Amazon CEO Jassy: AWS 2026 capacity is sold out, and most of 2028 is also booked

Amazon's cloud computing division is facing unprecedented market demand, with the vast majority of its core infrastructure capacity already snapped up for the next several years.

Amazon CEO Andy Jassy said in his latest statement that most of Amazon Web Services (AWS) capacity has been booked through 2028. This development highlights sustained and strong willingness among enterprises to invest in cloud services and artificial intelligence infrastructure.

Despite the surge in forward orders, Amazon has maintained a high degree of operational flexibility at the capital expenditure level. The company can adjust its procurement plans for servers and networking equipment within short cycles based on changes in customer demand, and this defensive mechanism effectively alleviates investor concerns that cloud service providers may over-expand production or that a future slowdown in demand could lead to overcapacity.

Beyond the supply and demand situation for computing power, the deepening of the internal hardware ecosystem is becoming a key support for Amazon in responding to market demand. Early strategic acquisitions and recent investments in external artificial intelligence startups are creating synergies and continuing to drive its self-developed chip strategy forward.

Forward computing power demand is strong, core services reserve flexible space

In an Andreessen Horowitz podcast episode released on Friday, Jassy elaborated on AWS's current capacity booking situation. He noted that as enterprise demand for computing resources continues to climb, AWS's large-scale capacity for the coming years has basically peaked.

"At this point, 2026 capacity is clearly sold out, and 2027 is basically sold out," Jassy said on the podcast. "A large portion of 2028 is also sold out."

Although large blocks of capacity have been locked in advance, AWS has not completely cut off supply for short-term demand. Jassy added that Amazon will always reserve a certain amount of computing power for on-demand services as well as EC2 for running applications and the inference service Bedrock for running AI models, to ensure the normal operation of core services and customer experience.

Capital expenditure is defensive, hardware procurement adjusted on demand

In response to risks that may arise from market cycle fluctuations, Amazon has adopted a more flexible strategy for data center hardware investment. Even if the overall growth rate of computing power demand slows in the future, Amazon's existing supply chain management model can effectively cope.

Jassy emphasized that AWS typically only needs to purchase servers and networking equipment a few months in advance. This shorter procurement cycle allows the company to adjust capital expenditure in close alignment with customers' actual needs, avoiding the lagging risks brought by heavy asset investment.

"If it turns out you don't need as much capacity to fill these data centers, you won't buy the related servers and networking equipment," Jassy stated clearly. This on-demand procurement logic provides direct protection for Amazon's profit margins and cash flow.

Self-developed chip strategy deepens, cooperation boosts technical capability

While discussing capacity, Jassy highlighted Amazon's long-term layout in the underlying hardware field in the wide-ranging interview. He described the acquisition of chipmaker Annapurna Labs more than a decade ago as possibly the best acquisition Amazon has ever made.

Annapurna Labs is currently the core driver of Amazon's internal self-developed chip projects, responsible for developing key hardware products including the artificial intelligence chip Trainium and the CPU processor Graviton. These self-developed hardware products not only reduce AWS's dependence on external suppliers but also provide customers with more cost-effective computing options.

In addition, ecosystem cooperation is feeding back into Amazon's hardware development. Jassy noted that Amazon's investment in artificial intelligence startup Anthropic has played a positive role in promoting its own chip manufacturing business.

"[Anthropic] has been developing on Trainium for some time, and I think this is very beneficial to Anthropic and also very beneficial to our Trainium team," Jassy said. "We learn a lot from each other." Risk disclosure and disclaimer

Markets carry risks, and investment requires caution. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial situation or needs of individual users. Users should consider whether any opinions, views or conclusions in this article are suitable for their particular circumstances. Investment based on this is at the user's own risk.