AWS Cloud Capacity Sold Out Through 2027

Amazon Web Services (AWS) cloud infrastructure capacity is nearly sold out through the end of 2027, it has been revealed. Amazon CEO Andy Jassy officially acknowledged this situation and disclosed that some 2028 allocations have already been committed. The supply shortage in cloud computing driven by surging demand has been publicly confirmed for the first time through statements from the company's top executive.

Behind this situation lies the rapid expansion of cloud computing demand accompanying the widespread adoption of generative artificial intelligence. Companies are increasingly securing cloud capacity through long-term contracts to reserve computing resources for training and inference of AI models. As these "commitment-based" long-term contracts accumulate, available capacity has been depleted to near nothing.

AWS maintains a policy of reserving a certain amount of on-demand capacity (a mechanism for using resources only when needed without prior contracts) even in this situation. Specifically, the company states it is maintaining some availability for EC2 (virtual server service) and Bedrock (a service that enables access to generative AI models through APIs). However, whether users can actually benefit from this availability is another matter.

Following Jassy's remarks, developers and engineers actually using AWS report growing difficulty in acquiring standard EC2 instances on demand. Reports also indicate that similar supply constraints are occurring at other cloud providers, suggesting this may become an industry-wide issue.

This situation is significant in highlighting the limits of the "supply side" in AI infrastructure. Cloud computing's greatest strength has traditionally been its flexibility to "use what you need when you need it." However, if this premise is eroding, it could directly impact the speed and scale of AI development, particularly for users such as startups and small-to-medium enterprises that cannot pre-commit to massive capacity.

A key point to watch going forward is when and to what extent supply shortages can be resolved through new data center construction and semiconductor procurement advances. The fact that AWS is already making 2028 allocations suggests demand is accumulating even further into the future, indicating that near-term resolution will not be easy. This can be characterized as a period in which securing cloud capacity becomes directly linked to competitive advantage in AI adoption.