AWS

AWS has confirmed it cannot restore resources or data hosted exclusively in the mec1-az2 zone in the UAE or exclusively in the Bahrain region following damage from conflict with Iran. The company states that the Bahrain incident spanned multiple availability zones and exceeded the resilience thresholds these regional and multi-AZ services are designed to withstand. This leaves customers with data trapped in the affected infrastructure unable to recover it through standard AWS restoration mechanisms.

• Data in mec1-az2 (UAE) is unrecoverable if not replicated elsewhere.
• Bahrain region damage exceeded multi-AZ design limits.
• Cross-region replication is critical for conflict-zone resilience.
• Single-AZ deployments in affected areas are total losses.
• Verify backup strategies for Middle East infrastructure immediately.

Alibaba Cloud has announced a six-year expansion plan to reach 20GW of datacenter capacity, betting that the AI demand surge will remain robust. To support this scale, the company is unveiling a new custom chip designed to power these facilities. This strategy contrasts with the current 37GW of capacity already under construction in the US alone.

• Alibaba targets 20GW total datacenter capacity over a six-year horizon
• New proprietary chip revealed to drive power efficiency for large-scale AI workloads
• Management remains bullish on AI boom longevity despite massive US construction pipeline
• Expansion signals continued capital intensity in global cloud infrastructure race

Recent incidents involving war damage, air traffic disruptions, and storage shortages highlight the fragility of cloud infrastructure when treated as an external concern. These events expose the hidden costs of relying on third-party resilience without adequate internal verification. The article argues that assuming cloud providers handle all failure scenarios is a dangerous oversight for engineering teams.

• War and physical chaos can breach cloud resilience assumptions instantly.
• Storage shortages reveal gaps in supply chain and capacity planning.
• Treating resilience as someone else's problem invites catastrophic failure.
• Engineering teams must verify, not just assume, provider guarantees.