Amazon Has Already Sold Nearly All of Its Major Cloud Capacity Through 2028

Amazon is struggling to keep up with the soaring demand for AI computing power. CEO Andy Jassy says nearly all of Amazon Web Services’ (AWS) cloud capacity for 2026 has already been reserved. Most of 2027 is booked too, and a significant share of capacity for 2028 has already been committed.

Jassy made the comments on Friday, October 23, on a podcast hosted by investment firm Andreessen Horowitz. His remarks, reported by Reuters, highlight the strain that AI’s rapid growth is putting on global digital infrastructure.

Amazon Web Services provides the servers and cloud services that companies use to run their websites, applications and AI systems. It is one of Amazon’s most important businesses.

Demand for powerful servers has surged in recent years. AI companies need vast amounts of computing power to train new models, while more organizations are bringing AI into their day-to-day operations.

Jassy says nearly all the large blocks of capacity reserved in advance for the coming years are already spoken for.

That does not mean AWS can no longer take on new customers. Amazon keeps capacity available for services such as EC2 and Bedrock, which let customers rent servers and use AI models.

The distinction matters: the pressure is mainly coming from large contracts in which customers reserve substantial amounts of computing power for extended periods.

As a result, companies that need this kind of capacity may find it harder to secure it immediately.

Amazon is trying to expand supply by building data centers, developing specialized AI chips and increasing its energy capacity.

The company has planned about $220 billion in capital spending for 2026, much of it tied to expanding its technology infrastructure.

Jassy said on the podcast that Amazon could invest even more if that would allow it to meet demand.

AI Wereld previously reported on AWS’s massive investment in AI infrastructure. Jassy’s latest comments make clear that demand remains strong despite the spending.

But expansion faces physical limits. New data centers need suitable sites, electricity, cooling, network connections and enough chips.

A company can order more servers, but that does not mean a fully operational data center will be ready quickly.

Amazon’s comments underscore a shift in the AI industry. The debate is no longer just about which language model is smartest; it is also about who can access enough computing power.

That could put smaller AI companies at a disadvantage. Large organizations can sign long-term contracts to lock in capacity, while startups are more likely to depend on cloud services that are available on demand.

Limited infrastructure could make expansion more expensive or complicated. But Jassy’s comments do not automatically mean that prices will rise for all AWS customers.

Energy use is also drawing more attention. New AI data centers consume huge amounts of electricity, while power grids in several countries are already under strain.

The same trend is playing out in the Netherlands. AI Wereld has previously covered the impact of data centers on the Dutch power grid.

The scale of these reservations reflects customers’ confidence in future AI applications. But it does not guarantee that every investment will ultimately pay off.

If demand for AI computing power grows more slowly than expected, technology companies could be left with costly infrastructure.

Jassy acknowledges that risk, but says Amazon can wait until relatively late to adjust orders for some components. That means the company does not have to commit years in advance to buying all its servers and networking equipment.

Investors will be watching how quickly Amazon can bring its new capacity online—and how much revenue it generates once it does.

Amazon is currently scheduled to report its quarterly results on October 29. The figures could shed more light on AWS growth and the financial impact of the company’s huge investments.

For now, the key takeaway is that the biggest cloud providers are competing not just for customers, but also for electricity, chips and the space needed to expand their services.