Bolt.new WebContainers cost advantage

Diving deeper into

Bolt.new

Company Report
Lovable's infrastructure relies on Fly.io's cloud containers rather than browser-native execution, making it structurally more expensive to operate than Bolt.new's WebContainers-based architecture.
Analyzed 6 sources

This cost gap shapes which business model each product can support. Bolt pushes most code execution into the user’s browser through WebContainers, so each new session avoids a rented server by default, while Lovable keeps thousands of apps running on Fly.io containers that are billed as cloud compute. That makes Lovable better suited to always on hosted apps, but it also means infrastructure expense rises more directly with usage and uptime.1 2 3 4 5

  • WebContainers run Node.js inside the browser tab, not on a remote VM. StackBlitz describes this as a browser based runtime for Node.js, and Bolt leadership ties that architecture to near 70% gross margins and profitability because spinning up an environment in the browser costs effectively nothing at the compute layer.1 5
  • Lovable is operating a much more server heavy path. Internal research notes more than 4,000 Lovable instances continuously hosted on Fly.io, and Fly bills Machines by the second while they run, with pricing tied to CPU and RAM. That means every active app can carry ongoing hosting cost even after generation is done.2 3 4
  • The tradeoff shows up in product behavior. Lovable syncs code to GitHub and bundles hosted integrations like Supabase and Clerk, which fits a managed app workflow. Bolt has focused more on instant local generation and then handing code off, including repo download, which reduces infrastructure load but puts more hosting responsibility elsewhere.4 5 6

The category is likely to split between low cost generation layers and higher cost managed app clouds. If Bolt keeps turning browser native execution into enterprise workflows and production ready handoff, it can preserve better margins while competitors that own persistent hosting will need either much higher ARPU or much tighter infrastructure efficiency.1 2 5 6