For most of the last decade, the standard move for any company modernizing its IT was to send everything to the cloud. AWS, Microsoft Azure, Google Cloud: pick one, migrate fast, and save money. That strategy is now fraying at the edges, and AI is the reason.
The 2026 ISG Provider Lens report on private and hybrid cloud services finds that enterprises are shifting from full cloud migration strategies toward mixed setups that balance on-site infrastructure with cloud capacity. The shift is not ideological. It is financial and regulatory.
Running AI at scale is expensive in ways that traditional cloud pricing models were not built for. Industry analysis shows that enterprise AI implementations routinely cost two to four times the advertised subscription price once integration, infrastructure scaling, and day-to-day management are counted. AI-driven workloads have pushed wasted cloud spending to 29% of total cloud budgets in 2026. Only 39% of organizations have clear, unified visibility over what they are spending across their cloud environments.
There is also the problem of being too tied to one provider. Many AI tools are built around specific cloud infrastructure, which makes moving to a different provider painful and expensive. Gartner estimates that data fees for moving data out of a cloud provider consume 10 to 15% of a typical enterprise cloud bill. The more AI you run, the harder and costlier it becomes to leave.
The sovereignty pressure is separate from cost, but equally real. The EU's Cloud and AI Development Act, proposed in June 2026, introduces a formal four-tier framework classifying cloud services by where their infrastructure sits, who owns and controls the company, and whether data can be compelled by foreign governments. This matters because storing data in an EU-based AWS or Azure data center does not fully protect it from US legal requests under the CLOUD Act. Regulators are now codifying the difference between where data physically sits and who actually controls it.
Gartner projects European spending on sovereign cloud infrastructure will nearly double between 2025 and 2026, reaching $12.6 billion, and approach $23 billion by 2027. For companies operating in financial services, healthcare, insurance, or any sector touching public-sector contracts, the question of which cloud provider you use is increasingly a compliance question, not just a procurement one.
A new class of cloud providers is responding to all of this. Gartner calls them neoclouds: companies built specifically to run AI workloads, often with a stronger focus on data sovereignty and more competitive pricing than the large established providers. Gartner expects these newer providers to capture 20% of a $267 billion AI cloud market by 2030. That is a meaningful chunk, and it signals that the big three cloud providers no longer have the AI infrastructure conversation to themselves.
What does this mean in practice? Companies that committed deeply to a single cloud provider for AI are now reviewing whether that commitment still makes financial and legal sense. The answer, increasingly, is that a split approach works better: run predictable, sensitive, or regulated workloads on infrastructure you control or that sits in a clearly defined legal jurisdiction, and use cloud for the flexible, bursty work that needs to scale fast.
This is not about abandoning cloud. It is about treating cloud as one tool in a broader setup rather than the only answer. The companies that locked in quickly, without negotiating data portability or exit clauses, are the ones facing the hardest conversations now. The lesson for anyone still building out their AI infrastructure: your cloud contract terms matter as much as your cloud choice.