in a suboptimal direction, momentum slows.
I’ve laid out each element explicitly, because many of the current narratives in tech regarding AI don’t make sense in the framework of the IDD Flywheel. Those who believe there will be a select few companies who “win” in the AI era, or that AI will spawn a new generation of solo-employee unicorns, are overlooking the laws of company building (and physics?).
Below I highlight each element of the flywheel in greater detail to explain how they work together.
Infrastructure Comes First
Every company starts with infrastructure – the core intellectual property that makes a product possible. The infrastructure is the underlying system that allows something new, cheaper, faster, or more reliable to exist in the world.
That infrastructure solves a real problem well enough that someone – a narrow Ideal Customer Profile (ICP) – is willing to adopt it.
“Infrastructure comes first” may seem obvious – but in the age of AI (and inflated tech prices, especially at early-stage), this reality is increasingly overlooked. In the Attention Economy, some start-ups focus on distribution first without recognizing that distribution is earned, not manufactured.
Infrastructure is what creates the right to distribute.
Distribution Is the Gateway
Distribution is often misunderstood as the end goal – after all, distribution is what yields revenue. But in reality, it’s the second step – the middle of the loop. Distribution is both the output and the input that drives successful, venture-backed companies.
Early-stage companies often define a beachhead market – a specific, manageable niche within a larger market that serves as their initial ICP. Distribution within an ICP gives a company proximity: proximity to users, proximity to workflows, proximity to behavior. That proximity produces something far more valuable than revenue alone: data. Proprietary, high-resolution, real-world data generated as a byproduct of customers using the product in production environments.
This is where the flywheel kicks into gear.
Data Is the Compounding Asset
Data gathered through real distribution is different from synthetic data, scraped data, or one-off datasets. It is contextual. It reflects actual decisions, constraints, and trade-offs made by real users all through the lens of how customers interact with your core product.
Crucially, this data cannot exist in a vacuum at a great company. The data and the learnings from the data feed directly back into infrastructure. Companies that understand the flywheel use this data to:
Improve the core product
Identify adjacent problems worth solving
Fine-tune pricing, risk models, and workflows
Launch new products that competitors cannot replicate without similar data access
Core infrastructure ultimately begets more infrastructure, which then unlocks wider distribution, more data, and so on.
Ramp as a Case Study in the Flywheel
Consider Ramp: the $32bn fintech started as a corporate card company that offered 1.5% cash back and a free spend management platform. Today, there’s no part of the CFO stack that Ramp doesn’t touch in some capacity. At its foundation, Ramp built infrastructure: a card network tightly coupled with spend management software. It offered cash back and embedded controls, real-time insights, and automation directly into how small businesses spent their money.
That core infrastructure unlocked distribution through a specific ICP: small and mid-sized businesses looking for better financial control. Once Ramp had that distribution, they began to observe, at scale, how businesses actually spend.
That spend data became a strategic asset that almost no amount of spend can replicate. It informed product expansion into areas like banking, accounting, procurement, and broader financial tooling. These products were shaped by real patterns in real customer behavior.
The IDD Flywheel explains why Ramp raised five new rounds of funding in 2025. Investors in the company appreciate how their infrastructure, distribution, and data advantages compound. Ramp’s reputation of having one of the most exclusive hiring processes in all of tech suggests that they continue to apply force correctly.
Why This Still Matters in the Age of AI
There’s a growing narrative that AI collapses moats. That as models commoditize, only a handful of winners will survive. The flywheel suggests something different. AI doesn’t collapse moats – rather, it accelerates capabilities. It helps new market entrants build core infrastructure faster, and it helps make sense of data in ways that may better inform how companies build new infrastructure.
AI cannot guarantee distribution (although companies like Unify can help) and try as it might via synthetic data, it cannot create the most proprietary data sets that companies have access to, which they earn through distribution.
OpenAI partners with the likes of T-Mobile because they understand the value of this proprietary data. Enterprises in heavily regulated industries offer a structured distribution channel – and with it, unique data environments that can inform future infrastructure development.
AI does not replace the flywheel – it actually reinforces the flywheel’s core elements. The strongest companies understand how the elements of the IDD Flywheel interact, and they build core systems around one component of the flywheel with each other element in mind.
A Timeless Model
The IDD Flywheel is not tied to any specific technology cycle. It worked before cloud. It worked before mobile. It worked before AI. And it will continue to work long after this Substack piece fades into online oblivion.
The core premise of the IDD Flywheel is that with each customer interaction, a platform’s core strengthens. Companies that understand this and have the capabilities to execute upon it will always have an opportunity to build category-defining businesses.