Amazon Web Services (AWS) has already committed its major cloud computing capacity for 2026, with most large allocations for 2027 also booked and substantial commitments extending into 2028, Amazon CEO Andy Jassy said on Friday, October 23, 2026.
Speaking on a video podcast hosted by venture capital firm Andreessen Horowitz (a16z), Jassy described demand from artificial intelligence laboratories and enterprise customers as growing faster than Amazon can expand its computing infrastructure.
The disclosure, reported by Reuters, highlights the scale of the AI infrastructure expansion underway at the world's largest public cloud provider. However, it does not mean AWS has stopped offering cloud services or that every customer must wait until 2028 for computing resources.
What Amazon Said About AWS Capacity Through 2028
During the conversation with a16z co-founder Ben Horowitz and general partner Erik Torenberg, Jassy outlined how far into the future customers have committed to AWS computing capacity.
According to his remarks, the major capacity allocations for 2026 are sold out, most of the corresponding capacity for 2027 is already committed, and a significant portion of 2028 capacity has also been reserved.
Crucially, Jassy distinguished those large commitments from resources AWS retains for customers accessing services on demand.
The distinction matters because AWS operates multiple services with different infrastructure needs and purchasing arrangements. Amazon Elastic Compute Cloud (EC2), for example, provides virtual computing resources used to run applications, while Amazon Bedrock enables developers and businesses to build applications using AI models.
AWS continues to reserve some capacity for on-demand use and services such as Bedrock and EC2, according to Jassy's explanation. The commitments therefore should not be interpreted as a platform-wide suspension of new customer access.
The figures also describe capacity commitments rather than independently verified utilization rates. Amazon has not disclosed in these remarks the precise percentage of total AWS computing resources reserved for each year, the number of GPUs involved, or how the commitments are distributed among customers and regions.
Why AI Demand Is Absorbing Cloud Capacity
Artificial intelligence has increased demand for computing infrastructure that can support both the development of increasingly capable models and their deployment into commercial products.
Training advanced AI models can require substantial amounts of specialized processing hardware operating across interconnected systems. Once models are deployed, inference—the process of generating outputs from trained models—creates additional computing demand as applications serve users.
These workloads have different technical characteristics, but both depend on available processing capacity, memory, networking, power, and cooling infrastructure.
AI research laboratories represent one important source of demand. Enterprises building AI-powered software, integrating models into business operations, or expanding existing cloud applications create another.
Jassy's remarks suggest that customers seeking large amounts of computing capacity are increasingly making commitments well ahead of deployment. Such arrangements can give cloud providers greater visibility into future demand while potentially limiting the amount of additional large-scale capacity available for new reservations.
However, the disclosure does not establish how much of AWS's total capacity is devoted to AI training, inference, or traditional enterprise workloads. It also does not demonstrate that every AWS region, instance type, or service is experiencing the same degree of demand pressure.
Why Amazon Cannot Expand Computing Capacity Overnight
Increasing cloud capacity requires far more than purchasing additional processors. Hyperscale infrastructure depends on data center buildings, electrical connections, cooling systems, servers, networking equipment, and the technical work needed to bring those components into operation.
Even after a data center building is constructed, its ability to serve customers depends on whether the necessary electrical, mechanical, and computing equipment has been installed, tested, and made operational.
Access to sufficient electricity and cooling is particularly important for dense AI computing installations. Advanced processors generate substantial heat, while large clusters require reliable high-speed connections between computing systems.
Amazon's investment plans demonstrate the financial scale of this expansion. The company previously outlined approximately $220 billion in capital expenditure for 2026, a figure that includes AI infrastructure and other business investments. That amount should not be confused with spending exclusively dedicated to AWS.
The company increased its 2026 spending plan as strong AWS growth reinforced demand for additional computing resources, according to Reuters' July 31 reporting on Amazon's cloud performance and investment outlook.
Jassy also indicated that Amazon retains some flexibility in infrastructure spending. As The Information reported on October 23, he explained that certain server and networking equipment purchases can be adjusted relatively close to when the hardware is needed.
That flexibility may help Amazon respond if expected customer demand changes. It does not eliminate the longer planning and construction requirements associated with expanding physical data center infrastructure.
What the AWS Capacity Commitments Mean for Customers
For businesses planning substantial AI training deployments or large infrastructure expansions, the commitments highlight the importance of understanding future resource availability and planning requirements.
Organizations that need extensive computing clusters may face different procurement considerations from customers running smaller applications or purchasing ordinary cloud resources on demand.
There are three important distinctions:
- Reserved capacity is not necessarily consumed capacity. A future customer commitment does not establish that the corresponding infrastructure is already operating at full utilization.
- Large infrastructure allocations differ from routine cloud access. The availability of substantial dedicated computing resources does not determine whether smaller EC2 deployments or other AWS services remain accessible.
- Future commitments do not guarantee future revenue or profitability. Actual financial results depend on contract terms, deployment, customer usage, costs, and other factors.
Jassy's statements do not establish that AWS has introduced new access restrictions, raised prices because of these particular commitments, or stopped accepting new customers.
Nor do they provide enough information to determine whether one geographic region, processing architecture, or instance category faces more significant constraints than another.
For existing customers and developers, the practical implications therefore depend on the specific services, regions, instance types, and capacity arrangements they require. Amazon's comments describe substantial pressure on future large-scale capacity, not a universal inability to provide cloud computing.
What to Watch Next
Amazon's next scheduled financial update could provide additional context about AWS growth and the company's expanding infrastructure investment.
Amazon has scheduled its third-quarter 2026 earnings conference call for October 29, when investors will be looking for information about cloud revenue, capital expenditure, and management's outlook for future capacity.
Several developments will help clarify the significance of Jassy's remarks: whether AWS provides more detailed capacity guidance, how quickly newly constructed infrastructure becomes operational, and whether growing supply translates into increased customer usage and revenue.
Another important question is whether demand continues to justify the pace of investment. Large customer commitments can support infrastructure planning, but they do not remove the financial and operational risks associated with building capacity years ahead of its expected use.
For now, Jassy's disclosure offers a clear indication of the intense competition for large-scale AI computing resources. The key unanswered question is not whether AWS has stopped serving customers, but how quickly Amazon can convert its substantial infrastructure investments into usable capacity while meeting commitments that already stretch into 2028.
0 Comments