Recent headlines have reignited debate over the sustainability of the artificial intelligence (AI) investment cycle. Announcements from major technology companies, including Microsoft¹, indicating a moderation in near-term data centre spending, have raised questions about whether AI’s rapid growth is already slowing. Yet, history shows that true innovation rarely moves in a straight line. Like automobiles, smartphones, and the internet before it, AI is entering a phase of consolidation and evolution, not decline. Beneath the noise, the structural demand for AI infrastructure remains intact, supported by sustained growth in compute workloads, semiconductor demand, and broader adoption across industries.

Key Takeaways

Recent data centre spending cuts reflect a recalibration of priorities, not a slowdown in AI adoption.
Structural demand for AI infrastructure remains intact, supported by sustained compute growth, semiconductor demand, and broadening AI adoption.
Market leadership is evolving across the AI value chain, with a current tilt towards adopters, while semiconductors remain critical to long-term growth.

A Recalibration, Not a Retreat

Over the past two years, capital expenditure in AI infrastructure surged at an unprecedented pace. Global data centre spending is forecast to exceed US$2 trillion over the next five years, reflecting the scale and speed of AI-driven compute demand.² This growth was never expected to continue unchecked. What we are witnessing is a recalibration following an intense build phase, not a structural downturn.

Much of the early-stage infrastructure required to support large language models (LLMs) and foundational AI systems has already been deployed. Companies are now shifting from aggressive capacity buildout towards optimising and monetising existing infrastructure. This pattern is consistent with previous technology cycles, where heavy upfront investment is followed by a period of consolidation and efficiency gains.

Structural Demand Remains Intact

Beneath the headlines, the drivers of AI infrastructure demand remain robust. Global AI workloads continue to grow at a pace that far exceeds capacity expansion.³ Critically, the majority of data centre power consumption still comes from non-AI workloads, meaning AI-related usage is only set to increase as adoption broadens.

Graph: Global and US Data Centre Power Demand

Efficiency improvements, including cost reductions from newer models like DeepSeek, are contributing to better return on investment for AI projects. However, these efficiencies do not offset the scale of infrastructure and power investment required to meet future demand. AI-driven power consumption is growing faster than grid capacity, highlighting structural bottlenecks in energy and compute supply.⁴

Furthermore, this growth is extending beyond the technology sector. AI adoption is accelerating across consumer, financial, and industrial sectors, particularly in Asia and emerging markets. This expansion is creating long-term demand for infrastructure and enablers, as well as opportunities for companies embedding AI into their business models.

Semiconductors Remain Critical to AI’s Growth

The semiconductor cycle provides a clear example of how recent concerns around AI spending have been misunderstood. Short-term headlines around memory oversupply, DRAM pricing pressure, and near-term margin compression, including recent commentary from Micron, reflect cyclical digestion, not structural decline.⁵ AI-related semiconductor revenues are still expected to surpass US$200 billion this year, driven by demand for high-bandwidth memory, server processors, and interconnect capacity.⁶

Importantly, the outlook for the broader semiconductor market supports this long-term growth story. The global semiconductor market size is projected to grow from approximately US$600 billion in 2021 to over US$1.3 trillion by 2030, with the largest contribution coming from computing and data storage, the backbone of AI infrastructure.⁷ This category alone is expected to nearly double over the next five years, underpinned by rising compute intensity and storage requirements from AI workloads.

Beyond core computer chips, the semiconductor value chain is also expanding to support the next phase of AI adoption. Capital equipment spending remains elevated, with growing demand for advanced packaging and inspection tools essential to enable more complex AI models. Additionally, emerging trends such as physical AI, robotics, and AI-enabled edge devices are creating new growth avenues across the semiconductor ecosystem.

Global X is dedicated to helping you build wealth

Global X is dedicated to helping you build wealth

Loading...

What do you think of our new website?