Most MVPs cost between $2,000 and $120,000. The spread is not mostly dishonesty. It is that “MVP” hides four completely different purchases, and a quote only makes sense once you know which one you are being sold.
| What you are buying | Typical cost | Real risk |
|---|---|---|
| US or UK agency build | $30k to $120k+ | Spending most of a pre-seed round before your first user |
| Offshore project agency | $25k to $40k | Fixed scope, then a handover cliff |
| Senior freelancer | $10k to $30k | One person’s availability is your timeline |
| No-code and templates | $2k to $10k | A ceiling you will hit, then a rewrite |
The four things people call an MVP
$30k to $120k+: the US or UK agency build. Senior product people, polished process, local timezone, rates of $100 to $200 an hour. The build is usually excellent. The problem is arithmetic, and we will come back to it below. Part of what you are paying for is the agency’s office, its sales team and its brand.
$25k to $40k: the offshore project agency. Eastern European and Indian firms with strong portfolios. Good ones deliver real quality at half the US price. The risks are the fixed-scope contract and the handover cliff. Change your mind mid-build, and you will, because learning is the entire point of an MVP, and change orders start arriving. Then the team disbands at delivery and bug fixes become a new negotiation.
$10k to $30k: the senior freelancer. Can be outstanding value. It is also a single point of failure: one person’s availability, health and motivation standing between you and launch. Continuity is the real cost. When they move on, your next developer inherits an undocumented codebase.
$2k to $10k: no-code and templates. Genuinely the right answer for validating demand. A Bubble app or a template landing page tests willingness to pay in weeks. The ceiling arrives fast though: scaling limits, integration walls, and technical investors who discount no-code builds because they know a rewrite is coming. Validate with no-code. Do not build your company on it.
How much should you actually spend?
This is the better question, and almost nobody answers it, because the honest answer is a fraction rather than a figure.
Look at what founders are actually raising. Carta’s data shows the pre-seed market has gone barbell-shaped: rounds between $1 million and $2.5 million fell from 24% of all pre-seed rounds in Q1 2023 to 18% by Q1 2026, while rounds under $250,000 hit a record 35% in Q4 2025. The comfortable middle is emptying out.
Put the two numbers next to each other. If you are in that 35%, a $60,000 agency build is roughly a quarter of everything you raised, spent before a single user has told you whether the thing is worth building. This is a runway argument rather than a moral one. The MVP is supposed to buy you information, and information you cannot afford to act on is not worth much.
A workable rule: the build should cost less than a third of your runway, and it should leave you able to pay for the version that comes after it. Almost every founder underestimates the second one. The first build is rarely the expensive part.
If you have not raised, the ceiling is whatever you can lose without the company ending. That number is usually a lot smaller than the quotes in your inbox, and that is a reason to change the shape of the deal rather than to find a cheaper builder.
What AI changed in 2026, and what it did not
It genuinely changed the floor. A technical founder with a coding assistant can now get a working prototype out in a weekend, and the $2k to $10k band has quietly absorbed a lot of work that used to sit above it. If you can code at all, build the first version yourself. It has never been cheaper to find out you were wrong.
What it did not change is everything that makes software survive contact with users. Auth that does not leak. A schema you can migrate. Payments that reconcile. Error handling for the paths nobody demoed. An assistant will write all of that quickly and will not tell you which parts are wrong, and reviewing generated code you do not understand is slower than writing it.
So the bands have not collapsed. They have pulled apart. The cheap end got cheaper and faster. The expensive end is still expensive, because what you buy there was never typing speed.
We are not putting a percentage on any of this. We have not seen a source for one we would trust, and a made-up number would be worse than none.
The question under every quote
Every band is answering the same question: how much certainty are you buying, and when do you pay for it? Agencies front-load cost and lock scope early. That is maximum certainty at the worst possible timing, given that an MVP exists precisely because you do not know the final scope yet. No-code minimises cost but caps the upside.
What the sticker price leaves out
The quote covers the build. These usually arrive separately, and together they are rarely trivial:
- Design, if it is not in the number. Ask explicitly.
- Infrastructure, from hosting to the third-party services you will not notice until the first invoice.
- App store fees and review time if there is a mobile app, which is a schedule cost as much as a money one.
- The second build. Everything you learn from the MVP is worthless unless you can afford to act on it. Budget for the changes, not just the launch.
A cheap build you cannot iterate on is more expensive than a fair one you can.
The subscription alternative
The model we run at the Studio restructures the timing: a flat monthly fee of $749 to $1,499, an MVP live in 8 to 12 weeks, then continuous shipping, one active request at a time, pause or cancel the cash-only plan anytime. Over a four-month build that is $3,000 to $6,000. Freelancer money for an agency-style team, with no $60,000 cheque before you have learned anything.
The catch, stated plainly: at the equity tier the cash price is low because we take a 5% to 15% stake, vesting monthly, in the companies we build. That is not a discount. It is a different deal, closer to a technical co-founder than a vendor, and we are selective about it because we only make money if the product wins. We have written up how the equity model works, including the cap-table questions your lawyer will ask, and how to size an equity grant against what a founding engineer normally receives.
Whatever you choose, insist on these
- Repo in your GitHub org from day one. Not delivered at the end. Day one.
- A deployed preview URL by week two, updated continuously. Progress you can click, not screenshots.
- Boring technology (TypeScript, PostgreSQL, standard frameworks) that your future hires already know.
- Weekly demos in plain English, not sprint jargon.
- A written answer to “what happens when we stop?” Handover docs, IP assignment, transition support.
Any builder, us included, should agree to all five in writing without flinching. The ones who flinch just answered your real question.