01What happened
By the company's own account: main 2026 capacity is sold out, 2027 is essentially sold out, and a substantial share of 2028 is already committed. Some capacity is held back for on-demand instances and model inference, but the whole blocks that suit large training runs and cluster deployments are largely reserved.
Demand is described as two-ended: model labs and a handful of fast-growing AI applications at one end, enterprises using AI to cut cost at the other.
AWS booked $42.2B of revenue in Q2, up 36.7% year on year; its AI business runs above a $25B annual rate.
02By the numbers
- $220B
- planned capex for 2026
- 2027
- the year capacity is essentially sold out
- $25B
- annualised AI revenue
Related items

China debuts its first fully domestic 100,000-card AI supercluster
On 18 October the first fully domestic 100,000-card AI supercluster, Sugon 8000, debuted in Shenzhen at the 2026 national HPC conference: a heterogeneous domestic-chip architecture that fuses HPC and AI natively, already running in Zhengzhou and connected to the national supercomputing network.
Read the record
Meta publishes the weights for Muse Spark Open
Meta released Muse Spark Open, an open-weights model that runs on two paired Mac desktops and scores 39 on the Artificial Analysis Intelligence Index. Meta calls it the most capable open-weights model from the US.
Read the recordWant to apply these facts to your business? Bring the question Start AI diagnostic
This section is maintained by CodeLeo: topic and structure are ours, imagery comes from a licensed library, and the text runs through fact sheet → AI-generated draft → human review. The copy therefore contains AI-generated content. It is a public record of facts, states facts only, carries no commentary, and does not provide newsgathering or syndication services.
Company, product and brand names appear only to identify them. This section has no affiliation, agency, sponsorship or partnership with those companies, and nothing here recommends or endorses their business.
