Skip to main content

AWS capacity largely booked through 2027

26 OCTOBER 2026·2 MIN READ·4 SOURCES

AWS cloud-computing capacity was largely sold out through 2027, with significant commitments extending into 2028, according to a report published by The News International on October 24, 2026.

AWS capacity largely booked through 2027

Key takeaways · 4

  • 01

    AWS capacity for 2026 was sold out, while major allocations for 2027 and some 2028 capacity were largely committed.

  • 02

    AWS keeps some capacity available on demand for services including Bedrock and EC2.

  • 03

    Amazon’s AWS AI business surpassed a $25 billion annualized run rate within three years of the generative AI boom.

  • 04

    Physical constraints cited in the report include data-center real estate, power availability and specialized memory.

Capacity is booked years ahead

AWS cloud-computing capacity for 2026 was already sold out, and its major allocations for 2027 were largely committed.[2] A good portion of 2028 capacity was also sold out, while significant commitments extended into that year.[2] The figures describe capacity commitments and allocations, rather than a claim that every AWS service or instance type is unavailable.[2] For customers planning deployments, the distinction makes it important to verify availability for the specific service and timeframe they need.

AI demand meets physical constraints

Jassy said Amazon was struggling to keep pace with demand from AI labs and enterprises.[2] The News International report said AWS’s artificial intelligence business surpassed a $25 billion annualized run rate within three years of the generative AI boom.[1] It also identified data-center real estate, power availability and specialized memory as physical constraints.[1] Those constraints help explain why adding capacity involves more than securing customer demand: infrastructure and components are also part of the challenge.[1]

On-demand services remain in the mix

AWS keeps some capacity available for on-demand services, including Bedrock and EC2.[2] EC2 provides AWS cloud servers for running applications, while Bedrock gives customers access to AI models for building generative AI applications.[2] This availability is a caveat to the sold-out figures, but the evidence does not specify how much on-demand capacity remains or whether it can meet a particular customer’s needs.[2] Teams should therefore check current service-level availability rather than treating the broad capacity figures as a guarantee of access or a complete outage.

Spending and hardware plans

Amazon had said earlier in 2026 that it would spend about $220 billion in capital expenditure that year.[2] The report also said investing substantially more would not be a bad use of capital.[2] Plans called for deploying two million additional NVIDIA GPUs across AWS’s global infrastructure and deepening work on AI factories, CPUs, networking, open models, data processing and robotics.[4] AWS’s hardware options include NVIDIA GPUs, AWS Trainium chips, or the two working together.[4] Amazon was scheduled to report third-quarter results on October 29.[2]

Capacity planning should account for the possibility that preferred AWS resources may be committed well ahead of a project’s launch. Teams can reduce uncertainty by checking availability for their specific workloads and considering whether on-demand services or alternative hardware options fit their needs.

Why it matters
Story quiz

Bring this to your team — one session a day, built for your industry.

kju.ai is built for teams: daily sessions, team analytics, and rollout support for your organization.

For teams

How this developed

  1. 26 October 2026

    AWS capacity largely booked through 2027

Sources

AI fluency, one session a day, built for your work.