Every hype cycle has its main character. This one, on the surface, belongs to AI models — chatbots, copilots, agents. But underneath the noise, something quieter is happening: the market is starting to price the pipes, not just the magic.
That’s what makes the current moment around data infrastructure so interesting. It’s the part of the stack that has been boring on the surface and absolutely decisive underneath.
When “Plumbing” Starts Getting Premium Valuations
Consider what it means when investors start talking about data platforms with the kind of energy usually reserved for consumer unicorns and frontier labs.
Databricks sits in that zone. It doesn’t look like a flashy consumer brand. It doesn’t publish viral demo videos the way model labs do. It positions itself as a unified platform for data, analytics, and AI — the lakehouse that tries to be the place where data is stored, processed, and turned into something models can actually use.
The pattern worth noting is that markets don’t typically assign these kinds of expectations to companies that are merely “nice-to-have plumbing.” When valuations for data infrastructure players start nudging into the same conversations as pure-play AI companies, there’s a signal hiding there: the belief that whoever owns the data layer owns a large part of the future value chain.
Why Models Without Infrastructure Keep Disappointing
We’ve seen this in earlier posts in the AI readiness thread: impressive models deployed on top of messy, fragmented, poorly-governed data tend to look great in demos and average in production.
The uncomfortable truth is that most AI failures in enterprises aren’t model failures. They’re data and organisational failures dressed up as model disappointment. Wrong data, stale data, siloed data, ungoverned data — pick your flavour. The outcome is the same: a “smart” system that can’t see clearly enough to be useful.
That’s why the conversation worth having is no longer “which model is best?” but “what does our data infrastructure actually look like?” Platforms that unify storage, processing, governance, and analytics in one place are trying to answer that question in a way that scales. Databricks is one version of that bet. Snowflake and Palantir are others. The exact architecture may differ. The underlying claim is similar: if AI is going to matter, the data layer can’t be an afterthought.
The Stack Is Where the Story Gets Interesting
There’s a temptation to think of AI as just “models + data,” but the stack that’s emerging is much richer.
Data platforms at the base. Feature stores and transformation layers above them. Model serving systems sitting on top. Monitoring, observability, and governance wrapped around it all. It starts to look less like a neat diagram and more like an ecosystem. And the companies playing across multiple layers of that stack are starting to look structurally different from those focusing on models alone.
The lens worth applying here is simple: in a world where models can increasingly be accessed as services, the defensible value shifts toward the foundations — the data platforms that know your organisation intimately, the governance frameworks that keep you out of trouble, the tooling that turns raw information into something models can actually learn from.
The headlines may still belong to the model labs. The durable value may quietly accumulate elsewhere.
A Moment for the Infrastructure Builders
For people building in this space, this moment feels different from earlier “data is the new oil” platitudes. Back then, the slogan ran ahead of the systems. Now, the systems are catching up.
Enterprises that are serious about AI are discovering that they’re accidentally in the data infrastructure business whether they like it or not. The question isn’t whether they will invest in data platforms, but how coherent and intentional those investments will be. The organisations that see this as core infrastructure rather than a side project are already behaving differently — in budgets, in architecture choices, in the questions their boards are asking.
The pattern worth watching over the next chapter is simple: as AI becomes more pervasive, which companies end up being known not just for what models they use, but for how robust their data foundations are?
Because in the long run, the story of AI may read less like a tale of model breakthroughs and more like the moment when the pipes finally stepped into the spotlight.
When you look at your own organisation, does it feel more like a company using AI on top of its data — or a company quietly becoming a data infrastructure organisation that happens to use AI?
Let’s keep learning — together.
harish.balasubramanian 
Share your thoughts