HOSTING_COST_CALCULATOR

Reading the transfer/speed line:"Mbps" is the connection speed (how fast data moves); "GB/mo transfer" or "port" is the monthly quota before extra charges kick in. Only outbound traffic counts — visitors downloading your site, images, files. Inbound (uploads to your server) is free everywhere.

Contabo(marked *) has no fixed monthly cap — traffic is "unlimited" under a fair-use policy, but sustained heavy usage can get the port throttled rather than billed extra. DigitalOcean caps transfer per plan (4,000–6,000 GiB/mo above); going over is billed at $0.01/GiB. Tencent caps transfer per plan too (2,560–7,168 GB/mo above); going over is billed at ~$0.081/GB (Singapore rate).

For a typical business or marketing site, actual usage is typically far below any of these caps unless serving heavy video/image traffic at scale — worth revisiting only if traffic grows a lot.

Estimated visitor capacity(the "~XXXK visits/mo" line on each card): calculated as transfer quota ÷ ~8MB average per visitor session — a working assumption for a content-heavy site with multi-page browsing (home → gallery → details → contact/checkout), where the browser caches repeated assets (CSS/JS/logo) across page views. Adjust upward if pages are heavier (uncompressed photos, video) or downward if visitors mostly view one page.

Two different things are being measured here, and it's easy to conflate them:
• Transfer (GB/mo)caps total monthly traffic — at these tiers it's 320K–896K visits/month, far beyond what a new site will see even after successful launch. This is essentially never the real constraint.
• Speed (Mbps)affects how fast a single page loads, not a hard visitor ceiling — at 30–45 Mbps, one visitor's full session downloads in ~1.4–2.1 seconds if the pipe were dedicated to them alone; with several people browsing at once the pipe is shared, but real bottlenecks only show up with many simultaneous heavy downloads, which is uncommon for a small site.

The actual ceiling on concurrent users is CPU/RAM (the 2/4/8 vCPU tiers), not the network — how many requests, database queries, and page renders the server can process at once. That depends on your specific backend setup, but as a rough order of magnitude, a well-configured 2 vCPU/4GB instance typically handles low hundreds of concurrent active users comfortably, scaling roughly with vCPU count. Recommend load testing before committing to a tier if real traffic volume is a known concern.