Value flows to AI infrastructure, not startups

In the constantly evolving landscape of artificial intelligence investment, a critical shift is occurring that savvy investors and business leaders need to recognize. Lo Toney of Plexo Capital recently shared insights that challenge conventional wisdom about where the most significant value creation is happening in the AI ecosystem. Rather than application-layer startups capturing the lion's share of value, infrastructure providers are emerging as the true winners in this technological revolution.

Key Points

• Infrastructure providers like Nvidia are capturing disproportionate value in the AI ecosystem through their essential hardware and computing resources, while application-layer companies face commoditization pressure.

• AI startups face a "squeeze" scenario where they must generate substantial revenue to offset high infrastructure costs while competing against well-funded incumbents who can rapidly replicate their innovations.

• Venture capital investments in AI applications may face diminishing returns as the barriers to entry decrease and differentiation becomes more challenging in an increasingly crowded market.

The Infrastructure Advantage

The most compelling insight from Toney's analysis is the fundamental economic imbalance emerging in AI value creation. While media attention often focuses on flashy AI applications and use cases, the companies providing the underlying computational infrastructure—particularly Nvidia with its specialized GPUs—are positioning themselves to capture the majority of economic benefits.

This matters tremendously because it represents a departure from previous technology paradigms. In the mobile revolution, application developers could create substantial value with relatively modest infrastructure investments. The AI landscape operates differently: computational requirements for training and running sophisticated models are extraordinarily resource-intensive, creating a toll-bridge dynamic where infrastructure providers collect "taxes" on virtually all AI innovation.

For business leaders, this suggests a fundamental rethinking of AI strategy. Rather than viewing AI primarily as a software or applications play, the hardware and infrastructure components deserve equal if not greater strategic consideration.

Beyond the Infrastructure Thesis

What Toney's analysis doesn't fully address is the potential for specialized AI applications in regulated industries to maintain higher margins. While general-purpose AI applications may face commoditization, domain-specific solutions in healthcare, finance, and other highly regulated sectors could sustain premium pricing due to specialized knowledge requirements and regulatory barriers to entry.

For example, Tempus in healthcare oncology has built significant value by combining AI capabilities with proprietary clinical data