How Much Does It Cost to Buildan App Like Spotify in 2026?

In 2025 Spotify booked about 17.2 billion euros of revenue and spent about 11.7 billion of it on cost of revenue, which its own annual report says is predominantly royalty and distribution costs. That's 68 cents of every euro before a single engineer gets paid.

Hold that number in your head, because it changes the question. When people ask what it costs to build an app like Spotify, they mean the software. The software is the smaller bill.

So I'll answer both. I'll price the app feature by feature in hours at our real rate of $150 to $225 an hour, and show every multiplication. Then I'll lay out the licensing stack, marking which prices are published by a rate court or a price page and which ones nobody outside a label deal will ever see.

The short version: my lean v1 scenario lands at about 1,140 to 1,710 hours, which at $150 to $225 is $171,000 to $384,750. A fuller launch is about 1,940 to 2,950 hours, or $291,000 to $663,750. Those are my scenario estimates, not quotes and not measurements, and they exclude music rights.

To be clear, I'm not a lawyer and none of this is legal advice. Spotify is a trademark of its owner, and Frenchy Digital has no affiliation with it. I use it here because it's the product people name when they describe what they want.

The ranges you find on agency blogs can't be checked, because none of them show hours, a rate, or which features carry which effort. I pulled three that rank for this topic.

Look at the spread. The low end of one page is $40,000; the high end of another is $400,000 and up. That's a factor of ten for what is supposedly the same product.

Why does that happen? Because a range with no inputs is a mood, not an estimate. If I tell you a bag of groceries costs between $40 and $400, I'm right, and I've told you nothing. The useful version is the receipt: this item, this price, this total.

One of those pages also says to budget $50,000 to $500,000 a year for licensing commercially released recordings. I looked for a source behind that and found none. Label terms are private (I'll show you Spotify's own filing saying as much), so any dollar figure for them is a guess wearing a suit.

I'm citing these pages as examples of the claim, not as fact. The same goes for the music streaming market size and growth percentages that circulate on these pages; I didn't find a primary source I'd print, so you won't see one here.

Therefore the method in this article is the receipt. Features first, from Spotify's own documentation. Hours per feature, labelled as my estimate. Rate applied in the open. Running costs from price pages you can open yourself. The same method runs through the rest of this series, including what an app like Netflix costs, where video delivery changes the arithmetic a lot.

Spotify, as a user sees it, is a player with quality tiers, offline downloads, playlists, search, recommendations and a few subscription plans. Its help center and product pages spell out enough detail to price each piece.

A few details from that table matter more than they look.

Quality tiers mean a transcoding pipeline.Spotify's audio quality page lists approximate settings of 24, 96, 160 and 320 kbps in its apps, AAC at 128 or 256 kbps on the web player, and lossless up to 24-bit/44.1kHz FLAC on Premium. Every tier you offer is another rendition of every track you store.

Offline has a royalty reason.Spotify lets Premium users download up to 10,000 tracks on each of up to 5 devices, and requires them to come online at least once every 30 days. Its help page says that check-in exists so Spotify can collect play data to compensate artists. That's the part clones forget: offline plays still have to be counted and reported.

Recommendations start with similarity search.Spotify's engineering blog describes Voyager, an open-source nearest-neighbor search library that replaced its older Annoy library and lets it recommend similar tracks, artists or albums without running expensive models in real time. The speed and accuracy multipliers in that post are Spotify's own claims; the design idea is the useful part.

Plans are a pricing engine.In the US, Spotify lists Individual at $12.99, Student at $6.99, Duo at $18.99 and Family at $21.99 a month, and some plans bundle audiobook listening hours. Each plan type is entitlement logic, and bundles matter for royalties too, which I'll get to.

What a v1 does not need: podcasts, audiobooks, lossless, a family plan, social features, device handoff. Those are years of product work built for a company with hundreds of millions of users. You're building for your first ten thousand.

The Spotify you're comparing against is run by an average of 3,787 research and development employees, according to its FY2025 Form 20-F. The company averaged 7,287 full-time employees in total.

Research and development expense was 1,393 million euros in 2025, 8% of revenue. Divide that by 365 and Spotify spent roughly 3.8 million euros on R&D per day.

Put my fuller launch scenario next to that. Its top end, $663,750, is well under a quarter of one day of Spotify's 2025 R&D spend, even allowing for the euro and dollar not being equal. Does that mean the scenario is too small? No. It means "an app like Spotify" is the wrong unit. You're building an app that does a few of the things Spotify does, for a narrower audience.

The filing also shows how concentrated the music supply is. Content licensed from Universal, Sony, Warner and Merlin accounted for about 72% of streams of audio content delivered by record labels on Spotify in 2025. Merlin represents the digital rights of hundreds of independent labels.

And it shows the scale of the audience: 751 million monthly active users and 290 million Premium subscribers at the end of 2025. Spotify itself notes those metrics come from internal data and have not been validated by an independent third party.

The cost of revenue section is where the annual report gets useful for a founder, because it shows how a music bill moves. Spotify says subscription music royalties are generally based on the greater of a percentage of relevant revenue and a per user amount, while ad-supported royalties are typically a percentage of revenue, with some deals using a per user or per stream floor instead. Read that twice. It means a label's cut doesn't shrink just because you discount the plan; the per user amount catches you.

The segment numbers show the same thing from the other side. Premium cost of revenue was 10,184 million euros in 2025, and the filing says it fell from 67% to 66% of Premium revenue. Ad-supported cost of revenue was 1,506 million euros, and it fell from 88% to 82% of ad-supported revenue. Think about that second figure. Out of every euro of advertising revenue on the free tier, 82 cents went back out as cost of revenue before anyone was paid to build the product.

And the filing breaks down what drove the Premium increase: higher royalty costs of 765 million euros (from revenue growth, increased rates for certain licensors, audiobook licensing and a partner program), a 34 million euro rise in payment processing fees, and a 23 million euro rise in streaming delivery costs. Delivery grew by less than a thirtieth of what royalties grew. That ratio is the whole article in one line: the pipes are cheap, the rights are not.

My lean v1 scenario is about 1,140 to 1,710 hours of senior work for native-feeling iOS and Android apps built in one cross-platform codebase, plus a backend and admin tool. Every hour figure below is my scenario estimate, not a quote and not a measurement of any past project.

Suppose a founder has a defined catalog: say an independent label group, a genre library, or a catalog of production music they own or have licensed directly. That's the realistic starting point, because it removes the part of the problem that no amount of engineering solves.

Let me add it up out loud. Low column: 60 + 120 + 80 + 160 + 80 + 100 + 100 + 120 + 120 + 80 + 120 = 1,140 hours. High column: 90 + 180 + 120 + 240 + 120 + 150 + 150 + 180 + 180 + 120 + 180 = 1,710 hours.

At $150 an hour, 1,140 hours is $171,000. At $225 an hour, 1,710 hours is $384,750. So the lean v1 is about $171,000 to $384,750.

Why is the player so big? Because background audio, lock screen controls, headphone button handling, interruptions from calls and a queue that survives the app being killed are each small, and together they are most of what makes an audio app feel real. The player is the product.

Why is royalty reporting in a v1 at all? Because it's the feature regulators and rights holders actually read. If you stream anyone else's music, you owe usage reports, and reconstructing plays after the fact is miserable. Logging every play (online and offline, with the plan the listener was on) from day one is cheap insurance.

Subscriptions get their own line because you're integrating three billing systems: Apple, Google and a web checkout. The entitlement logic that decides what a listener may play has to agree across all three. Our write-up on implementing in-app purchases goes through the receipt validation and server notifications in detail.

A fuller launch adds about 800 to 1,240 hours to the lean v1, for a total of about 1,940 to 2,950 hours. Again, every figure is my scenario estimate.

The arithmetic: 200 + 80 + 80 + 160 + 120 + 160 = 800 added hours at the low end, and 320 + 120 + 120 + 240 + 200 + 240 = 1,240 at the high end. Add those to the lean totals: 1,140 + 800 = 1,940 and 1,710 + 1,240 = 2,950.

At $150, 1,940 hours is $291,000. At $225, 2,950 hours is $663,750. So the fuller launch is about $291,000 to $663,750.

Notice the overlap in that table: the top of the lean v1 at $225 ($384,750) sits above the bottom of the fuller launch at $150 ($291,000). That's normal. Rate and scope both move the total, and a small scope at a higher rate can cost more than a larger scope at a lower one.

The ad-supported tier is the sneaky one.It looks like "just add ads." It's actually a second product: ad insertion between tracks, frequency caps, advertiser reporting, and a second set of royalty calculations, because free listeners generate revenue differently from subscribers. Spotify's own filing shows the difference: 34% gross margin on Premium against 18% on ad-supported in 2025.

Recommendations are cheaper than people fear, at first.A first version is embeddings for tracks plus a nearest-neighbor index, the same basic idea Spotify describes with Voyager. It will be worse than Spotify's. It will be much better than nothing, and it's where I'd spend after the player works.

Here's the downside of scenario two, said plainly: it roughly doubles the engineering bill without adding a single licensed track. If your catalog is the reason people come, spend on catalog before you spend on a web player.

An on-demand music app in the US needs rights to two different copyrights in every song, and they come from different places with different price transparency. The sound recording is the master (the actual audio). The musical work is the composition (melody and lyrics) owned by songwriters and publishers.

One more consequence for the build: rights are per track and per territory. Your catalog tool needs availability windows, territory flags and fast takedowns. That's already in the lean v1's catalog and admin lines, and it's why I gave them real hours.

Delivering audio is cheap at startup scale: about $1,826 a month for infrastructure in my 10,000-subscriber scenario, using Amazon's published prices. Here's how that number is built.

The scenario: 10,000 paying subscribers, each listening 30 hours a month, at about 160 kbps (Spotify's "High" setting). Those usage figures are my assumptions.

Bandwidth first. 160 kilobits per second is 20 kilobytes per second. Times 3,600 seconds is 72 MB an hour. Times 30 hours is 2,160 MB, or about 2.16 GB per subscriber per month. Times 10,000 subscribers is 21,600 GB a month.

CloudFront's pay-as-you-go page lists 1 TB a month free, then $0.085 per GB for the next 9 TB and $0.080 for the next 40 TB in the US and Europe. So: the first 1,000 GB are free; the next 9,000 GB cost $765; the remaining 11,600 GB cost $928. Delivery is about $1,693 a month.

Requests are a small extra. If audio is served in 10-second segments (my assumption), 30 hours is 10,800 segments per subscriber, about 108 million across 10,000 people. CloudFront includes 10 million HTTPS requests and charges $0.01 per 10,000 after that, so 98 million extra requests cost about $98.

Storage is smaller still. Assume 100,000 tracks averaging 3.5 minutes, stored at 96, 160 and 320 kbps. That's 576 kbps combined, times 210 seconds, which is about 15.1 MB per track, or about 1,512 GB for the catalog. At S3's published $0.023 per GB-month, storage is about $35 a month.

CloudFront also lists flat-rate plans, including a $15 a month Pro plan with 50 TB of data transfer and a $1,000 Premium plan, advertised with no overage charges. I haven't reviewed their terms for an audio streaming workload, so I priced the pay-as-you-go option. If the flat plans fit your use, delivery could be cheaper still.

This table leaves out compute, databases, monitoring, email and error tracking. Those depend on architecture choices and are usually smaller than delivery for audio. The big lesson is the ratio: infrastructure is a four-figure monthly bill, while licensing, as the next section shows, scales with every dollar of revenue.

Per subscriber, the published costs you can actually calculate are app store or card fees plus the musical works royalty; the label line stays blank because its price is not public. Here's the arithmetic at Spotify's US Individual price of $12.99, which I'm borrowing as the scenario price.

• Revenue: 10,000 subscribers x $12.99 = $129,900 a month.
• Musical works, all-in: the greater of 15.3% x $129,900 = $19,874.70 and the content cost prong, which for a portable subscription cannot exceed $1.10 x 10,000 = $11,000. So the all-in is $19,874.70, assuming service provider revenue equals the subscription price (the regulation defines revenue in detail I have not modelled).
• Of that, the mechanical to the MLC is the all-in minus your PRO fees, but never less than the floor of $0.60 x 10,000 = $6,000. So PRO fees up to $19,874.70 minus $6,000 = $13,874.70 sit inside the all-in rather than on top of it.
• Web signups through Stripe: 2.9% x $12.99 = $0.377, plus $0.30, is about $0.68 per subscriber, or about $6,767 a month if everyone pays on the web.
• iOS signups under Apple's Small Business Program: 15%, so about $1.95 per subscriber.
• Sound recording royalties: not publicly disclosed.
• Public performance license rates: not publicly disclosed, but in this scenario credited against the all-in above.
• Delivery and storage from the previous section: about $1,826, or about $0.18 per subscriber.

Now the trap in the Apple line. The Small Business Program rate applies only while proceeds stay at or under $1 million in the prior and current calendar year; cross it, and Apple's page says the standard 30% applies to future sales. If all 10,000 subscribers paid through iOS for a year, gross is $129,900 x 12 = $1,558,800, and proceeds at 85% are about $1,324,980. That crosses the line within the year.

Google Play is changing too. Its help page lists 15% for auto-renewing subscriptions in markets still on the old schedule, and for the US, UK and EEA from June 30, 2026, a 10% service fee plus a 5% billing fee on auto-renewing subscriptions.

So on the lines we can see, a $12.99 subscriber costs roughly $2 in musical works royalties (mechanical and performance together), $0.68 to $1.95 in payment fees, and under $0.20 in delivery. The lines we can't see are the big ones. Spotify's 68% cost of revenue is a reference point for how large they can get at scale, not a rate you'll be offered.

This is also why the total cost of an app is mostly what happens after launch, a point our guide to app total cost of ownership makes for every category. Music is the extreme case.

Year two costs about $25,650 to $76,950 in maintenance labor in my planning assumption, plus infrastructure that scales with listening and licensing that scales with revenue.

The labor figure is my planning assumption of 15% to 20% of the lean v1's build hours per year, not an industry benchmark. 15% of 1,140 hours is 171 hours, and at $150 that's $25,650. 20% of 1,710 hours is 342 hours, and at $225 that's $76,950.

That work isn't optional. Apple and Google ship major OS releases every year, audio session behavior changes, billing APIs move, and a music app lives in the background where OS changes bite hardest.

Infrastructure grows faster than you'd expect. Take 50,000 subscribers on the same usage assumptions: 50,000 x 2.16 GB = 108,000 GB a month. On CloudFront's tiers: 1,000 GB free; 9,000 x $0.085 = $765; 40,000 x $0.080 = $3,200; the remaining 58,000 x $0.060 = $3,480. That's about $7,445 a month for delivery, five times the listeners for about 4.4 times the bill.

Licensing grows in straight proportion. At 50,000 subscribers paying $12.99, monthly revenue is $649,500 and the musical works all-in is 15.3% of that, $99,373.50 (the $1.10 content cost cap would be $55,000, lower, so the percentage governs). The published rate doesn't get cheaper with scale; the per-GB delivery rate does.

The first thing to break in a new music app is usually play counting, and the second is background playback on specific phones. Both are invisible in a demo.

Most founders who say "Spotify for X" should launch one of three smaller products first, each chosen by the license it needs rather than the features it copies. These are scenarios, framed as scenarios, with the arithmetic from the tables above.

Which one is right? The one whose license you can actually get this quarter. A beautiful player with nothing licensed to play is the most expensive demo you'll ever build, like opening a restaurant before you've found a supplier.

The clearest red flag in an agency quote for an app like Spotify is a total that arrives before anyone asks whose music you'll stream. Licensing decides the architecture, so a quote that skips it is a quote for a different product.

• A dollar range with no hours, no rate and no feature list you can cross items off. You can't check it, so you can't negotiate it.
• A ready-made clone with a music catalog included, from a vendor who can't name the license behind it.
• A yearly licensing budget for major label music printed without a term sheet in hand. Those terms are not publicly disclosed.
• A per-stream royalty presented as a fixed price per play. Spotify's filing describes royalties as the greater of a revenue share and a per user amount, not a price per stream.
• Mechanicals and PRO fees added on top of each other at full rate, which misreads the Phonorecords IV all-in.
• No line for play logging, MLC usage reports or offline play sync in the first release.
• No separate line for app store and web billing entitlements, which is three integrations, not one.
• A music streaming market size or growth figure with no primary source, used to justify the budget.

None of these mean the agency can't code. They mean it hasn't priced the part of this business that costs the most. Ask for the hours table, and if they won't send one, you have your answer.

A real licensed streaming product exceeds our starter packages, and I'd rather say that here than in a proposal. Our MVP development packages are published at $15,000 to $25,000, $30,000 to $50,000, and $55,000 to $75,000 and up. My lean v1 scenario is $171,000 to $384,750.

So where does a package fit? At the front. A validation prototype on a small owned catalog can test whether people will pay for your niche before you negotiate a single license. That's the cheapest way to find out the thing that matters.

For the full build we price in hours at $150 to $225, senior-led, and send a fixed-price phased proposal after discovery. We'd usually build the listener apps in React Native so iOS and Android share one codebase, with native modules where audio needs them. Full source code and IP transfer to you on full payment, and there's a 30-day post-launch warranty.

We've been building since 2016 (founded in France, a US company since 2019), from Los Angeles. We have not published a music streaming case study, and I won't imply one. The closest published work is LIVV Audio, a headphone control app with native iOS and Android apps, a NestJS backend and a React admin dashboard.

If you're on the creator side rather than the listener side, our roundup of music production apps covers the tools artists use before a track ever reaches a streaming catalog.

Several numbers people expect in this article are missing because I couldn't source them, and I'd rather leave a blank than fill it.

• Label royalty rates, advances and minimum guarantees for on-demand streaming: not publicly disclosed.
• ASCAP, BMI, SESAC and GMR rates for on-demand streaming: not publicly disclosed in anything I found.
• How the Phonorecords IV definition of service provider revenue applies to your exact plan and bundle; I used the rate table and steps in 37 CFR 385.21 and treated revenue as the subscription price.
• Whether final 2026 rates for commercial webcasters (as opposed to commercial broadcasters) have replaced SoundExchange's interim rates.
• The status of the MLC's amended complaint against Spotify after October 1, 2025.
• Terms of CloudFront's flat-rate plans for audio streaming workloads.
• Per-stream payout figures that circulate online. Spotify's filing describes royalties calculated from many variables, not a per-stream price.
• Spotify's $11 billion 2025 payout figure is the company's own statement in its newsroom.
• Every hour estimate in this article is my scenario, not a measured project.

Do these three this week, before you talk to any developer, including us.

• Write down whose music you will stream and how you will get the rights to it: owned, direct license, distributor, or radio-style statutory license. If you can't fill that in, stop there.
• Take the lean v1 table above and cross out every row you don't need. Add up what's left and multiply by $150 and by $225. That's your honest range.
• Pick a subscriber count and listening hours, and redo the CloudFront arithmetic with your numbers. Then add 15.3% of revenue for musical works (mechanical and performance together) and leave the label line blank until you have a term sheet.

If your product leans toward video rather than audio, the same method applied to an app like TikTok shows how much faster delivery bills grow.

Time to get to work.