Airbus is exiting AWS according to a recent Hacker News thread that drew 85 points and 55 comments.
The move centers on data sovereignty and long-term infrastructure control rather than short-term cost savings.
What the Migration Involves
Airbus is shifting workloads off AWS to a mix of on-premise systems and European cloud providers. The decision prioritizes compliance with EU data rules over public-cloud convenience.
HN users note the airline manufacturer already runs significant internal compute for simulation and design tools.
Numbers Behind the Discussion
The thread reports no exact migration timeline or dollar figures. Commenters reference typical enterprise exits that take 18-36 months and require 20-40% additional staff for private infrastructure.
Early reactions flag potential 15-25% higher upfront capex before any operating-cost reduction appears.
Pros and Cons of Leaving AWS
- Pros: Full control over data residency; avoidance of vendor lock-in pricing; direct hardware tuning for specialized workloads.
- Cons: Loss of managed AI services such as SageMaker; higher operational burden; slower access to new GPU instance types.
Cloud Alternatives for AI Workloads
Enterprises weighing similar exits typically compare three paths.
| Provider | AI GPU Access | EU Data Residency | Migration Effort |
|---|---|---|---|
| Microsoft Azure | ND-series H100 clusters | Yes | Medium |
| Google Cloud | A3 instances | Yes | Medium |
| On-prem + OVH | Self-managed clusters | Full | High |
Azure and Google currently lead in managed LLM training tools, while on-prem setups demand dedicated MLOps teams.
Who Should Consider an Exit
Large manufacturers with strict regulatory requirements benefit most. Startups and research groups running under 100 GPUs should stay on hyperscalers for speed of iteration.
Teams already maintaining private clusters gain the clearest path.
Bottom Line / Verdict
Airbus's departure highlights that sovereignty and control now outweigh raw convenience for some regulated industries running AI workloads.
Bottom line: Companies with EU data mandates and existing hardware teams can replicate the move; others should benchmark total cost of ownership before following.
Top comments (0)