Global cloud spending might be booming, but AWS is trailing Microsoft and Google
Global cloud infrastructure spending is booming, according to new research from Canalys, reaching $90.9 billion in the first quarter of 2025.
The spending surge marks a 21% year-on-year increase, the consultancy found, with widespread AI adoption programs playing a key role in recent growth.
“Large-scale investment in both cloud and AI infrastructure remains a defining theme of the market in 2025,” Canalys said.
Notably, this growth is being driven by both the enterprise end-user and provider sides, with leading hyperscalers now “intensifying efforts to optimize infrastructure” to accommodate AI adoption.
“As AI transitions from research to large-scale deployment, enterprises are increasingly focused on the cost-efficiency of inference, comparing models, cloud platforms, and hardware architectures such as GPUs versus custom accelerators,” said Rachel Brindley, Senior Director at Canalys.
“Unlike training, which is a one-time investment, inference represents a recurring operational cost, making it a critical constraint on the path to AI commercialization.”
AI inference costs have been a key talking point in recent months, with Google Cloud in particular focusing heavily on the challenges faced by customers at its annual conference earlier this year.
To tackle these challenges, Canalys noted that hyperscalers are now “deepening their investments in AI-optimized infrastructure”.
Providers such as AWS, Microsoft, and the aforementioned Google Cloud have all introduced in-house chips and “purpose-built instance families”. The aim here is to drive inference efficiency and thereby reduce the overall cost of AI.
Google told ITPro in April its investment in TPUs, or ‘tensor processing units’, are crucial to help it meet demand for AI inference. The company unveiled the launch of its seventh generation TPU, dubbed ‘Ironwood’, during the Google Cloud Next conference.
Microsoft and Google record solid growth
Canalys’ research shows the ranking for AWS, Microsoft Azure, and Google Cloud remained unchanged in the previous quarter. The trio’s combined market share also accounts for 65% of global cloud spending.
“Collectively, the three hyperscalers recorded a 24% year-on-year increase in cloud-related spending,” the consultancy said.
Broadly speaking, this paints a positive picture for the leading providers. Google Cloud and Microsoft both recorded growth rates of over 30%, for example.
The latter recorded significant growth of 33% alongside a 16-point growth rate lift to Azure from AI. The consultancy said this marked the largest single-quarter uplift since Q2 2024.
“In April, Azure announced the availability of the GPT-4.1 model series on both Azure AI Foundry and GitHub, further broadening developer access to advanced AI capabilities across its ecosystem,” Canalys said.
“Azure AI Foundry, Microsoft’s platform for building and managing AI applications and agents, is now used by developers at more than 70,000 enterprises.”
Google Cloud, meanwhile, recorded a revenue backlog of $92.4 billion. While this marked a slight dip compared to the previous quarter, Canalys attributed this to “supply constraints – particularly with regard to compute capacity.
The company made some positive announcements across the quarter, especially in the AI space with the launch of its Gemini 2.5 model range.
Gemini 2.5 Pro in particular received positive feedback from customers and showed solid benchmark performance, Canalys said.
AWS troubles
There are warning signs for AWS; analysts note that growth momentum “diverged” during this period.
Amazon’s cloud division recorded growth of 17%, which marked a deceleration compared to the final quarter of 2024 in which the hyperscaler notched 19%.
Canalys once again attributed this deceleration to “supply-side constraints”, which stunted the firm’s ability to meet rising AI-related infrastructure demands.
Despite sluggish growth, AWS’ AI business is still growing at a triple-digit annual rate, but the consultancy said this “remains in the early stages of development”.
Earlier this year, the company announced a new price-cutting strategy to promote its Trainium chips and drive adoption rates. The hyperscaler has been keen to tout its in-house chips as a viable alternative to costly Nvidia options, regularly citing Trainium 2’s 30-40% price-performance advantages.
Bedrock, AWS’ flagship AI model service, which gives users access to a range of third-party and in-house models, also saw some new additions.
Anthropic’s Claude 3.7 Sonnet and Meta’s Llama 4 models were added to the service over the last quarter, and AWS also became the first cloud provider to offer DeepSeek R1 and Mistral’s Mixtral Large options.
MORE FROM ITPRO
Source link
