The next decade of outsized returns will not come from applications. It will come from owning the layer beneath them, and almost no institutional capital is pricing this correctly.
There is a version of technology investment that has dominated the past twenty years, and it is built on a simple and largely validated insight: software eats the world, and the companies that build the most widely used software capture enormous value. The playbook produced extraordinary returns across multiple generations of companies and investors. It also produced a market structure in which application-layer valuations are now compressed, competitive moats are eroding at an accelerating rate, and the marginal dollar of capital is chasing increasingly incremental improvements in increasingly crowded categories.
The next investment cycle will not be won at the application layer. It will be won at the infrastructure layer beneath it, the systems that every application, every AI model, every autonomous agent, and every digital business depends on to function. And that layer is currently mispriced by most institutional capital.
The infrastructure bet is not a defensive bet. It is the offensive bet of the next decade. The highest-conviction, longest-duration, most defensible return profile belongs to whoever owns what the applications run on.
The reason for the mispricing is structural and understandable. Infrastructure businesses do not produce the narrative metrics that drive investment committee decisions. They do not have viral growth curves, daily active user counts, consumer brand stories, or the kind of founder-driven narratives that make compelling LP updates. They have uptime statistics, capacity utilization rates, long-duration contract structures, and the kind of financial profiles that look, to most technology investors, like utilities.
And utilities have historically been valued like utilities: steady, defensible, income-generating, but not the source of venture-scale returns. The historical analogy led investors to systematically underweight infrastructure relative to its strategic importance in the technology stack. That analogy is now breaking down, for reasons that are structural rather than cyclical.
The convergence driving infrastructure demand is the product of three simultaneous forces, each significant on its own, collectively unprecedented in their combined effect.
The first is AI. Every large language model, every agentic system, every AI-powered application requires compute infrastructure, data infrastructure, networking infrastructure, and increasingly, governance and trust infrastructure. The infrastructure demand created by the current AI development wave is not incremental. It is exponential and supply-constrained. The companies that own the compute, the interconnects, the cooling systems, and the data center capacity are not commodity providers in a competitive market. They are critical path assets in the most consequential technology transition of the decade. Their pricing power, contract duration, and competitive defensibility are all increasing as a function of structural scarcity.
The second force is sovereignty. Governments across Europe, Asia, and the Gulf are making significant investments in domestic technology infrastructure as a matter of national security rather than commercial optimization. Data localization requirements, sovereign cloud mandates, domestic AI development programs, and critical infrastructure legislation are creating infrastructure demand that is policy-driven, long-duration, and explicitly not subject to the competitive dynamics that compress commercial returns. The intersection of infrastructure and sovereignty is producing some of the most durable investment opportunities of the current decade.
The third force is the transition from the application paradigm to the agent paradigm described elsewhere in this series. As computing moves from interfaces that humans navigate to systems that act autonomously on human intent, the value of application-layer differentiation decreases and the value of the infrastructure those systems depend on increases. Every agent needs compute. It needs data access. It needs authentication and trust infrastructure. It needs connectivity. None of these requirements are optional, and all of them are infrastructure plays.
The investment implication is a significant reallocation of attention and capital toward categories that have been systematically underweighted relative to their structural importance. Private infrastructure, sovereign technology partnerships, data center and compute capacity, connectivity infrastructure, and the governance and security systems that make autonomous systems trustworthy are all categories where the supply of institutional-quality investment opportunities is growing faster than the supply of capital equipped to evaluate them correctly.
The infrastructure bet is not a defensive bet. It is the offensive bet of the next decade. The application layer will continue to generate returns for the companies that successfully navigate the compression. But the highest-conviction, longest-duration, and most structurally defensible return profile belongs to whoever owns what the applications, and the agents that are replacing them, run on.