In 2019, enterprise cloud adoption was a competitive advantage. In 2025, it is table stakes. The question is no longer whether companies should move to the cloud — it is how much of their infrastructure has already migrated, what remains on-premise and whether the bills are under control.
According to Gartner, worldwide end-user spending on public cloud services is projected to reach $723 billion in 2025, up from roughly $490 billion in 2022. That is a 47% increase in three years. The three dominant players — Amazon Web Services (AWS), Microsoft Azure and Google Cloud — control roughly two-thirds of the market, though their relative positions keep shifting.
AWS remains the market leader by revenue, generating $27.5 billion in the third quarter of 2024 alone, with a 19% year-over-year growth rate. But Microsoft Azure has been gaining ground, particularly in enterprise accounts that already run on Office 365 and Windows Server. Azure’s revenue grew 33% in the same period, driven heavily by AI workloads that leverage Azure’s existing relationships with OpenAI and other model providers. Google Cloud, though smaller at $11.4 billion in Q3 2024 revenue, has been growing faster than both — posting 35% growth, in large part because of its strength in data analytics and machine learning infrastructure.
The Real Cost Debate: Cloud vs Repatriation
One of the most intense debates in enterprise IT right now is whether the cloud actually saves money. The 2023 repatriation movement — companies moving workloads back from the cloud to on-premise data centres — was fuelled by some very loud examples. 37signals (the company behind Basecamp and Hey) famously published a detailed accounting of how they saved $1.5 million per year by leaving the cloud. DHH’s blog posts on the topic went viral.
But those examples, while real, are outliers. Most companies do not run a single-product SaaS business with predictable traffic patterns. A 2024 survey by Flexera found that 72% of enterprises have a “cloud-first” strategy, but only 32% said they are actually achieving their cost savings targets. The gap between expectations and reality is driving a more nuanced approach: not cloud repatriation, but cloud optimisation.
FinOps — the practice of bringing financial accountability to cloud spending — has become one of the fastest-growing job categories in tech. The FinOps Foundation’s 2024 survey found that organisations practising mature FinOps reduce cloud waste by an average of 25%. The biggest wins come not from moving off the cloud, but from right-sizing instances, buying reserved capacity and shutting down idle resources — none of which require leaving the public cloud.
AI as the New Cloud Accelerant
If there is one force pushing cloud adoption faster than anything else, it is artificial intelligence. Training and running large language models requires enormous compute resources that most organisations simply cannot build in-house. GPT-4, for context, was trained on an estimated 25,000 NVIDIA A100 GPUs running for 90 to 100 days — infrastructure that would cost hundreds of millions of dollars to replicate on-premise.
Microsoft’s Azure has been the primary beneficiary of the AI boom, as OpenAI’s exclusive cloud provider. But AWS is fighting back with its own AI chips (Trainium and Inferentia), and Google Cloud’s TPU v5p pods are specifically designed for large-scale training workloads. The AI infrastructure market, estimated at $45 billion in 2024, is projected to reach $105 billion by 2028 according to Bloomberg Intelligence.
Multi-Cloud Is Now the Default
Gone are the days when companies picked one cloud provider and stuck with it. The 2024 Flexera State of the Cloud report found that 89% of enterprises use a multi-cloud strategy, and 73% use a hybrid cloud approach that mixes public cloud with on-premise infrastructure. The reasons are practical: avoiding vendor lock-in, optimising costs by workload type, meeting regulatory requirements for data sovereignty and ensuring redundancy.
Kubernetes, originally created by Google and now maintained by the Cloud Native Computing Foundation, has become the de facto standard for multi-cloud orchestration. According to the CNCF’s 2024 annual survey, 84% of organisations use Kubernetes in production, up from 78% in 2022. The rise of managed Kubernetes services — Amazon EKS, Azure AKS, Google GKE — has lowered the barrier significantly, though running Kubernetes at scale still requires significant operational expertise.
What Comes Next
The cloud market in 2025 is moving toward maturity, but it is not done evolving. Several trends are worth watching: sovereign cloud (data centres operated within specific national borders to comply with regulations like GDPR and the EU’s planned AI Act), edge computing (processing data closer to where it is generated, reducing latency for applications like autonomous vehicles and industrial IoT) and the continued blurring of the line between cloud and on-premise through services like AWS Outposts and Azure Arc.
For enterprise IT leaders, the message is clear: the cloud is no longer a destination you arrive at. It is an operating model you refine continuously. The winners will be those who master FinOps, embrace multi-cloud pragmatically and position their infrastructure to absorb whatever AI workloads demand next.