New Delhi: Amazon Web Services (AWS), the cloud computing arm of Amazon, has run out of available capacity for 2026, with most of its major capacity allocations for 2027 already booked and a significant portion of 2028 commitments also taken. Amazon CEO Andy Jassy has said the company is struggling to meet the rapidly increasing demand for computing infrastructure, driven largely by artificial intelligence (AI) development and adoption.

Speaking on a video podcast hosted by venture capital firm Andreessen Horowitz (a16z), released on October 23, Jassy described the growing pressure on AWS infrastructure. He indicated that Amazon could invest more than its planned capital expenditure of around $220 billion for 2026 if additional spending helps the company address demand.

The comments highlight the growing challenge faced by cloud providers as AI companies, research laboratories and enterprises compete for computing resources needed to train, deploy and operate AI models.

AWS capacity booked years in advance

Jassy said AWS had effectively committed its major capacity allocations for 2026 and that most of the significant allocations for 2027 were also taken. He added that a substantial portion of capacity for 2028 had already been booked.

The company continues to retain some resources for on-demand services, including Amazon Elastic Compute Cloud (EC2) and Amazon Bedrock. This means customers may still be able to access certain computing resources, depending on availability, even as large capacity allocations are committed in advance.

EC2 provides cloud-based computing infrastructure, including virtual servers that businesses can use to run applications and workloads. Amazon Bedrock allows developers and organisations to build generative AI applications using supported AI models.

The distinction between long-term capacity commitments and on-demand availability is important. Jassy’s remarks do not mean that every AWS service or computing instance is unavailable to new customers. Rather, they indicate that much of the company’s large-scale capacity has already been allocated, limiting the flexibility available to accommodate additional demand.

The situation also reflects how AI infrastructure is becoming a long-term planning priority for technology companies, which need to secure computing resources well before they are required for commercial deployment.

AI demand puts pressure on infrastructure

The expansion of AI applications is increasing demand for data centres, servers, memory and electricity. These resources are essential for training AI models and running inference, the process through which an AI model generates responses or performs tasks after receiving a prompt or instruction.

Jassy highlighted constraints involving data-centre space, power availability and memory supplies. Expanding cloud capacity is therefore not simply a matter of ordering additional servers. Companies must also secure suitable sites, obtain adequate electricity and procure specialised components.

AI accelerators rely on high-bandwidth memory (HBM), which supports the movement of large amounts of data during demanding computing operations. Traditional cloud services also require dynamic random-access memory (DRAM). When supplies are constrained, cloud providers must manage competing requirements across AI and conventional workloads.

AWS serves customers beyond the AI sector, including businesses that depend on cloud infrastructure for applications, databases, websites and other digital services. Amazon must therefore balance the expansion of AI computing with the needs of its established cloud customers.

The resource constraints described by Jassy suggest that infrastructure availability could remain a challenge even as technology companies increase investment. Building data centres and bringing new computing capacity online require coordination across construction, electricity supply, networking equipment and semiconductor procurement.

Amazon could increase its infrastructure spending

Amazon had indicated plans to spend approximately $220 billion on capital expenditure in 2026. The investment supports the company’s broader infrastructure requirements, including data centres, servers and equipment needed to serve growing demand.

Jassy’s latest comments suggest that the planned amount may not represent a strict ceiling if the company identifies additional opportunities to expand capacity. He indicated that further investment could be justified given the volume of demand AWS is unable to accommodate.

However, the comments do not establish that Amazon has formally revised its spending target. Any increase would depend on the company’s decisions and its ability to secure the infrastructure, components and power required to expand operations.

The scale of planned spending also illustrates the financial commitment required to compete in AI infrastructure. Cloud providers need to invest ahead of demand, while ensuring that new facilities and equipment can be brought into service and generate revenue.

For Amazon, the challenge is to expand quickly enough to serve customers without compromising the reliability of existing cloud services. The company has not publicly disclosed a detailed breakdown of the capacity commitments underlying Jassy’s remarks.

Amazon develops its own AI chips to lower costs

Alongside infrastructure expansion, Amazon is developing its own chips to improve computing performance and reduce the cost of operating AI systems. One of its initiatives is Trainium, a family of processors designed for AI workloads.

Jassy said Amazon’s Trainium chips offer a price-performance advantage of around 30 to 40 per cent compared with comparable alternatives. Such improvements could help the company serve more workloads with its available infrastructure, although actual results depend on the applications and systems involved.

Lowering the cost of AI inference is another priority. Jassy argued that inference remains considerably more expensive than it needs to be for AI to become widely used across everyday business operations.

Reducing these costs could make it easier for businesses to deploy AI features in customer service, software development, data analysis and other applications. It could also encourage customers to move from limited experimentation towards broader use of AI tools.

Amazon’s chip development and cloud infrastructure investments are connected: more efficient processors could improve how effectively the company uses its data centres, while additional capacity would allow it to serve a larger volume of customer workloads.

What the capacity crunch means for cloud customers

The shortage of available capacity highlights the importance of infrastructure planning for companies that depend on cloud computing. Organisations preparing large AI deployments may need to plan capacity requirements well in advance, assess different infrastructure options and monitor the availability of the services they intend to use.

For smaller businesses, the immediate impact will depend on the specific AWS services, locations and computing configurations they require. Jassy’s remarks do not indicate that all customers will face interruptions or that all on-demand services are unavailable.

Amazon is scheduled to report its third-quarter financial results on October 29, 2026. Investors and customers will be watching for further information on AWS growth, infrastructure investment and the company’s plans to address capacity constraints.

The broader picture is that demand for AI computing is growing faster than available infrastructure can readily expand. AWS’s capacity commitments through 2027 and into 2028 underline the scale of the challenge, while Amazon’s spending plans and development of in-house chips show how the company is attempting to respond. Its ability to bring additional capacity online will be an important factor in determining how effectively it can meet future demand.