Plan transit gateway attachments, routing and bandwidth across connected VPCs.
Transit gateway planning spans three dimensions: cost (a flat hourly fee per attachment plus a usage-based fee for data actually processed — cost = attachments × $/hr × hours + data_processed_GB × $/GB), route capacity (each attachment contributes routes to a shared route table with a practical ceiling, commonly cited around 10,000 routes), and bandwidth (aggregate throughput divided across attachments, each with its own burst ceiling).
Total monthly cost
cost = attachments × price_per_hr × hours_per_month + data_processed_GB × price_per_GB
Route tables are held in memory on the gateway's control plane and propagated to every attachment, so there's a practical ceiling (commonly around 10,000 routes per route table on major cloud platforms) to keep propagation and lookup performant at scale — large deployments should plan for route summarization rather than one route per fine-grained subnet.
Most cloud providers charge a flat hourly fee per attachment (VPC, VPN, or Direct Connect/ExpressRoute) plus a separate per-GB fee for data actually processed through the gateway — cost scales with both how many things you connect and how much traffic flows between them.
When total routes across attachments approach the route table's practical limit, summarizing many small routes into fewer, larger aggregate routes (where addressing hierarchy allows) reduces route table pressure and keeps the gateway comfortably under its capacity ceiling.