From asset-light to AI-heavy: infrastructure shift

Over the last 18 months, the narrative that “software is eating the world” has faded as attention shifts back to long-term assets and infrastructure.

Driven by demand from AI models, more data centers are currently being constructed than at any time in history. This demand is flowing through into foundries and semiconductor production equipment, as well as into raw materials, construction machinery, power generation and cooling systems, and the engineers needed to manage both construction and operations.

Surprisingly, AI-related infrastructure is not the only sector experiencing a capital expenditure super-cycle. The pace of nuclear power plant construction has reached levels not seen since the 1980s. Capital expenditure is rising significantly across global energy grids, particularly transmission networks and renewable power generation, including solar and battery storage systems, while investment is also increasing in satellite constellations, space infrastructure and even high-speed rail in China and Japan.

Figure 1: Technology companies accelerate capital expenditure on AI infrastructure

For the last twenty years, “asset-light” became the holy grail of investing, and the “Rule of 40”¹ was the admission ticket to an elite members’ club of the most successful cloud software providers. In contrast, when traditional industries grew, that growth typically required a linear increase in variable costs; to produce more widgets you might need a larger factory, more raw materials, more machines and more people to operate them. Levers to improve returns exist, but can be difficult to identify.

Digital businesses, by comparison, tend to be “asset-light”, since the cost of building more software or adding a user to a platform once the software or platform is built, is typically negligible. Many of the leaders in the digital world also benefit from positive network effects, where the value of their solution increases as more people use it, and since AI models need vast quantities of data to be trained on, companies with the largest platforms often have access to the deepest pools of digital data.

With so many advantages favoring the digital giants and their “asset-light” business models, what went wrong? On the one hand, AI, although still at an early stage of adoption, has challenged the “asset-light” model and raised questions about the durability of traditional software economics, as increasingly powerful models threaten to disrupt, marginalise, or even push parts of the traditional software industry toward obsolescence. On the other hand, AI also requires a vast wave of compute power to support it, creating a capital expenditure super-cycle across the physical infrastructure needed to process, transmit and power the next generation of digital workloads. This infrastructure includes the data centers needed to handle growing AI data volumes, as well as the transmission grids and power generation assets – from nuclear and solar to natural gas – required to meet the resulting increase in electricity demand.

Figure 2: US electric utility capex expected to surge

Source: Surging energy demand puts US utility capex forecast near USD 1.3T in 2026–30 | S&P Global, 26 April 2026

This shift from “asset-light” to “asset-heavy” models extends well beyond AI. Investment is also picking up across more traditional parts of the infrastructure universe, including midstream pipelines, transport networks, and telecom infrastructure. Long seen as part of the old economy, these assets have regained importance as the post-COVID shock, supply-chain disruption, the war in Ukraine and higher interest rates reminded investors of the importance of resilience, security, and control over critical networks.

This is the broader point behind HALO assets – heavy assets with low obsolescence – which are difficult to replicate, slow to build and often essential to the functioning of the real economy, making them increasingly valuable in a world of higher interest rates and rising strategic competition.

In a world increasingly shaped by AI, energy security and supply-chain security, the strategic importance of “asset-heavy” infrastructure is becoming harder to ignore. Whether this marks a lasting structural shift or simply a long-awaited rebalancing after years of “asset-light” dominance remains to be seen. However, the outlook for infrastructure appears extremely positive.

For long-term investors, the shift underscores the importance of companies enabling and supporting critical infrastructure. Our approach focuses on identifying businesses with strong positioning across data centers, power generation, grid modernization and industrial systems, where sustained capital investment and rising demand are likely to drive durable growth.

¹ A simple calculation used to measure the performance and health of fast-growing software companies: a company’s revenue growth rate (%) plus profit margin (%) should equal or exceed 40% to indicate a healthy balance between growth and profitability.