Thanks to its ability to support compute-intensive workloads, the cloud is becoming central to artificial intelligence and its related applications. According to Capgemini experts, this shift is already apparent in public cloud spending, which “is expected to almost double from $723 billion in 2025 to $1.47 trillion by 2029”. The report also indicates that classical architectures may prove insufficient for the sector’s demands.
Businesses have to fine-tune models with their own information, maintain control over their data and meet current regulations. This is driving them to adopt hybrid, private, multi-cloud and sovereign architectures so they can operate securely and in compliance. Particularly, AI agents “require scalable, distributed, low-latency infrastructure where cloud and edge operate as a single intelligent fabric” that responds quickly and efficiently.
Hybrid platforms will become the standard in 2026. Rather than operating as rigid tools, “AI workloads will flow seamlessly between edge, private cloud and public cloud environments.” In this sense, Cloud 3.0 marks the point at which the cloud evolves from an IT platform into a distributed execution layer that enables AI at scale.
A new operating model
This transition to Cloud 3.0 is reflected in three clear signals. First “is the emergence of AI-optimised hybrid architectures, where enterprises deliberately split workloads across private, public and edge environments.” AI is no longer tied to a single provider at scale; instead, it is deployed wherever it delivers value. Second, companies often run critical applications across multiple providers simultaneously to strengthen resilience and data sovereignty. Finally, highly regulated industries — including the public sector, healthcare and banking — are combining the power of major providers with strict controls over where information is stored and how it is managed.
Capgemini notes that as organisations move from experimenting with AI agents to putting them into production, they need “robust orchestration frameworks that coordinate multiple specialised agents.”