Skip to content

How much does it cost to build an AI app?

An AI app has two price tags: what it costs to build, and what it costs every time someone uses it. The second one decides whether the business works, so this guide covers both. Four of the five apps we run ourselves are AI apps, so this is the arithmetic we do for our own products too.

Pull House · 8 min read · Updated

Book a 30-min call

The short answer

Building an AI iPhone app costs about what any app of the same size costs: our projects typically range from $10,000 to $100,000+, with a fixed price after discovery. On top of that comes a running cost per request, which depends almost entirely on where the model runs: a hosted API, your own server, the device itself, or Apple’s Private Cloud Compute.

What an AI app adds to the build

The model call is usually a small part of the work. These are the parts that take time:

  • The wait. A generation takes seconds. Progress, retries when a request fails, and results that are still there when someone comes back to the app all need designing and building.
  • Moderation. Guideline 1.2 treats AI content creation apps like user-generated content: filtering, a way to report content, a way to block abusive users, and published contact details.
  • Consent. Guideline 5.1.2(i) requires a clear disclosure and explicit permission before personal data, such as a photo, goes to a third-party AI service.
  • Payments. Credits or a subscription, sold through in-app purchase, with restore and receipts handled properly.
  • Quality checks. Prompts and outputs change as models change. iOS 27 adds an Evaluations framework for testing AI features beyond unit tests.

Running costs: where the model runs

Where the model runsCost per requestTrade-offs
A hosted API (OpenAI, Anthropic, Google and others)Billed by the provider, per request or per tokenThe strongest models and the fastest start. Costs grow with use, and personal data leaves the device, so 5.1.2(i) applies.
Your own serverHosting, whether it is used or notFull control over the model and the data, plus the work of running it.
On the device (Foundation Models, Core AI, Core ML)NonePrivate and works offline. Limited by model size and by the device; Apple’s model needs Apple Intelligence to be available.
Apple’s Private Cloud ComputeNo cloud API cost for eligible appsApple’s next-generation models, for Small Business Program members with fewer than 2 million first-time downloads.

The last two rows changed the arithmetic in 2026. According to Apple, apps in the App Store Small Business Program with fewer than 2 million total first-time downloads can use the next generation of Apple Foundation Models on Private Cloud Compute at no cloud API cost. And the Foundation Models framework can now also call third-party models such as Claude and Gemini through the same Swift API (iOS 27). More in our guide to on-device AI.

Pricing so every request pays for itself

Our own AI apps earn in two ways: credits spent per generation, and a Pro subscription. Credits suit apps where each request has a clear cost, such as image generation. A subscription suits apps people use often and where the cost per use is small or predictable.

  • Apple’s share comes first. On digital sales through in-app purchase, Apple keeps 15% for members of the Small Business Program and for subscriptions after their first year, and 30% otherwise (subscriptions).
  • Then the request. What you keep from one credit has to cover what one generation costs you, including the ones that fail and are retried.
  • Credits cannot expire. Credits bought through in-app purchase may not expire under guideline 3.1.1, so a pack priced today has to make sense for as long as people hold it.
  • Free tiers get abused. A free allowance needs limits per account and per device, or it becomes your largest cost.

Costs people forget

  • Failed generations. Timeouts and retries cost money and earn nothing.
  • Model changes. A provider updating or retiring a model can change your results, and your prompts need checking again.
  • Moderation at scale. Reports have to be read and acted on, by people or by tools.
  • Yearly iOS updates, the same as any app. See what it costs to maintain an iOS app.

Ways to spend less

  • Validate with a hosted model, then move. Start with an API to learn what people use, then move the frequent, simple requests on to the device.
  • Use Apple’s models where they are good enough. Summaries, extraction, tagging and short text often are, at no cost per request.
  • Check Private Cloud Compute eligibility before you budget for a cloud API.
  • Cache what repeats. The same prompt with the same input should not be paid for twice.

Related: AI app development · On-device AI on iPhone · How much an iOS app costs

FAQ

Quick answers

Not covered here? Write to hello@pull.house.

  • About what any app of the same size costs: our projects typically range from $10,000 to $100,000+, with a fixed price after discovery. Model and server costs per request come on top.

Start a project

Have an iOS app to build?

Tell us what you are making. If we are not the right team for it, we will say so.

Book a 30-min call
hello@pull.house