This acceleration is all the more striking as it follows an already robust trend observed over recent years. After a slight 2% decline in 2023, capex for the nine major global cloud service providers surged by 58% in 2024 and 70% in 2025, and are now expected to grow by 79% in 2026. In other words, the pace of expansion continues to intensify despite an increasingly high comparison basis. In absolute terms, spending is projected to have risen from $172.6bn in 2023 to $272bn in 2024, then to $462.7bn in 2025, before reaching $830bn in 2026.

Following Q1 financial results, North American hyperscalers have emerged as the primary drivers of this revision. Microsoft has raised its investment spending forecasts to $190bn, representing a y-o-y increase of about 130%, while Google is now targeting between $180bn and $190bn. Meta has also lifted its guidance to $125bn-$145bn, while AWS is expected to exceed $230bn this year.

Beyond the headline figures, the composition of this spending indicates that AI is becoming a long-term infrastructure theme rather than a mere server cycle, according to TrendForce, a market research and consulting firm specializing in the technology sector. Investments are concentrated on GPU clusters, in-house developed ASIC chips, power supply systems, and data centers capable of supporting higher power density. TrendForce estimates that global installed data center capacity will reach approximately 155 GW in 2026, a 29% increase.

In this context, the favorable cycle no longer concerns only advanced semiconductors but also extends to electrical equipment, high-voltage direct current solutions, and liquid cooling, for which demand could continue to grow with the next generations of AI servers expected in 2027 and 2028.