When you trace the arc of the current technology cycle, a clear pattern emerges across a string of companies—Palantir, then DataDog, then InnoData, then Snowflake. Read together, these names tell a single story: the data layer has become the point of monetization. The center of gravity in artificial intelligence is shifting. It is no longer just about chips. The value is migrating up the stack to where data is stored, organized, and put to work.
From Silicon to Software
The early narrative of the AI boom fixated almost entirely on hardware—on the silicon that makes large models possible. That focus was understandable, but it was incomplete. What recent earnings have revealed is that monetization is increasingly happening on the software side, at the data layer. This is the most important signal to take from these results. Snowflake matters not merely as a single stock that keeps climbing higher, but because of what its trajectory says about the broader shift in how value is captured. The same dynamic showed up in Salesforce, particularly around its agent offering, where the move toward monetizing data-driven software products became visible.
Data Warehousing as the New Gold
Think carefully about what data warehousing actually represents. The data layer is the new oil, the new gold. It is the moat. Palantir's defensibility rests on it. DataDog's moat is built on it. And it is, in all likelihood, what Oracle's enduring advantage will ultimately be as well. The reason is straightforward: whoever controls the data infrastructure controls the substrate on which every intelligent application depends. Models are only as useful as the data they can reach, and the systems that house, structure, and govern that data hold a position that is exceptionally hard to dislodge.
The "AI Ghost Train" and Its Limits
There is a pervasive anxiety circulating through markets—call it the "AI ghost train." It is the fear that the large model labs will steamroll every existing software company, rendering them obsolete. This worry is not entirely baseless. Some companies could indeed be disintermediated, cut out as foundational AI providers absorb functions that once belonged to standalone products. We are already seeing fragments of that play out.
But the fear is overblown when it is applied indiscriminately. The decisive countervailing force is this: data infrastructure and the installed base are key. A company that owns the data layer—and the entrenched base of customers built on top of it—possesses a defensibility that a powerful model alone cannot easily overcome. The frontier labs may be formidable, but they do not automatically inherit the data, the integrations, and the customer relationships that have been accumulated over years.
Conclusion
The lesson of this moment is that the real prize in artificial intelligence is being claimed not at the level of raw computation but at the level of data. Monetization has moved to the software, and within software, to the data layer specifically. For investors and builders alike, the question worth asking is not simply who has the best model, but who controls the data and the infrastructure that the models cannot function without. That is where the moat lies, and that is where durable value will accrue.