Canalys: Global cloud infrastructure service spending reached US$90.9 billion in the first quarter, up 21% year over year

The Zhitong Finance App learned that according to Canalys's latest data, global cloud infrastructure service spending reached US$90.9 billion in the first quarter of 2025, an increase of 21% over the previous year. Massive investment in cloud and AI infrastructure has become one of the key topics for market development in 2025. At the same time, in order to accelerate large-scale adoption of AI by enterprises, leading cloud vendors are speeding up infrastructure optimization, especially through self-developed chips, with the aim of reducing AI usage costs and improving inference efficiency. In the first quarter of 2025, the market rankings of the three major cloud vendors (Amazon Web Services, Microsoft Azure, and Google Cloud) remained unchanged, accounting for 65% of global cloud spending. The combined cloud-related spending of the three leading cloud vendors increased 24% year over year.

In the first quarter of 2025, the growth trend of leading cloud vendors diverged. Both Microsoft Azure and Google Cloud achieved year-on-year growth of more than 30% (although Google Cloud's growth rate was slightly slower than in the previous quarter), while Amazon Web Services (AWS) grew 17%, slowing down from 19% in the fourth quarter of 2024. This deceleration was mainly affected by supply-side restrictions, which limited its ability to meet rapidly growing AI-related demand. In response, leading cloud vendors continue to increase their investment in AI infrastructure, expand computing power capacity, and grow their layout over the long term.

Overall, in the first quarter of 2025, the global cloud services market maintained steady growth, focusing on two major strategic priorities: one is to accelerate cloud migration, including moving more workloads to the cloud or restarting the transformation of local systems that were originally shelved; the other is to actively explore the application of generative AI. The rise of generative AI is highly dependent on cloud infrastructure, which in turn has further strengthened enterprises' cloud strategies and accelerated the migration process.

Rachel Brindley, senior director of Canalys, said, “As AI moves from the research stage to the deployment stage, companies are paying more and more attention to the cost effectiveness of the inference stage and are beginning to compare and evaluate models, cloud platforms, and hardware architectures such as GPUs and custom accelerators. Unlike training, which is a one-time investment, reasoning is an ongoing operating cost, and has become a key constraint on the path to commercializing AI.”

Canalys analyst Zhang Yi added, “Currently, many AI services use a pay-per-use model — usually billed per token or number of API calls — and cost forecasting becomes more difficult as usage expands. When the cost of inference fluctuates greatly or is too high, companies often have to limit use, reduce model complexity, or only deploy in high-value scenarios, so the broader potential of AI has not been fully unleashed.”

To meet these challenges, leading cloud vendors are increasing their investment in AI-optimized infrastructure. Leading cloud vendors, including AWS, Microsoft Azure, and Google Cloud, have launched self-developed chips such as Trainium and TPU and dedicated instance series to improve inference efficiency and reduce overall AI usage costs.

Amazon Web Services (AWS) continued to maintain its leading position in the first quarter of 2025, accounting for 32% of the global market share, and revenue growth of 17% year over year. Its AI business is still in its early stages of development, but it is maintaining a three-digit annual growth rate. In March 2025, AWS launched a price reduction strategy to push customers to use its self-developed Trainium AI chip to replace Nvidia, which is more expensive, and emphasizes that Trainium 2 has a 30-40% advantage in terms of price-performance ratio. At the same time, AWS accelerated the expansion of its Bedrock service, added Anthropic's Claude 3.7 Sonnet and Meta's Llama 4 models, and became the first cloud vendor to fully host DeepSeek R1 and Mistral's Mixtral Large models. Further demonstrating its long-term commitment to global infrastructure, AWS announced an investment of over $4 billion in May 2025, with plans to establish a new cloud region in Chile by the end of 2026.

Microsoft Azure continued to be the world's second-largest cloud service provider in the first quarter of 2025, with a market share of 23% and a year-on-year increase of 33%. Microsoft reports that AI brought Azure a 16 percent increase in growth, the biggest increase in a single quarter since the second quarter of 2024. In April 2025, Azure announced that the GPT-4.1 model series was officially launched on Azure AI Foundry and GitHub, further expanding the channels for developers to obtain advanced AI capabilities in their ecosystem. Azure AI Foundry is Microsoft's platform for building and managing AI applications and agents, and is currently used by developers in more than 70,000 enterprises. This quarter, the number of tokens handled by the platform exceeded 100 trillion, a fivefold increase over the previous year. Microsoft also continues to work to reduce the cost of AI applications. The report shows that its AI performance has increased by nearly 30% under the same power consumption, while reducing the cost per token by more than 50%. As part of a global infrastructure expansion plan, Microsoft built new data centers in 10 countries on four continents within the first quarter.

Google Cloud maintained a 10% market share in the first quarter of 2025 and achieved 31% year-on-year growth. As of March 31, its total revenue backlog reached US$92.4 billion, a slight decrease from the previous quarter. The decline was mainly due to supply restrictions, particularly a shortage of computing power capacity, which limited Google Cloud's ability to meet the full needs of customers. In March 2025, Google launched the Gemini 2.5 model series. Among them, Gemini 2.5 Pro was widely recognized for its leading benchmark performance and rankings in the Chatbot Arena. The model has significantly improved both reasoning and programming capabilities, bringing more application possibilities to developers and enterprise users. Since the beginning of the year, active usage of Google's AI Studio and Gemini API has surged by more than 200%, reflecting the strong enthusiasm of developers to adopt and the strong market demand for generative AI solutions. Google also opened its 42nd global cloud region in Sweden and announced that it will invest $7 billion to expand its data center in Iowa to further support its growing AI and cloud workloads.