From Silicon To Software: The Great Divergence

By Vigilant Wealth Management on February 25, 2026

Introduction: The Dawn Of A New Industrial Revolution

Artificial Intelligence (AI) is driving the most significant capital investment cycle of the modern era. Infrastructure spending is accelerating, while the rise of agentic AI—systems capable of autonomous decision-making and task execution—is beginning to reshape software development and the nature of knowledge work. While AI promises transformative gains in productivity, history suggests that value accrues unevenly across industries and stages of the business cycle. There is no guarantee that today’s incumbents will define the era of agentic AI.

Technology transitions are not linear, and market leadership can shift as monetization pathways evolve and new catalysts emerge. This paper examines the forces driving divergence across the AI value chain and where we believe the risks and opportunities are most consequential for investors.

AI Infrastructure: A Multi-Year Investment Cycle

The first wave of the AI cycle has been defined by extraordinary levels of investment in digital infrastructure (i.e., data centers). While companies and governments across the globe are investing in AI infrastructure, large U.S. technology companies account for the largest share of overall spending. In Calendar Year 2025, the hyperscalers – Amazon, Microsoft, Google, Meta, and Oracle – collectively spent $412B, up 72% year-over-year. Based on FactSet consensus estimates, aggregate spending is expected to exceed $650B in Calendar Year 2026. The chart below illustrates the scale and velocity of this investment cycle.

The primary beneficiaries of this investment cycle have been companies that are embedded in the semiconductor supply chain. This is the segment of the technology ecosystem where earnings growth has been most directly tied to AI capital expenditures. We believe this level of spending is rational given that demand for AI compute remains materially above supply. At the close of 2025, the hyperscalers had combined remaining performance obligations (RPO) of approximately $1.6 trillion. These backlog figures suggest that enterprise demand for cloud and AI services continues to expand.

While we remain constructive on the durability of the AI infrastructure cycle, uncertainty around timing and the scale of demand introduces risks. These same hyperscalers have materially increased capital intensity, defined as capital expenditures as a percentage of cash flow from operations, placing further pressure on near-term free cash flow generation. The capital intensity ratio has risen from 35% in Q3 2023 to 72% in Q4 2025, and consensus expectations suggest this ratio will move higher in the coming quarters.

To date, strong demand signals and expanding backlog have supported this elevated spending. However, sustained capital intensity increases sensitivity to execution risk. Should revenue growth moderate or utilization rates fall short of expectations, returns on incremental capital investment could decline, potentially pressuring valuations across the infrastructure supply chain.

Additional risks warrant consideration. First, a meaningful portion of hyperscaler backlog is concentrated among a limited number of large AI model builders. Second, infrastructure cycles have historically overshot, and the current supply-demand imbalance may not persist indefinitely. Third, physical constraints such as power availability and leading-edge wafer capacity remain gating factors.

Looking ahead, we believe the long-term structural demand for AI infrastructure supports continued investment across the supply chain. However, elevated capital intensity underscores the importance of selectivity and valuation discipline as this cycle evolves.

Software: A Structural Reset, Not A Cyclical Correction

In contrast to infrastructure, the enterprise software sector is undergoing a significant valuation reset. Multiples have compressed materially from historical peaks, reflecting growing uncertainty around AI’s implications for the traditional Software-as-a-Service (SaaS) model. The chart below illustrates how Enterprise Value (EV) to Next-Twelve-Months (NTM) Sales – a standard valuation metric – has diverged sharply between software (IGV) and semiconductors (SOXX).

We believe this correction represents a fundamental reset of expectations rather than cyclical weakness. For years, software commanded premium valuations based on the durability of recurring revenue models, economies of scale (build once, sell many), and high customer switching costs. While near-term fundamentals remain solid for many vendors, the sustainability of these competitive advantages faces unprecedented pressure.

The rise of AI-native development tools represents more than a productivity enhancement, it challenges the very defensibility of the traditional SaaS model. The cost of software development is collapsing. When software can autonomously generate, test, and deploy new functionality, the barriers to building competitive applications decline materially. What once required scaled engineering teams and long development road maps can now be achieved by smaller teams leveraging AI copilots and agents. As development cycles compress and replication risk rises, the durability of moats rooted in proprietary code and development scale becomes less certain. We see three structural challenges confronting traditional SaaS vendors:

• Disintermediation Risk: AI agents and workflow automation tools threaten to bypass traditional application interfaces entirely. If agents increasingly are the intermediary between interactions across systems, the interface layer could lose strategic importance.
• Seat-Based Compression: The traditional per-user pricing model breaks down when AI agents perform work previously done by human users. Productivity gains may translate to fewer seats or increased pricing scrutiny, pressuring the economics of traditional licensing agreements.
• AI Monetization Uncertainty: While software vendors race to embed AI capabilities into their products, the path to monetization remains unclear. Capturing value commensurate with the productivity delivered (i.e., consumption-based pricing) will be critical to sustaining growth and protecting margins.

Traditional systems of record face particular vulnerability. While they possess valuable data moats and benefit from high switching costs, they are also primary targets for disruption by AI-native competitors unburdened by legacy architectures and business models.

Despite these headwinds, we do not believe that this moment represents the demise of all enterprise software vendors. Instead, this period of transition will reward differentiation and execution. Incumbents possess advantages: proprietary data sets, knowledge of workflows, and proven expertise in governance, compliance, and security. Companies that position themselves as foundational platforms for building and deploying AI agents may ultimately strengthen their strategic relevance.

Conclusions: Navigating the Great Divergence

The divergence across the technology sector reflects a fundamental reassessment of where AI creates durable economic value. Transformative technology cycles have consistently rewarded those closest to the infrastructure layer first, while the broader application layer navigates a longer, less certain path to monetization. In this cycle, semiconductor companies have been the earliest beneficiaries, as their earnings growth is directly tied to infrastructure spending. Software faces a more complex transition. The structural headwinds confronting the traditional SaaS models – AI native development, pricing disruption, and uncertain monetization pathways – are not transitory. While selective winners will emerge, broad multiple expansion appears unlikely without clearer evidence of durable AI-driven revenues.

Near-term capital allocation continues to favor infrastructure, where the earnings backdrop remains well supported. The software recovery will be uneven and, in our view, contingent on demonstrated progress toward AI-native monetization. Across both segments, selectivity and valuation discipline remain the appropriate posture.