AI Application Layer May Deliver Bigger Returns Than Infrastructure

The infrastructure versus application divide

Tech giants are investing unprecedented capital in AI infrastructure—power grids, chips, data centers, and foundation models. Yet history suggests the largest returns may come from a different layer entirely: the applications built on top of that infrastructure.

NVIDIA CEO Jensen Huang describes AI as a five-layer stack, with energy at the bottom, followed by chips, infrastructure, models, and finally applications at the top. While the first four layers make AI possible, only the fifth answers the critical question: what can we actually do with it?

Today, most capital and attention flows to foundational layers. Companies are securing GPUs, constructing massive data centers, and funding increasingly powerful models. This buildout is necessary, but it represents the means rather than the end product.

Lessons from previous technology revolutions

The internet boom offers instructive parallels. Investors poured billions into fiber-optic cables and server farms during the 1990s. Many of those infrastructure companies didn't survive to see the web they helped create. The infrastructure was eventually bought and repurposed, but the companies that defined the internet era—Amazon, Google, Facebook—were built on top of that foundation, not within it.

The smartphone revolution followed a similar pattern. Telecom companies upgraded mobile networks at enormous expense, while the application layer produced the era's defining companies: Uber, Instagram, WhatsApp, and Spotify. These applications didn't just use the infrastructure; they became the industry itself.

Why it matters

AI enables a fundamentally new business model: vertically integrated companies that don't sell AI tools to an industry but instead use AI to rebuild the industry's economics entirely. Traditional service businesses scale by adding people—more customers require more employees, creating natural limits on margins and growth. AI changes this equation by encoding knowledge-based tasks into systems that operate continuously, improve over time, and serve dramatically more customers without proportional staffing increases.

Legal services, accounting, insurance operations, healthcare administration, and publishing are among the sectors where AI could transform labor-scaled businesses into intelligence-scaled ones. The next generation of winners may not look like traditional SaaS companies selling tools to these industries, but rather like the industries themselves, rebuilt on an AI foundation.

The vertically integrated AI opportunity

Consider a law firm where software performs legal research and drafting, a publishing company where AI handles formatting and production, or an insurance operation where AI processes claims that once required entire teams. These companies could combine advantages rarely seen in traditional services: large existing markets, high growth, high margins, and unprecedented scalability.

The infrastructure race remains real and necessary, with major winners emerging in energy, chips, data centers, and models. But infrastructure buildouts eventually mature, and models become more commoditized over time. What endures is the question every technology revolution must answer: what can we do now that we couldn't before?

The details in this analysis were first reported by Yehuda Niv, founder and CEO of Spines and founder of Niv Publishing, writing for Calcalist.