AWS lets one EventBridge custom bus serve every account in an organization through RAM

AWS's launch post describes the setup it is replacing. Its best practice puts each team in a separate account. Moving events between those accounts meant several buses joined by cross-account rules or bus-to-bus configurations ^[1]. The post lists what followed: platform teams lost sight of who subscribed to which events, cross-account and bus-to-bus routing charges compounded, and teams that needed ordering built workarounds or moved to other technologies ^[9]. In a candid line for a launch post, AWS wrote that the workaround "reintroduces the operational complexity that serverless architectures are meant to eliminate" ^[14].

The cross-account part of the fix is a permissions change. The post's author wrote: "Sharing uses AWS Resource Access Manager (AWS RAM), so I did not have to set up cross-account permissions or bus-to-bus routing myself" ^[5]. I think reusing RAM is the right design. The bus gets its sharing model from an existing AWS service. EventBridge does not have to invent a cross-account permission scheme of its own.

On the new bus, publishers send events, subscribers consume only what they need, and EventBridge handles ordering, retention, routing and delivery ^[15]. AWS says platform teams keep visibility into all event flows and fine-grained control over who can publish and consume ^[10]. It also says the larger subscriber quota cuts the fragmentation that happens when limits force a split across buses ^[7]. That visibility covers the shared bus. A rule left on a classic bus in a team account keeps routing through rules and targets ^[3], and those flows sit outside the shared bus.

Ordering is where teams have the most homemade code to reconsider. AWS's example is a logistics application where driver location updates must arrive in sequence, because out-of-sequence events make routing algorithms act on stale data ^[11]. Ordered and unordered consumers can run on the same bus ^[12]. A team can delete its own sequencer once the documentation answers three questions. Does order hold per group for every subscriber? What happens to later events in a group when one delivery fails? How much throughput does a single group get?

AWS says the new pricing model "delivers improved economics at scale" and adds cost allocation for publishers and subscribers ^[13]. The post does not include per-event prices. For the saving to transfer to a given estate, the new per-event charge has to come in below what that estate pays today for an event plus every cross-account hop it takes. Estates that fan events through several team buses stand to gain the most. A two-account setup with one forwarding rule may see little change.