How Much Does It Cost to Build an App in 2026?

Most answers to this question are useless in one of two ways. Either they are a single scary number with no context, or they are a range so wide ($10,000 to $500,000) that it tells you nothing about your own project.

The honest answer is that app cost is a function of four things: how much behavior the app has to encode, who builds it, how you contract for it, and how long you intend to keep it alive. Get those four right and you can predict your number within about 20% before writing a line of code. Get them wrong and no vendor quote will save you.

We have built and shipped products on both sides of this equation. Arhivix is our own document platform, now running for 500+ companies, so we pay its infrastructure and maintenance bills every month. FENIX was a client build where an AI quoting workflow cut quote turnaround from about three days to same day. MaxPlayer is a cross-platform media player we took from seed to launch. That mix is why the numbers below include the costs that only show up after launch, not just the build quote.

The four variables that set your number

Before any range is useful, calibrate on these.

Behavioral surface. Not screen count. Count the number of distinct rules the software has to enforce. A marketplace with one payout rule is cheap. A marketplace with regional tax handling, partial refunds, disputes, and split payouts is not, even though both have roughly the same screens.

Who builds it. The same specification varies by a factor of six depending on where the team sits. This is the largest single lever, and we cover it below with sourced rates.

How you contract. Fixed price, time and materials, and dedicated team produce genuinely different totals for identical scope, because they distribute risk differently.

How long it lives. A build quote covers roughly the first six months of a product’s life. If you plan on three years, the build is somewhere between a third and a half of what you will actually spend.

Cost by complexity, with real ranges

Here is the market as of 2026. These brackets assume a competent team, custom design, and a production launch, not a prototype.

Simple app: $15,000 to $50,000

One user type, one platform, five to twelve screens, authentication, a straightforward backend, one or two third-party integrations (email, analytics, maybe payments through a hosted checkout). No offline mode, no real-time features, no complex permissions.

This is booking tools, internal field apps, content apps, single-purpose utilities, and most first-version MVPs. Timeline for a mobile app in this bracket is typically six to twelve weeks.

Clutch, which aggregates verified buyer reviews rather than vendor self-reporting, puts the majority of mobile app projects in the $10,000 to $49,999 band at $25 to $49 per hour, which lines up with this bracket almost exactly.

Medium app: $50,000 to $150,000

Two or three user types with meaningful permission differences, both mobile platforms plus a web admin, ten to thirty screens, real payment flows, file handling, push notifications, search and filtering, and three to six integrations into systems you do not control.

This is where most funded startups and most serious internal business tools land. Timeline is typically four to nine months. FullStack’s 2026 price guide puts mid-sized custom applications at $100,000 to $500,000, which is the same territory viewed from the higher end of the vendor market.

Complex app: $150,000 to $500,000+

Multi-tenant architecture, offline-first sync, real-time collaboration or messaging, regulated data, audit trails, SSO, hardware or sensor integration, or ML in the critical path. Anything where correctness is a legal or safety question rather than a UX question.

This is the territory of custom software rather than app-shaped products. The cost driver here is rarely feature count. It is that every feature needs to survive adversarial conditions: bad networks, hostile users, auditors, and concurrent edits. Testing alone can approach the size of a whole simple app.

A useful gut check: if you cannot describe your app’s core rules in a single page of plain English, you are in this bracket, whatever your feature list says.

Where the money actually goes

Most quotes present a lump sum. Here is how a typical medium build distributes, roughly, and what happens if you starve each phase.

Discovery and technical architecture: 5% to 10%. Requirements, user flows, data model, integration mapping, stack decisions. Skipping this does not save 8%. It costs you 20% to 40% later in rework, because you discover the real requirements halfway through the build instead of before it.

Design: 10% to 15%. Wireframes, UI, design system, prototype. Changing a flow in Figma costs an afternoon. Changing the same flow after it is built costs a sprint, plus retesting.

Engineering: 45% to 60%. Frontend, backend, integrations, infrastructure setup. This is the part everyone budgets for and the only part most people compare quotes on.

QA and testing: 15% to 25%. If a quote has no explicit QA line, testing is being done by whoever wrote the code, on the same day they wrote it, which is the least effective possible arrangement. Assume the cost is hidden in the engineering number or is not happening.

Launch and stabilization: 5% to 10%. Store submission, monitoring, analytics, crash reporting, the two weeks of real-world bugs that only appear when strangers use the product.

Project management: 10% to 15%, usually distributed across the above rather than itemized.

Team geography: the biggest lever

Same specification, different postcode, wildly different invoice.

Accelerance surveys global outsourcing rates annually. Its 2026 data puts senior developers at $64 to $76 per hour in Central and Eastern Europe, $60 to $75 in Latin America, and $31 to $41 in Asia, with rates down 4.4% year over year in CEE and 7.1% in LATAM. Those are broker-network rates. Direct engagement with a studio typically lands lower.

At the other end, FullStack’s guide puts enterprise-class consultancies at $400+ per hour, big-business firms at $250 to $350, mid-market firms at $120 to $250, and small firms at $90 to $160.

The underlying reason is salary arbitrage, not talent arbitrage. Stack Overflow’s 2025 developer survey reports median backend developer salaries of $175,000 in the United States, $108,913 in the United Kingdom, $87,011 in Germany, and $22,086 in India. A team in Belgrade using the same frameworks, reading the same documentation, and reviewing code against the same standards costs less because rent in Belgrade costs less.

The honest caveat: geography sets your rate, not your outcome. A $30 per hour team that takes three times as long is more expensive than a $90 per hour team that gets it right. Accelerance makes the same point in its own data release, noting that hourly rate is a poor proxy for total cost once you account for rework and delay. If you are weighing regions, we go deeper in nearshore vs offshore development and in our guide to software development companies in Eastern Europe.

Engagement model changes the total

Fixed price buys budget certainty and costs you a risk premium, typically 20% to 30%, because the vendor is pricing in everything that might go wrong. It works when scope is genuinely frozen. It works badly for anything discovery-shaped, because every change becomes a contract negotiation instead of a conversation.

Time and materials is cheaper on average and riskier on any single project. You pay for what happens. It requires you to be an active participant, because you are the one deciding what gets built next.

Dedicated team (a fixed monthly cost for a fixed set of people) makes sense past roughly six months of continuous work. You stop paying per-project overhead and get a team that accumulates context, which is the thing that actually makes month nine cheaper than month one.

Our recommendation for most first builds: time and materials with a hard budget ceiling and a milestone review every two weeks. You keep the flexibility to change your mind, and the ceiling keeps that flexibility from becoming a blank check.

Agency, freelancer, or in-house

The honest version of this tradeoff, with no sales spin.

Freelancers are the cheapest hourly and the highest variance. FullStack’s guide puts experienced freelance developers at $100 to $300 per hour and newcomers at $30 to $60. For a simple app with a clear spec and a technical founder who can review the work, a good freelancer is genuinely the right answer. The failure modes are real though: no design capability, no QA, no cover when they get sick or take a better offer, and no one to hand the code to afterwards.

Agencies cost more per hour and less per outcome, when they are good. You are buying a team that already knows how to sequence a build, plus continuity. You are also buying overhead: sales, management, and bench time are in your rate. The risk is bait and switch, where you meet the seniors and get the juniors.

In-house is the most expensive to start and the cheapest at steady state, if the product is core to your business. Below roughly two years of continuous roadmap, the hiring cost, ramp time, and management burden usually do not pay back.

We wrote a longer comparison in freelancers vs agency vs Toptal if you are actively choosing.

Hidden costs nobody puts in the quote

Store fees and platform commissions. The Apple Developer Program is $99 per year and a Google Play developer account is a one-time $25. Those are trivial. The commissions are not: Apple’s Small Business Program charges 15% for developers under $1,000,000 in annual proceeds and the standard rate above that, and Google Play applies 15% on the first $1,000,000 in annual earnings in most markets, with a restructured service-fee-plus-billing-fee model rolling out in the EEA, UK, and US. If your revenue model runs through in-app purchases, price this into your unit economics before you price the build.

Infrastructure. $50 to $500 per month for a small-to-medium app, considerably more once you have real data volume, redundancy, or compliance requirements.

Third-party SaaS. Email, monitoring, error tracking, CDN, auth. Individually $20 to $300 per month, collectively $200 to $2,000.

Maintenance. Budget 15% to 25% of the build cost annually. FullStack’s guide uses the same range. This is not optional spend: it is dependency upgrades, OS version compatibility, security patches, and the small fixes that keep the app from quietly rotting.

Post-launch iteration. Your v1 is a hypothesis. Reserve at least three to six months of runway for acting on what real users tell you. This is often the highest-return money in the entire project.

AI assistance is not the discount it looks like. Stack Overflow’s 2025 survey found 84% of developers using or planning to use AI tools, but only 3.1% highly trust the output, 66% cite solutions that are “almost right, but not quite,” and 45.2% report that debugging AI-generated code takes more time than expected. AI has compressed boilerplate. It has not compressed architecture, integration, or the debugging of subtle wrongness. Treat any quote that prices in a large AI discount with suspicion.

How to cut cost without killing the product

Cut user types before you cut features. Every additional role multiplies permission logic, testing paths, and edge cases. Launching with one user type and adding the second later is often a 30% saving.

Ship web first when you can. A responsive web app reaches everyone, deploys without store review, and avoids paying for two platforms. Add native when you actually need the camera, the sensors, or offline. We compared the tradeoffs in PWA vs native mobile apps.

Buy the commodity parts. Authentication, payments, notifications, file storage, and search all have mature hosted options. Building them yourself is engineering effort spent on something no user will ever notice.

Manual-first the expensive automation. If a workflow will run twenty times a month at launch, do it by hand and automate it when volume proves it matters. For Srecna Trojka we shipped a working AI RFP system in six weeks by sequencing hard, and bid preparation dropped from about 18 hours to 6 with win rate up 14 points.

Do not cut discovery, design, or QA. These are the three lines people cut first and the three that cost the most to cut. Cutting scope is a saving. Cutting rigor is a loan at a bad interest rate.

What this looks like at Notix

Our minimum project size is $10,000, our rates sit in the $25 to $49 per hour band, and a focused MVP typically ships in about six weeks. Those three numbers together define what we can honestly do: a well-scoped first version of a real product, not a twelve-module enterprise platform on a starter budget.

For anything complex, we run a paid discovery phase first, usually $2,000 to $5,000, which produces an architecture document, a feature-level effort breakdown, and a phased plan with milestone costs. That is deliberately not free, because a free estimate is a guess and a paid one is a document you can budget against. If you want the deeper breakdown by project type, our software development cost guide goes further, and the MVP cost guide covers first versions specifically. If you are still shaping the idea, start with the MVP development guide.

The most useful thing you can do next

If you have read this far, you probably have a number in your head and a project that does not quite fit any of the brackets above. That is normal. Every real app sits between two rows of a table like this one.

Send us the shape of what you want to build, even roughly. We will tell you which bracket it actually falls into, what would push it up a bracket, and what would pull it down, including when the honest answer is that a freelancer or an off-the-shelf tool would serve you better than we would. We would rather lose that project than quote you into one that was never going to work.

Book a free consultation and we will give you a real range, not a range so wide it means nothing.