How Much Does It Cost to Make an App in 2027?

Key takeaways

• Most custom apps cost $45,000 to $120,000. A ready made white label app costs a small fraction of that.
• Complexity, especially the number of user roles, drives cost more than platform, industry or location.
• Regulated categories like fintech, healthcare and betting start where ordinary apps finish, around $60,000 and up.
• For AI features, model choice is a 40x lever on running cost. It matters more than the build.
• The hourly rate is the most misleading number in outsourcing. Total hours decide the bill.
• Google rewrote its Play Store fees in 2026, so the "15% for small developers" rule no longer holds in the US, UK and EU.
• Budget about 1.7x the build price over three years once maintenance and hosting are included.

The short answer: cost by complexity

Complexity is the single biggest driver, so start here before anything else.

Most people asking this question assume they need tier three and actually need tier one or two. The honest test is simple: if a product already exists that does roughly what you want, you are buying a rebrand, not commissioning a build.

What separates the tiers is rarely the number of screens. It is the presence of four things: more than one type of user, money moving between parties, anything happening in real time, and regulation. An app with none of those is simple however many screens it has. An app with all four is complex however small it looks.

Where the money actually goes

A total is only useful if you can see inside it. Here is the real line item breakdown from our Depop style marketplace build, a two-sided resale app with buyer and seller flows, payments and an admin panel:

Two things in that table generalise to almost every project. First, the client apps are only around 37 percent of the build. People budget for the screens they can see and are surprised by the other two thirds. Second, payments and the admin panel together take 20 percent. They are invisible to users and almost always missing from a first-draft scope, which is why so many quotes grow after kickoff.

Those proportions hold as projects scale. A $120,000 marketplace spends roughly the same share on its backend as a $15,000 one. What changes is the absolute number, and how many times you pay the client-app share, because a marketplace with a separate driver or vendor app builds that 37 percent twice.

App cost by feature

This is how agencies actually estimate: list the features, estimate hours for each, multiply by a rate. The table below gives typical build-from-scratch hours for a cross-platform app with a backend, priced at a $50 per hour blended rate. Rescale with your own rate.

Add 15 to 25 percent on top for design, QA and project management, which a feature list never shows. And note the difference between this table and the Depop breakdown above: building marketplace payments from scratch costs $6,000 to $12,000, but the whole payments line on that white label build was $1,200 to $1,500. That gap is the white label argument in one row. A proven codebase already contains most of these features, so you pay to adapt them rather than to build them.

The expensive features share a trait. Chat, live tracking, video and marketplace payouts all involve state that must stay correct across several people at once. That is genuinely hard engineering, and it is where most budget overruns happen.

App cost by type

These are the headline ranges from our own build guides, each one researched separately.

The pattern is clear once the categories sit side by side. Marketplaces cluster at $40,000 to $80,000 for a credible MVP because they all need the same bones: two user types, listings, search, payments and an admin panel. Media apps climb fastest at the full-build stage, because streaming infrastructure and content delivery scale with every minute watched. Single-purpose tools sit lowest.

Popular categories at a glance

Several of the most-searched app categories do not have their own guide on our site yet, so the ranges below are built from the feature table in the previous section: we list the features each category needs, price them at the same $50 hourly rate, then add 20 percent for design, QA and management and 20 percent contingency. Treat them as MVP planning estimates you can reproduce yourself.

These are MVP figures. A full platform in the same category typically costs two to three times more, which is consistent with the guides that do exist: grocery delivery moves from $40,000 to $80,000 at MVP up to $80,000 to $250,000 for a full platform, and a social app with live video and gifting starts where the table above ends, at $60,000 to $120,000. Dating and fitness are the categories where founders most often overbuild, because the features that make the top apps sticky, such as AI matching or personalised training plans, are usually second-year features rather than launch requirements.

App cost by industry

Industry changes the number mainly through compliance, not features. A chat screen costs the same everywhere. A chat screen that must be auditable under healthcare rules does not.

Regulated industries share a shape in our research: compliance frequently costs more than the product features. Sports betting is the starkest example. A full Paddy Power style platform costs $100,000 to $390,000 to build, and our guide puts UK Gambling Commission licensing and first-year compliance at a further £200,000 to £500,000 on top. Telehealth follows the same curve: a HIPAA-ready MVP is $40,000 to $80,000, a production platform $100,000 to $200,000, and adding real AI takes it to $250,000 to $400,000.

Streaming carries a different hidden cost. Content protection is mandatory once you license third-party catalogue, and our DRM cost breakdown covers what that adds. Budget compliance and licensing as line items from the first estimate, not as surprises discovered in month four.

App cost by tech stack

Stack choice changes cost less than people expect, but it changes it predictably.

Cross-platform is the default for good reason: you write once and ship to both stores, which is where most of the saving in a modern build comes from. Going native roughly doubles mobile engineering because you build the same screens twice, in two languages, with two review cycles.

iOS, Android or both? On a native build, one platform costs roughly half of two, since the backend is shared and only the client is duplicated. On a cross-platform build, adding the second platform costs far less, typically 10 to 20 percent extra for platform testing, store submission and the occasional native module. That asymmetry is why "launch on one platform to save money" is good advice for native teams and much weaker advice for cross-platform ones. iOS users typically spend more in-app, while Android reaches more of the global market, so pick whatever your first thousand customers actually carry.

Backend choices matter less than teams argue about. The expensive decisions are architectural rather than linguistic: real-time features, search infrastructure, and anything requiring high availability. Specialised infrastructure follows its own curve, as our custom MCP server cost breakdown shows, running $8,000 to $15,000 for a straightforward build and $60,000 to $120,000 for a production one.

AI app development cost

Almost every app brief in 2027 includes an AI feature, and it changes the cost model in a way traditional estimates miss. An AI feature has two costs: building it, and running it. The build is usually modest. The running cost scales with every user interaction, forever.

Building it. Adding an AI chatbot or assistant through a model API typically takes 60 to 150 hours, roughly $3,000 to $7,500 at a $50 blended rate. A retrieval system that answers from your own documents adds more. Training or fine-tuning your own model is a different order of magnitude and rarely necessary for an app.

Running it. Model providers charge per token, and the price difference between model tiers is enormous. Using Anthropic's published API rates as a worked example, at the time of writing:

*Assumes 1,000 input and 500 output tokens per reply at standard rates. Haiku rates apply to prompts up to 100,000 tokens.

At 100,000 replies a month, that is about $35 on the small model and $1,400 on the largest. The same feature costs 40 times more depending on one configuration choice. Most chatbot answers do not need the largest model, and routing simple questions to a small one while reserving the large one for hard cases is the single most effective AI cost control. Batch processing halves the rate for anything that does not need an instant answer, and prompt caching cuts the cost of repeated context.

AI-heavy products are a different category entirely. A full Character.AI style companion app built from scratch runs $150,000 to $400,000+ in year one with a team of five to seven senior engineers plus machine learning and trust-and-safety specialists, against a $15,000 one-time white label route. At the top end, clinical AI software sits at $60,000 to $250,000 because accuracy and data handling carry real liability.

Can you just build it with AI tools?

AI app builders can produce a working prototype in an afternoon, and for validating an idea they are genuinely excellent. The catch is the gap between a prototype and a production app. Our AI-built app cleanup guide prices professional rescue work at $800 to $3,000 for light cleanup, $4,000 to $12,000 for a standard cleanup with security review and deployment fixes, and $10,000 to $30,000+ for a full refactor to production grade. Budget for that step if you start with an AI builder. It is still often cheaper than a custom build, but it is not free.

App cost by team type

Who builds it changes the cost as much as what you build.

The in-house line is worth spelling out. The US Bureau of Labor Statistics puts the median annual wage for software developers at $135,980 as of May 2025. Two developers is over $270,000 a year in salary alone, before benefits, recruitment, a designer or a product manager. In-house makes sense when the app is the business and will be developed continuously for years. For a first version, it is almost always the most expensive route to the same result.

Freelancers are cheapest on paper and riskiest in practice. The rate is low, but project management, QA, design and continuity all land on you, and a single illness or better offer can stall the project entirely. They work best for a tightly scoped piece of a larger effort.

App cost by region

Developer location changes the rate, and the rate is what most comparisons fixate on. Approximate market rates for app developers:

Treat these as indicative. Published rate surveys vary widely because most are produced by vendors with an interest in the answer. The one thing they broadly agree on is worth repeating: the hourly rate is the most misleading number in outsourcing. A $40 per hour team that needs 1,400 hours costs $56,000. A $90 per hour team that needs 550 hours costs $49,500, and delivers months sooner. Compare total quoted cost and timeline, never the rate alone.

Clone app costs

This is where the biggest savings sit. "Clone" here means building a product with the proven mechanics of an existing app under your own brand, not copying its code or trademarks. There are two completely different prices depending on route.

Our own white label pricing runs in three published tiers, using our Carvana clone as the reference:

Disclosure: Make An App Like publishes this guide and sells these packages. Hosting and third-party API costs are billed separately at actuals.

Set against custom builds of the same models:

Across the wider market, ready made vendors publish entry prices from about $2,499. Our comparison of sites to buy white label apps covers who sells what, what is actually included, and the contractual checks worth doing before you pay a deposit.

The honest trade-off: a white label app gives you a proven product fast and cheaply, but it is proven for someone else's market. If your differentiation is the product itself, you need a custom build. If your differentiation is distribution, pricing, content or geography, buy the codebase and spend the saved six figures on customers.

App store fees in 2027: what Apple and Google take

Store commission is a cost of revenue rather than a build cost, but it belongs in any honest budget because it decides whether the economics work. It also changed significantly in 2026, and most cost guides have not caught up.

Google Play rewrote its fee structure in 2026. In the US, UK and EEA from 30 June 2026, and in Australia and Japan from 30 September 2026, Google now charges:

In markets not yet moved to the new structure, the older rates still apply: 15 percent on subscriptions and on the first $1 million of annual revenue, and 30 percent above that. The practical consequence is that the "15 percent for small developers" rule quoted almost everywhere no longer describes one-off purchases in the US, UK or EU. Subscriptions are now the cheapest model to run through Google Play in those markets.

Apple charges a standard 30 percent commission, reduced to 15 percent under its App Store Small Business Program for developers with up to $1 million in annual proceeds, with further reductions for subscriptions retained beyond their first year. In the United States, the Epic v. Apple litigation now requires Apple to let apps link to outside payment options, and what commission Apple may charge on those purchases is still being decided by the courts. Treat US payment-link economics as unsettled and check current terms before modelling revenue around them.

Both stores also charge account fees: the Apple Developer Program is $99 a year and a Google Play developer account is $25 one time.

Total cost of ownership: the three-year view

Build cost is not total cost. An app is not a one-off purchase: operating systems update twice a year, dependencies drift, security patches are mandatory, and an unmaintained app degrades within about 18 months. Here is what a $60,000 medium-complexity app costs over three years, using maintenance at 20 percent of build per year and hosting that grows with usage:

The headline from the chart is the one founders most often miss: years two and three add roughly 70 percent to the original build price, before marketing. If you can raise $60,000 to build an app but not another $40,000 to keep it running for two more years, your real budget is smaller than you think.

What actually moves the number

• Number of user roles. A customer app is one build. Customer plus driver plus admin is three. This is the biggest hidden multiplier in marketplace and delivery apps.
• Real-time anything. Live location, chat, presence and streaming all need infrastructure that a simple request-and-response app does not.
• Money between parties. Taking a payment is simple. Splitting it, holding it in escrow and paying it out is not.
• Compliance. In regulated sectors, frequently the largest single line item, and sometimes larger than the build.
• AI model choice. A 40x swing in running cost from one configuration decision.
• Design originality. A custom design system costs several times a well-implemented standard one, and users rarely notice the difference.
• Scale assumptions. Architecting for a million users before you have a thousand is the most common way to burn a budget.

How to estimate your own app cost

The formula agencies use is simple: total hours × blended hourly rate + 15 to 25 percent for design, QA and management. Getting the hours right is the hard part. A practical sequence:

1. Write down every user type. Customer, provider, admin, support. Each one is close to a separate application.
2. List the features per user type. Be specific. "Payments" is not a feature; "customer pays, platform holds funds, provider is paid out weekly" is.
3. Price each feature from the table above. Use the high end of each range for anything real-time or involving money.
4. Add the multipliers. Native instead of cross-platform, compliance requirements, and AI running costs.
5. Add 15 to 25 percent for overhead and 20 percent contingency. Every project discovers scope during the build.
6. Check whether a white label version exists. If one does, compare its price against your total before going further.

Worked example: a food delivery app

Here is the method applied end to end to a food delivery MVP with three user types (customer, courier and restaurant) plus an admin panel, priced from the feature table at $50 an hour:

That lands squarely in the medium tier, and it agrees with our dedicated grocery delivery cost guide, which puts a delivery MVP at $40,000 to $80,000. Two lines in that table deserve attention. Payments and live tracking together are more than a third of the feature cost, which is why delivery apps cost more than they look. And the admin panel, which no customer ever sees, costs twice as much as the entire search experience. At a $25 hourly rate the same scope halves to roughly $22,000 to $45,000; at $100 an hour it doubles to roughly $86,000 to $181,000. The hours stay the same. Only the rate moves.

Our free app cost calculator runs this arithmetic for you from your features, platforms and complexity, and gives a range in a couple of minutes.

Common budget mistakes

• Budgeting only for the build. Three-year cost is about 1.7x the build. Plan for it from the start.
• Choosing on hourly rate. Cheap rates with more hours frequently cost more, and later.
• Leaving out the admin panel. Every app needs one, almost nobody scopes it, and it is around 10 percent of the build.
• Building every feature for launch. Most launch features go unused. Ship the core loop and let users tell you what is missing.
• Ignoring AI running costs. A demo that costs nothing at 100 users can cost thousands a month at 100,000.
• Treating compliance as a later problem. In regulated categories, retrofitting it costs far more than building it in.
• Paying everything upfront. Fixed-price milestones tied to deliverables protect you. Large upfront payments do not.

How to spend less without wrecking it

• Cut scope, not quality. Three features built properly beat ten built badly, and everything you remove saves design, build, test and maintenance cost forever.
• Start cross-platform. Go native later for the specific screens that need it, if any ever do.
• Buy the commodity parts. Authentication, payments, chat and notifications are solved problems. Paying to rebuild them is a straightforward waste.
• Evaluate a white label base seriously. At a small fraction of a custom build, it deserves a real look before you reject it.
• Route AI requests by difficulty. Small models for simple answers, large ones only when needed.
• Prefer subscriptions on Android. Under Google's 2026 fees, subscriptions carry the lowest commission in the US, UK and EU.

Conclusion

If you need one number: a credible custom app costs $45,000 to $120,000, which is where most real projects land once the scope is honest. Below that, you are buying an MVP or a rebranded codebase. Above it, you are building a platform, and the cost is driven by compliance, real-time infrastructure, AI or multiple user roles rather than by the app itself.

The more useful question is not "how much does an app cost" but "which route am I on, and what will it cost to run". A founder validating an idea and an enterprise replacing a core system are both making an app, and the gap between them is two orders of magnitude. Work out your route, budget for three years rather than one, check the 2026 store fees against your business model, and then use the specific guide for your category, every one of which is linked above with its own worked breakdown.

Sources

• Developer salaries: US Bureau of Labor Statistics, Occupational Outlook Handbook, Software Developers (median annual wage, May 2025).
• AI model pricing: Anthropic API pricing, checked 30 October 2026.
• Google Play fees: Google Play Console Help, "Service fees", published at support.google.com.
• Apple fees: Apple's App Store Small Business Program page, published at developer.apple.com/app-store/small-business-program.
• Epic v. Apple: reporting on the Supreme Court declining to pause the contempt order (May 2026).
• All app-type, industry and clone ranges: our own costing and build guides, linked in each table above.