MigrationNew

    What Leaving Lovable Hosting Actually Costs You

    A migration quote prices the move. It rarely prices the month after. Here is every line that shifts from Lovable's managed plan onto your own bill and your own calendar.

    A migration quote covers the move. It rarely covers the month after the move.

    That is convention more than dishonesty. Build and migration work is advertised as a project with an end date, so the recurring scope that arrives the day you take delivery sits outside the line item. Nobody is hiding it. It is simply not what the quote is for.

    If you are weighing a move off Lovable's managed hosting onto your own Next.js stack, the engineering is the part you can already see. What follows is the part you inherit.

    What the managed plan is actually paying for

    Lovable's hosting documentation is unusually plain about what is included. Hosting is fully managed, every published app is served over HTTPS through a global CDN, and there is no separate server to set up. Then there is the line most founders never register until it is gone: “Your plan does not cap visitors, requests, or bandwidth.”

    Rendering is included too, and which kind you get depends on the project's age. Apps created from 13 May 2026 use TanStack Start with server-side rendering. Older React and Vite projects get on-request pre-rendering on deployed public URLs, served only to verified search and AI crawlers. Either way, somebody else operates it, patches it and keeps it up.

    One asymmetry is worth knowing before you plan around it. Lovable separates hosting from its built-in Cloud backend, and the docs note that running out of credits halts development and backend functions but does not take the published site off the internet. The site stays up. The ability to change it does not.

    The line items, before and after

    Line itemOn Lovable's managed hostingOn your own Next.js stack
    Hosting billFlat plan; documented as no cap on visitors, requests or bandwidthMetered against actual usage, on your account
    Page renderingSSR via TanStack Start, or crawler-only pre-rendering on older React + Vite projectsYour rendering strategy, chosen per route and operated by you
    Build pipelineNot your concernA cost centre with machine-size and caching decisions
    TLS and custom domainCertificates provisioned automatically; custom domain on a paid planYours to configure and keep valid
    ObservabilityPlatform levelYours to instrument, route and answer
    Framework upgradesPlatform moves; an in-place upgrade path is documentedYour calendar and your regression surface
    Backend availabilityCredits gate development and backend functionsProvider plan and limits, on your account

    Four costs that do not appear on the quote

    1. The bill changes shape, not just size

    A flat plan with no documented cap on visitors, requests or bandwidth is a fixed number you can put in a spreadsheet in January and still recognise in June. Metered infrastructure is not. You may well pay less in absolute terms; what you lose is the ability to know the figure in advance. For a pre-revenue product with unpredictable traffic, predictability is worth something, and it is the first thing a migration hands back.

    2. Builds become something you tune

    Once the build pipeline is yours, it is a dial rather than a given. Vercel added the ability to opt into enhanced Turbo build machines on individual deployments on 17 September 2026, specifically framed as a way to manage costs, and on 18 September 2026 added Spend Management for Enterprise Flexible Commitment plans with budgets, email alerts, webhooks and optional pausing of production deployments. On the framework side, Next.js 16.3, released 3 August 2026, turned on Turbopack disk caching for next build by default so repeat builds read unchanged artifacts from cache.

    All of that is good news. It is also four decisions that used to be nobody's job at your company.

    3. Observability becomes a job, not a feature

    The tooling here got markedly better in September 2026, which is easy to mistake for the work getting smaller. Supabase shipped four prepared agent roles - Health, Security, Performance and Capacity - on 13 September 2026, along with a query_logs MCP tool giving structured access to logs across API, Auth, Storage, Edge Functions and Postgres. On 18 September 2026 it added service error rate checks to Advisors, covering PostgREST, Auth, Storage and Edge Functions.

    Better instruments do not read themselves. Once you are operating your own stack, somebody has to own the alert that fires at 2am and decide whether it matters.

    4. The upgrade calendar becomes yours

    On a managed platform, the platform moves and you follow. Off it, you choose when to move and you absorb what breaks. Next.js 16.3 is a fair illustration: it landed on 3 August 2026 with Instant Navigations and Partial Prefetching behind opt-in flags, and the Next.js team has said those behaviours will become the default in a future major version. Opting in early is a choice. So is not opting in, and then doing it all at once later.

    When the move earns its cost

    Migrating is the right call when you need control the hosted platform does not give you: your own caching rules, your own edge and middleware behaviour, your own build pipeline, per-route rendering decisions you make rather than inherit. If your product's value lives in something the platform will not let you run, the recurring scope is simply the price of the product existing.

    It is the wrong call when the honest answer is that your app is mostly marketing pages on a stable data model and nothing about the platform is actually in your way. In that case Lovable's documented in-place upgrade to TanStack Start is the cheaper and less disruptive path, and the flat uncapped plan is a feature rather than a limitation you are escaping.

    The question to answer before you sign anything is not whether Next.js is better. It is which of these seven line items you are ready to own.

    Where to start

    Price the month after, not just the move. Write down the seven rows above for your own app and mark the ones you have a plan for. If most of them are blank, the migration is not ready, whoever is quoting it.

    If you would rather have someone else run that exercise and tell you plainly what moving would cost you in recurring scope, our Strategy Audit is $499 one-time and is credited toward any build. If you already know the answer is a rebuild, the Publishable MVP package is $4,999 one-time, and the Founding Client package is $2,999 one-time with a 14-day build guarantee and a full refund if the deadline is missed. Every build carries the same promise as the rest of our work: first paying customers within 90 days of launch, or a full refund. We have completed 80 jobs on Upwork on exactly this stack.

    If we are not the right fit, we will say so on the call. Book your free 20-minute intro call.

    Sources

    Related posts