Global cloud infrastructure spending has entered an era of unprecedented acceleration. Recent research from Canalys has revealed that spending reached an astonishing $90.9 billion in the first quarter of 2025—a 21% increase over the previous year. This surge isn’t just a reflection of broader digital transformation trends; rather, it’s being powered by rapid, widespread adoption of artificial intelligence (AI) technologies across the public and private sectors. As enterprises increasingly shift from AI research toward at-scale deployment, the commercial dynamics, provider competition, and technical strategies in the cloud market are evolving rapidly.
At the heart of the current spending wave is the dual push from both enterprise users and cloud providers. The Canalys report, widely cited in industry coverage, says “large-scale investment in both cloud and AI infrastructure remains a defining theme of the market in 2025.” In concrete terms, this means that nearly every major hyperscaler—Amazon Web Services (AWS), Microsoft Azure, and Google Cloud—is urgently rethinking its infrastructure to handle the demands of modern AI deployments.
AI’s commercial trajectory is changing the economics of cloud computing. Training a massive model (such as those underpinning generative AI services) is a capital-intensive, but one-time, undertaking. However, the process of “inference”—deploying these models to actually generate insights, text, or predictions in real time for businesses—incurs ongoing operational costs. As Rachel Brindley, Senior Director at Canalys, explains: “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.”
With model inference becoming a key bottleneck, both in performance and cost, hyperscalers are pouring billions into optimizing their offerings. The outcome isn’t just new chips or cloud instances, but a shift in market leadership and provider narratives.
However, the “winners’ circle” now looks different than it did just a few years ago:
For enterprise leaders, the takeaways are clear: cloud spending will continue to grow, but the calculus around AI will center increasingly on cost efficiency, model availability, and the proven performance of each provider’s bespoke hardware. The path to AI commercialization passes through the cloud, but the market’s topography is in flux. Those making infrastructure decisions today must balance vision with vigilance—scrutinizing performance claims, spreading risk through multi-cloud strategies, and demanding both transparency and innovation from their vendors.
What’s beyond doubt is that as AI advances from research to reality, the battle for cloud supremacy is both fiercer and more consequential than ever before.
Source: IT Pro Global cloud spending might be booming, but AWS is trailing Microsoft and Google
At the heart of the current spending wave is the dual push from both enterprise users and cloud providers. The Canalys report, widely cited in industry coverage, says “large-scale investment in both cloud and AI infrastructure remains a defining theme of the market in 2025.” In concrete terms, this means that nearly every major hyperscaler—Amazon Web Services (AWS), Microsoft Azure, and Google Cloud—is urgently rethinking its infrastructure to handle the demands of modern AI deployments.AI’s commercial trajectory is changing the economics of cloud computing. Training a massive model (such as those underpinning generative AI services) is a capital-intensive, but one-time, undertaking. However, the process of “inference”—deploying these models to actually generate insights, text, or predictions in real time for businesses—incurs ongoing operational costs. As Rachel Brindley, Senior Director at Canalys, explains: “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.”
With model inference becoming a key bottleneck, both in performance and cost, hyperscalers are pouring billions into optimizing their offerings. The outcome isn’t just new chips or cloud instances, but a shift in market leadership and provider narratives.
Ranking the Top Three: Microsoft, Google Surge as AWS Stumbles
Despite the overall growth, Canalys’ analysis highlights a subtle but important market shift: while AWS maintains the largest single market share, Microsoft Azure and Google Cloud are recording far steeper growth rates. The “big three” still command about 65% of the global cloud market together, and all posted solid year-on-year spending increases—collectively up 24%.However, the “winners’ circle” now looks different than it did just a few years ago:
- Microsoft Azure saw its cloud business grow by a remarkable 33%. Azure’s AI-related services in particular were credited with adding a 16-point growth rate lift—the largest single-quarter gain since mid-2024. Notably, the introduction of the GPT-4.1 model to Azure AI Foundry and its integration with GitHub have fueled a wave of developer activity. Azure AI Foundry reportedly serves developers inside over 70,000 enterprises, a milestone Microsoft is keen to promote.
- Google Cloud recorded growth over 30%, despite minor fluctuations in quarter-to-quarter revenue backlog. With a reported revenue backlog of $92.4 billion in Q1 (down slightly from the previous quarter), Google’s challenges mostly stem from supply constraints in available compute capacity. Yet, in the AI field, Google has made significant headway, particularly with the launch of its Gemini 2.5 model range. The Gemini 2.5 Pro model has received positive reviews for its real-world benchmark performance.
The AI Inference Battle: Chips, Cost, and Custom Hardware
Driving much of the infrastructure arms race is a renewed focus on AI inference efficiency. Here, each hyperscaler is racing to offer more cost-effective, purpose-built solutions:- Microsoft and Google have both deepened investments in AI-optimized hardware. For instance, Google’s innovation in in-house chips came into focus with the release of its seventh-generation Tensor Processing Unit (TPU), dubbed “Ironwood.” TPUs are specifically designed to speed up AI operations, especially inference, and are featured prominently in Google’s generative AI offerings. Google claims this innovation is central to making inference affordable and scalable.
- AWS has doubled down on its Trainium chips, which are engineered for both AI training and inference. The company has announced aggressive price cuts to promote Trainium chip adoption, seeking to offer a compelling alternative to the more expensive Nvidia GPUs that dominate the current market. Amazon regularly cites benchmark data showing Trainium 2 with a 30-40% price-performance advantage over comparable Nvidia products. However, such internal benchmarking claims require careful scrutiny. Independent, third-party evaluation of Trainium’s real-world performance across workloads remains limited, so prospective enterprise customers are wise to run their own tests or seek cross-industry confirmation.
Service Differentiation: Models, Backlog, and Developer Ecosystem
Beyond hardware, cloud leaders are competing on the breadth and accessibility of AI models and the strength of their developer ecosystems.- Google Cloud’s AI Services have been bolstered by the addition of new models. Gemini 2.5 and Gemini 2.5 Pro provide improved performance on a variety of benchmarks, positioning Google alongside OpenAI-based platforms on Azure. Additionally, Google’s backlog—the amount of cloud revenue already committed but not yet recognized—clocks in at over $92 billion, which illustrates the scale of ongoing large deals with enterprise customers.
- Microsoft Azure’s Model Access has expanded notably. The availability of GPT-4.1 in both Azure AI Foundry and GitHub demonstrates Microsoft’s strategy to appeal not just to IT decision makers, but also to the burgeoning developer community tasked with integrating AI into production applications.
- AWS Bedrock, Amazon’s flagship for AI model hosting, has grown its portfolio to include both third-party and proprietary models. During the past quarter, it became the first provider to offer DeepSeek R1 and Mistral’s Mixtral Large, in addition to adding the popular Anthropic Claude 3.7 Sonnet and Meta’s Llama 4 models. This “open model marketplace” strategy is clearly designed to keep avid developers engaged and to position AWS as an innovator—even as its infrastructure growth lags rivals.
Critical Analysis: Strengths, Strategic Risks, and Industry Implications
Strengths and Market Drivers
- Record Cloud Spending: The $90.9 billion Q1 figure and 21% annual growth mark a robust industry trajectory. Cloud remains the backbone of digital transformation and AI adoption worldwide.
- AI-Driven Infrastructure: Providers’ race to optimize for inference, not just training, acknowledges the real-world economics of operationalizing AI at scale.
- Custom Hardware: The willingness of AWS, Microsoft, and Google to design and mass-produce their own chips indicates both deep pockets and a realization that general-purpose hardware may no longer suffice for AI-centric cloud workloads.
- Model Breadth: Expanding model libraries and opening platforms to competitive, third-party AI options foster innovation and customer choice. This is critical as “winner-take-all” narratives in cloud and AI have softened, due partly to open-source momentum and shifting enterprise priorities.
Strategic Risks and Weaknesses
- Supply Constraints: Both Google and AWS cite supply-side issues, particularly in access to top-end GPUs and specialized compute hardware. As demand for inference-capable infrastructure outpaces supply, cloud providers face tough decisions on capacity allocation, pricing, and partnership strategy. Clients with time-sensitive AI rollouts may feel the friction.
- AWS Momentum Faltering: While still dominant in absolute terms, AWS’s slowing growth is cause for concern. The fact that its AI business, while growing quickly, is “still in the early stages of development”—per Canalys—raises the risk that AWS could cede leadership to Microsoft or Google if its infrastructure investments can’t catch up. With AWS’s historic perception as an innovation leader, this narrative reversal (whether temporary or lasting) could shake customer confidence and open the door to greater multi-cloud or migration activity.
- Unverified Performance Claims: Claims such as Trainium 2’s “30-40% price-performance advantage” should be approached with caution. Vendors naturally highlight the benchmarks most favorable to their products, but the diversity of enterprise workloads and the fast-evolving nature of AI models means that real-world returns will vary. Prospective customers are well-advised to demand third-party performance validation and pilot new hardware before wide deployment.
- Revenue Backlog Volatility: Google’s minor dip in revenue backlog—attributed to compute supply shortages—suggests that not all demand can be converted to revenue right away. If supply chain or manufacturing woes continue, providers risk billing delays, competitive customer switching, or even reputational impact.
The Outlook for the Remainder of 2025
As organizations of all sizes accelerate AI deployments, the cloud infrastructure market shows little sign of cooling. If anything, the shift from research prototypes to at-scale production models is likely to intensify demands around efficiency, flexibility, and cost predictability. Key trends to watch include:- The Rise of Multi-Cloud Strategies: As enterprises become more sophisticated in comparing not just pricing but operational AI costs, many are opting for multi-cloud or hybrid deployment models. This reduces the risk of lock-in, exploits the unique strengths of each provider, and can help spread supply risk.
- Accelerated Hardware Innovation: The trend toward proprietary accelerator chips—TPUs, Trainiums, and as-yet-unannounced contenders—will only ramp up. Custom silicon is now a table-stakes play in the race for AI leadership.
- Open Models and Marketplaces: Expect a continuing broadening of available models and platform openness. With open-source AI growing and new entrants like Meta and Mistral gaining traction, the era of single-provider AI dominance is ending. Customers are demanding choice, and providers will continue to differentiate by integrating the latest and most reliable models quickly.
Conclusion: A Market at a Crossroads
The first quarter of 2025 cements several realities. Cloud infrastructure is not just growing; it’s being fundamentally reshaped by artificial intelligence and the demands of its operationalization at enterprise scale. While Microsoft and Google are registering higher year-on-year growth rates and expanding developer traction, AWS faces strategic headwinds but remains a formidable competitor thanks to its breadth and ongoing innovation in AI hardware.For enterprise leaders, the takeaways are clear: cloud spending will continue to grow, but the calculus around AI will center increasingly on cost efficiency, model availability, and the proven performance of each provider’s bespoke hardware. The path to AI commercialization passes through the cloud, but the market’s topography is in flux. Those making infrastructure decisions today must balance vision with vigilance—scrutinizing performance claims, spreading risk through multi-cloud strategies, and demanding both transparency and innovation from their vendors.
What’s beyond doubt is that as AI advances from research to reality, the battle for cloud supremacy is both fiercer and more consequential than ever before.
Source: IT Pro Global cloud spending might be booming, but AWS is trailing Microsoft and Google