BREAKING
Ramu, Tharanesh break national records at junior athletics championshipsPrabhas-starrer ‘Salaar Part 2’ to go on floors in 2027: Prashanth NeelAditya storms into World Boxing U-19 semifinals, assures medalManipur: 17 Kuki militants held after discrepancy in identity checks raise questions | DetailsKheerthi Ganta, Natura Bethi and Aarohi Mathur win third-round gamesDelhi man arrested in alleged Siddaramaiah sting operation case‘Hyderabadi boys’ did the job, says Nitish after New Zealand heistAWS capacity sold out for 2026, much of 2027 already booked: Andy JassyAshley Westwood breaks silence on Vibin Mohanan and Adrian Luna captaincy buzzFarooq Abdullah says Pakistan’s diplomacy raises Kashmir issue globally
Technology

AWS capacity sold out for 2026, much of 2027 already booked: Andy Jassy

hellohimachalse@gmail.com 24 Oct 2026 1 hour ago

The artificial intelligence boom is creating a problem for cloud providers. The artificial intelligence boom shows that there is not enough computing capacity to meet demand. The artificial intelligence boom forces cloud providers to search for computing capacity.

Amazon CEO Andy Jassy says Amazon Web Services has already sold out most of its cloud capacity for 2026. Many of the spots for 2027 have also been taken. Even 2028 is starting to get filled up. Companies are trying hard to get the computing power they need for their AI workloads.

In a video podcast from a16z released on October 23 Jassy said that demand for AI infrastructure is growing fast and that Amazon is struggling to take on new customers. The company might end up spending more than its planned $220 billion, in capital expenditures for 2026 as it pushes to grow its infrastructure.

AWS struggles to keep up with AI demand

Jassy said Amazon is having to balance the rapidly growing demand for AI computing with its traditional cloud business, which is also expanding. The challenge is that both businesses depend on resources that are increasingly difficult to secure, including electricity, data centre space and memory.

“If you think about it, there’s only so much data center and power today. There’s only so much memory, and I don’t think that that shortage ends in 2027,” he said.

The bottleneck extends beyond simply constructing more data centres. AI workloads require specialised high-bandwidth memory (HBM) for AI accelerators, while Amazon’s conventional cloud services also rely heavily on DRAM. With supplies under pressure, Amazon has to carefully decide where its available resources should be allocated.

Jassy said the tight supply situation is also a factor behind the possibility of higher infrastructure spending.

“’26 is obviously at this point sold out. ’27 is largely sold out. We always leave a certain amount for on-demand in our inference service, Bedrock or EC2, but largely sold out, all the big chunks. A good chunk of ’28 is sold out,” he said.

AWS provides services including EC2, which offers virtual servers, and Bedrock, which allows businesses to build AI applications using different AI models.

Amazon wants to bring down AI costs

Capacity isn’t the only concern. As companies move from experimenting with AI to using it at scale, the cost of running models remains a major hurdle.

Jassy believes inference- the process of generating a response or completing a task after an AI model receives a prompt – needs to become significantly cheaper.

“It’s still an order of magnitude too expensive in my opinion for it to be used as pervasively as possible,” he said.

Amazon is also developing its own Trainium AI chips as part of its effort to improve performance while lowering costs. Jassy said the chips deliver 30 per cent to 40 per cent better price-performance compared with comparable alternatives.

Related News