AI boom drove cloud infrastructure spending by 21% in Q1 2025
Major cloud providers are spending billions to build systems specifically to support the next phase of enterprise AI.
In 2025, the cloud isn’t just supporting artificial intelligence; it’s being reshaped by it. As enterprises push AI out of the lab and into real-world applications, demand for scalable, high-performance cloud infrastructure continues to surge. That demand is stretching the capacity of even the largest cloud providers and forcing a reengineering of the systems that power modern AI.
To keep up, cloud providers are increasing their infrastructure investments at an unprecedented scale, expanding data centers, deploying custom chips, and optimising systems for AI inference workloads.
In the first quarter of the year, global cloud infrastructure spending reached $90.9 billion, a 21% increase from Q1 2024, according to Canalys.
The investment boom is being driven by forces of demand for cloud computing. On the enterprise side, there’s growing urgency to embed AI into everyday workflows—from customer service to software development. We are already seeing generative AI models—especially large language models—now being part of core enterprise operations. This is pushing them to migrate more workloads to the cloud and invest in environments optimised for AI performance.
Another major factor is the rising cost and complexity of inference. As enterprises deploy AI at scale, the demand for sustained, high-volume inference is growing rapidly. Inference—the stage where trained models generate outputs—is computationally intensive and constant. Unlike training, which is a one-time process, inference runs continuously in production environments. This puts pressure on existing cloud infrastructure, exposing both cost inefficiencies and capacity limits.
As a result of these two major factors, enterprises are demanding more optimised cloud capabilities, and cloud providers are responding with large-scale infrastructure upgrades centered on custom silicon and workload-specific optimisation, consequently driving up cloud infrastructure spending.
Cloud hyperscalers like AWS, Microsoft Azure, and Google Cloud, which together command 65% of the global cloud market spending, are leading that charge, each pursuing strategies to reduce the cost of AI delivery while expanding capacity.
AWS maintained market leadership with a 32% share after growing 17% YoY, a slowdown attributed to supply-side constraints. To close the gap, AWS is expanding support for models like Claude 3.7 Sonnet and Meta’s Llama 4 on its Bedrock platform while promoting its in-house Trainium 2 chip, which offers a claimed 30–40% better price-performance than Nvidia-based solutions. It’s also investing $4 billion in a new Chilean region to extend its global infrastructure footprint.
Meanwhile, Microsoft Azure, with 23% market share, posted 33% growth, with AI alone adding 1 percentage point to its total. Its AI platform, Azure AI Foundry, processed over 100 trillion tokens this quarter. At the same time, Microsoft reported a 30% gain in throughput at constant power and a 50% drop in cost per token. These are direct responses to the rising burden of inference and cost.
Lastly, Google Cloud, holding 10% of the market, grew 31% year-on-year. In March, it rolled out the Gemini 2.5 model series, with Gemini 2.5 Pro performing strongly in reasoning and code tasks. It has driven a sharp increase in usage of Google AI Studio and the Gemini API, up over 200% since January. Still facing capacity limits, Google is investing $7 billion to expand its Iowa data center and has opened a new region in Sweden to support demand.
As AI adoption deepens, the cloud is no longer just a delivery channel. It’s becoming the operational backbone of enterprise AI, and the major providers are building systems specifically to support the next phase of enterprise AI—efficient, large-scale, and production-ready.
Content Partner
