AWS is making AI infrastructure more expensive, and the timing is not subtle.
Amazon Web Services raised prices on its EC2 Capacity Blocks for ML by roughly 20%, effective July. Capacity Blocks let customers reserve GPU clusters for fixed windows - a model designed for bursty training workloads rather than always-on inference. The increase applies to those reserved slots, meaning teams budgeting for upcoming model runs will pay materially more than they planned. The move was first reported by Business Insider.
The hike is a signal, not just a line item. GPU memory constraints have been a quiet bottleneck across the industry for months, and cloud providers are now passing that pressure to customers. When the largest cloud vendor raises reserved-compute prices 20% in one move, it recalibrates what AI development actually costs - and it hits startups and research teams with fixed budgets harder than hyperscalers with negotiated contracts.
AWS is not alone in facing tight GPU supply, but it is the first major provider to make the crunch this visible in its public pricing. Rivals like Google Cloud and Azure have their own waitlists and capacity constraints; they have just been quieter about it. Expect the others to watch how customers absorb this before deciding whether to follow.