Cloud Capacity Saturation, Supply Constraints, and Record Capex: AWS Projections
Amazon Web Services (AWS), the world's largest cloud services provider, has its data center capacity virtually fully booked through 2027. During an interview on the Andreessen Horowitz (a16z) podcast, Amazon CEO Andy Jassy stated that the company faces insatiable demand across both traditional and artificial intelligence workloads, which has led to pre-selling even a substantial portion of the capacity planned for 2028.
Unprecedented Capital Expenditure and Component Bottlenecks
Despite executing a record capital expenditure budget, the pace of physical expansion cannot match market demand:
- $220 Billion Capex: Amazon will spend $220 billion this year on cloud and AI infrastructure. Jassy stated that the company could allocate significantly higher figures without it representing a misuse of capital, and even then, it would not satisfy all existing demand.
- Availability by Year: Capacity for 2026 is 100% sold out, while 2027 is largely sold out, and a considerable portion of 2028 has already been pre-committed. This excludes only quotas reserved for On Demand services, EC2, and the Bedrock inference platform.
- Structural Supply Chain Constraints: Data center scaling faces severe bottlenecks in hardware components, particularly shortages of memory components. Jassy warned that these supply limitations will not be restricted to 2027, but will persist across various components for several years.
Strategic Prioritization Dilemmas Between AI and Traditional Cloud
The simultaneous growth of conventional cloud and artificial intelligence architectures forces AWS to define operational priorities:
- Commitment to Leadership: Jassy maintained that AWS is deploying more active capacity than any other competitor to preserve its leadership position in cloud and AI infrastructure.
- Vertical Trade-offs: Since growth is accelerating across all fronts, AWS management has had to make complex allocation decisions (trade-offs) to determine which verticals and customers to prioritize when delivering computing power.
At next+ we believe that Andy Jassy's statements confirm that the digital economy's bottleneck has definitively shifted from software availability to the physical limits of data centers and the hardware supply chain. The fact that the global cloud leader has its capacity booked through 2027 despite investing $220 billion in a single year demonstrates that the deployment of generative AI is consuming computing resources at a rate exceeding global manufacturing capacity. For Chief Technology Officers (CTOs), Chief Financial Officers (CFOs), and digital transformation leaders in Latin America, this landscape demands securing multi-year computing capacity contracts well in advance and planning contingencies in the face of persistent hardware shortages and rising cloud infrastructure costs.