Video platform server cost estimation guide calculating infrastructure expenses for 2025

Key takeaways

• Egress dominates the bill. For almost every video platform, outbound traffic is 70–95% of your video streaming server cost. Pick the CDN / cloud first and everything else after.

• A clean cost model has four lines. Egress, storage, compute (origin + transcoder + signalling), and third-party (CDN, DRM, analytics). More than four categories in your estimate usually means you’re double-counting.

• AWS is no longer automatically the expensive option. Pay-as-you-go CloudFront still lists $0.085/GB after a 1 TB free tier, but its flat-rate plans (launched November 2025) cover a 50 TB monthly allowance for $15–$200. Bunny lists $0.005–$0.01/GB, Hetzner dedicated servers include unlimited traffic on a 1 Gbit port, and DigitalOcean bills $0.01/GiB beyond the pooled Droplet allowance (November 2026 list prices).

• Architecture choices move cost more than provider choices. Adaptive bitrate, H.265/AV1 instead of H.264, P2P for 1:1 calls, an SFU instead of an MCU, and aggressive CDN TTLs routinely cut total bills by 40–70%.

• Model per-user-hour cost, not monthly totals. Whether you charge per seat, per class or per lesson, the unit economics come down to cost per active user-hour. That’s the number investors and CFOs look at.

Updated November 2026: CloudFront flat-rate request caps and Premium levels ($1,000–$10,000), a new video hosting pricing comparison with Vimeo’s October 2026 plans, and corrected Hetzner AX42 and live-stream figures.

Founders and CTOs routinely get blindsided by their first real video bill. A free tier hides the real cost, the product launches, the egress meter spins, and 90 days later there’s a $50,000 surprise on the AWS invoice. This guide turns “how much will my video platform cost to run” from a guess into a simple model you can put in a spreadsheet. It covers the four cost lines that actually exist, November 2026 list pricing from every provider we use in production, a cost-by-audience table, three worked examples (9-person conference, webinar, VOD catalogue), and the architecture levers that usually matter more than the cloud choice.

Target audience: founders, CTOs, product leads and engineering managers who are scoping a video product, pricing it, or trying to get an existing bill under control. Every number below is conservative and tied either to a public price list or to a deployment we have shipped. When numbers move fast, we price the discovery sprint first and let real measurements replace estimates.

Short answer: cost at a glance

A video streaming server costs about $50–$300 a month at MVP stage, roughly $100–$900 a month for a daily 50-viewer webinar (~10 TB of egress), and $450–$6,400 a month for a 90 TB VOD library, at November 2026 list prices. Almost all of the spread inside each range comes from the CDN: egress is 70–95% of the bill, and storage plus compute together rarely pass 30%.

The table prices delivery only, for three audience sizes. It assumes 720p H.264 with an ABR ladder (~1.35 GB per viewer-hour plus a 1.25× ladder uplift, so ~1.7 GB), North America/Europe rates, and 6-second HLS segments.

Viewer-hours / monthEgress (720p ABR)CloudFront pay-as-you-goCloudFront flat-rateBunny CDNMux delivery (720p)
1,000~1.7 TB~$60$15 (Pro)~$17$0 (inside 100K free min)
10,000~17 TB~$1,325$15 (Pro) for VOD; $200 (Business) for live~$170~$400
100,000~170 TB~$10,800$3,500 (Premium, 200 TB level)~$850–$1,700~$4,700

Quick formula. Monthly egress cost = viewer-hours × GB per viewer-hour × 1.25 (ABR uplift) × price per GB. Add 5–30% for compute, storage and third-party services, and you have a first budget you can defend. The rest of this guide shows where each of those numbers comes from.

Live vs VOD. Live streaming CDN pricing per GB matches VOD, so per viewer-hour, live and VOD egress cost the same. Live costs more in total: ingest, transcoding and the origin run for the whole event whether 10 or 10,000 people watch, and nothing can be pre-encoded or cached ahead of time. Budget a live event’s compute as fixed per streamed hour and its egress as variable per viewer-hour. Live also sends about twice as many CDN requests as VOD, because players reload the playlist with every segment. 10,000 viewer-hours of 6-second live HLS come to roughly 12M requests, which is past the 10M included in CloudFront’s Pro flat-rate plan.

Reach for CloudFront’s flat-rate plan when: you deliver HLS or DASH under ~50 TB a month from one distribution and want AWS in the stack. AWS says the allowances are not hard limits and there are no overage charges, but sustained usage well above the allowance can get delivery adjusted (fewer or more distant edge locations). Pro includes 10M requests a month and Business 125M, and AWS lets the first spike up to 3× the allowance pass without any effect on service. Real-time logs and a few legacy features need pay-as-you-go.

Why Fora Soft wrote this playbook

Fora Soft has spent 20+ years building real-time video, streaming and AI products. Our video and audio streaming software development team has delivered 250+ of them since 2005, mostly in video conferencing, scalable live streaming and VOD. We run production workloads on AWS, GCP, Azure, Hetzner, DigitalOcean, Bunny and Cloudflare, across WebRTC, HLS, CMAF and custom SFU stacks, so the video streaming server cost math in this guide is the same math we run before every build. We built BrainCert, a WebRTC classroom platform at $3M ARR with 100K+ customers and 500M+ classroom minutes; Worldcast Live, HD live concert streaming to 10,000+ concurrent viewers; and dozens of other video products where the infra bill had to be predictable to within ~10% before launch.

This playbook is the simplified spreadsheet we hand to clients during a first scoping call: the four cost lines, the 2026 price ranges that actually matter, the architecture decisions that move them, and three worked examples. If you want it tailored to your product, book a call at the end. With our Agent Engineering workflow (AI agents amplifying our senior engineers), the model comes back fast, the numbers are conservative, and we’ll tell you honestly where building in-house beats buying managed.

Need a real server-cost model for your video product?

Share your expected active users, concurrent streams and recording policy. In 30 minutes we’ll turn that into a monthly bill range, per-user-hour cost and the cheapest provider mix for your architecture.

Book a 30-min call → WhatsApp → Email us →

The four-line cost model every video platform shares

Every video platform bill decomposes into four cost lines, whether it’s a conference, webinar, VOD library, streaming OTT, live classroom or security VMS. Once you know these four, you can model any video streaming server cost in an afternoon, from an MVP to a 10,000-viewer VOD library.

Cost line What it is Typical share of bill Main lever to reduce
Egress / bandwidth Outbound traffic from servers / CDN to viewers 70–95% CDN choice, ABR, codec, P2P
Storage Recorded clips, VOD masters, transcoded renditions 2–10% Retention policy, cold tiers, codec
Compute Origin, SFU, transcoder, signalling, API, DB 5–20% Autoscale, spot, ARM, right-sizing
Third-party services CDN, DRM, transcode SaaS, analytics, observability 2–15% Volume commits, buy vs build

Rule of thumb: if your early model says storage or compute is the dominant cost, something is wrong. Either the product barely streams video, or you haven’t modelled egress honestly.

Four-line video platform cost model: egress is 70-95% of the bill, then compute, storage and third-party services

Figure 1. Every video platform bill decomposes into four lines, and egress dwarfs the rest.

Line 1: Egress — the number that sinks startups

Inbound traffic is free almost everywhere. Outbound (egress) is where every big cloud makes its money. Start by pinning down two numbers for your product: GB per viewer-hour and concurrent streams in the peak hour.

Profile Codec Bitrate GB / viewer-hour
360p mobile H.264 ~0.8 Mbps ~0.36 GB
480p SD H.264 ~1.5 Mbps ~0.68 GB
720p HD H.264 ~3 Mbps ~1.35 GB
1080p FHD H.264 ~5 Mbps ~2.25 GB
1080p FHD, H.265 H.265 ~3.3 Mbps ~1.5 GB
4K H.265 / AV1 ~15 Mbps ~6.75 GB

Multiply GB/viewer-hour by total hours watched per month and by an ABR uplift factor (~1.2–1.3, since the lower rungs get served too), and you have monthly egress. Then pick a provider.

Provider Egress price / GB (2026) Notes
AWS CloudFront 1 TB/month free, then $0.085 (to 10 TB) → $0.08 → $0.06 → $0.04 (150–500 TB) → $0.03 → $0.025 → $0.02 (over 5 PB) Flat-rate plans since Nov 2025: Pro $15 / Business $200 for 50 TB (10M / 125M requests); Premium $1,000–$10,000 for 50–600 TB; no overage charges
AWS S3 direct egress ~$0.09 per GB Avoid direct serving at scale — always put a CDN in front
Cloudflare (Pro/Business) Flat (effectively free egress bundled) Huge ToS headroom, R2 object-store pairs well
Bunny CDN $0.01 Standard (EU/NA), $0.03 Asia/Oceania, $0.045 South America; Volume network $0.005 to 500 TB, $0.004 to 1 PB, $0.002 to 2 PB Lowest cost at scale, strong VOD support, global POPs
Fastly 100 GB free, then $0.12 (to 10 TB, NA/EU) → $0.08 next 10 TB; lower on contract Best edge compute (VCL), premium SLA
Hetzner (dedicated) Unlimited on 1 Gbit uplink; the 10G uplink add-on bills traffic above 20 TB EU champion; bring-your-own CDN or pair with Bunny
DigitalOcean Droplet Pooled Droplet allowance (500 GiB on the smallest plan), then $0.01 / GiB Great predictability, Spaces for S3-compatible object storage
GCP / Azure CDN $0.08–$0.12 tiered Pick for colocation with Google or Microsoft stacks

Our AWS vs DigitalOcean vs Hetzner comparison goes deeper into cross-provider pros and cons. For the bigger “where does the egress go” picture, the edge computing playbook walks through POP selection and worker-level tricks. Every egress and storage number below is a 2026 published list rate. Check the source sheets for your region and volume: AWS CloudFront pricing and Cloudflare R2 pricing.

2026 egress price per GB by provider: AWS CloudFront $0.085 vs Bunny $0.005, with Cloudflare and Hetzner bundling egress

Figure 2. First-tier outbound price per GB, 2026. CDN choice alone swings the bill by more than 10x.

Line 2: Storage — small bill, outsized impact

Most platforms spend 2–10% of their bill on storage, but the architecture choices made here affect every other line. A good storage policy quietly shrinks egress (fewer renditions stored), compute (fewer transcodes) and retention risk.

Typical file sizes. 1 hour of 720p H.264 is ~1.35 GB; 1080p H.265 is ~1.5 GB; 4K AV1 is ~4–5 GB; a full ABR ladder (4–5 rungs) for an hour is ~5–7 GB. On conference workloads, merging per-participant recordings into a single composite cuts storage by 30–50%.

Storage tier Price / GB-month (2026) When to use
AWS S3 Standard ~$0.023 Hot origin for VOD & live recordings
AWS S3 Infrequent Access ~$0.0125 Archived clips accessed < once / month
AWS S3 Glacier Instant / Deep Archive $0.004 / $0.00099 Compliance retention, evidentiary clips
Cloudflare R2 $0.015 S3-compatible, zero egress fees to Cloudflare network
Bunny Storage $0.005–$0.015 Pairs with Bunny CDN; cheapest end-to-end
Hetzner Object Storage Flat monthly base incl. 1 TB storage + 1 TB egress, pay-as-you-go above Flat EU baseline, S3 operations free
DigitalOcean Spaces $5/month incl. 250 GiB + 1 TiB transfer; then $0.02/GiB Predictable for small teams, S3-compatible

Retention policy is a cost lever. Default to 30–90 days on hot storage, archive to IA/Glacier after 90 days, and delete at 365 days unless the use case (compliance, courtroom, regulated industry) says otherwise. S3 lifecycle rules, Object Lock and immutability are among the cheapest CapEx line items to implement, and they survive every audit.

Line 3: Compute — origin, SFU, transcoder, signalling

Video compute splits into four workloads. Size each one on its own instead of forcing all of them onto a single instance type.

Signalling and API. A couple of burstable EC2 t3/t4g instances handle thousands of concurrent WebSocket clients without strain. So do 2–3 €5–€20 cx22/cpx21 servers at Hetzner (CPX and CCX cloud servers in Helsinki got 15% cheaper on 10 November 2026) or DO $12–$24 Droplets. This layer is stateless and scales horizontally, and for sub-10k DAU products it usually stays under $100/month.

SFU / media server. LiveKit, mediasoup, Jitsi and Janus are all CPU-heavy: 1 vCPU handles ~150–250 participant-legs at 480p, depending on bitrate. The classic choices are a c5.4xlarge ($0.68/hr on AWS), a Hetzner AX42 dedicated box (€99/month excl. VAT plus €49 setup on Hetzner’s AX42 page in November 2026, after the 15 June 2026 price adjustment) or a DO CPU-optimised droplet (~$150/month for 8 vCPU). Our comparisons of SFU architecture live in the P2P vs SFU vs MCU breakdown.

Transcoding / packaging. This runs either batched (AWS MediaConvert Basic tier at $0.0075 per normalized minute in US East (Ohio), where SD counts 1× and HD 2×; Mux and Cloudflare Stream encode for free; FFmpeg on GPU) or live (MediaLive, LiveKit Egress, custom). Batched VOD almost always comes out cheaper than DIY GPU unless your throughput is sustained. Hetzner dedicated servers with NVENC-capable GPUs start to make sense above ~2,000 encode-hours/month.

Origin and application backend. Databases, app servers, worker queues, scheduling, auth. Most products land in the standard $200–$2,000/month band. Scale up as soon as DB CPU crosses 60% or slow queries start dragging p95.

For deeper server-sizing math see our scalable video streaming app playbook.

Line 4: Third-party services — CDN, DRM, analytics, TURN

CDN. Already covered under egress: Bunny and Cloudflare for cost, CloudFront for AWS native integration, Fastly for edge compute, Akamai for enterprise SLA.

DRM. Widevine / PlayReady license servers cost nothing when you go to Google/Microsoft directly. Third-party providers (EZDRM, PallyCon, Axinom, Irdeto) run $500–$3,000/month plus per-license fees.

TURN. This is the WebRTC fallback when direct P2P fails, which happens in ~15–20% of sessions on enterprise networks. Running coturn on your own Hetzner box costs $50/month, against $40 per 100 GB relayed for managed Xirsys / Twilio, so self-hosting pays back quickly.

Analytics & QoE. Mux Data, Conviva, Bitmovin Analytics, or a Kafka + ClickHouse stack you run yourself. Managed analytics costs ~$0.002–$0.005 per viewer-session; at low scale, open-source is effectively free.

Observability. Datadog, Honeycomb, Grafana Cloud and New Relic cost $500–$2,000/month at mid-scale. Self-hosted Prometheus + Grafana + Loki starts free on a modest VM.

AWS bill climbing faster than your DAU?

We run cost audits on live video platforms and routinely cut bills 40–70% by re-architecting egress, storage and transcoding. Share your last three invoices and we’ll highlight the biggest levers.

Book a 30-min call → WhatsApp → Email us →

Architecture levers that move the bill more than the provider

1. Adaptive bitrate. A 3–5 rung ABR ladder can cut average egress by 20–30% compared with serving everyone 1080p, since most viewers end up on 720p or 480p. Use HLS/DASH with ABR from day one.

2. Codec upgrade. H.265 saves 25–50% bitrate over H.264 at the same quality, and AV1 saves another 20–30%. Encoding gets more expensive while egress drops sharply. Past ~10 TB/month of VOD egress it almost always pays off; our video encoding primer walks through the codec trade-offs.

3. P2P where possible. 1:1 WebRTC calls connect via STUN when the network path allows, so ~80–85% of sessions never touch your TURN/SFU. For 1:1 telemedicine or tutoring, P2P costs effectively nothing on the server side.

4. SFU not MCU. An SFU forwards streams without decoding them, so compute scales much better. An MCU mixes streams in the cloud, which is CPU-expensive. With 3+ participants, SFU is almost always the better choice.

5. CDN with long TTLs. With proper versioning, VOD manifests and segments can stay cached for 24–48 hours, which keeps 90+% of requests off your origin. For live, manifest TTLs of 2–5 seconds and segment TTLs of a few seconds still cut egress substantially.

6. Regional origins. Multi-region origins placed near your users cut inter-region egress and lower first-frame latency. This matters most in APAC and LATAM.

7. Autoscaling + spot instances. Scaling out for peak hours with autoscaling groups, and running batch transcoding on spot/preemptible instances, cuts compute by 40–60% compared with always-on reserved capacity.

8. ARM / Graviton. ARM instances on AWS (Graviton), Ampere on Hetzner or Tau on GCP cost ~30–40% less than x86 for the same throughput on modern video pipelines. Test them and switch wherever your stack supports it.

Architecture levers cut a $10,000 monthly egress bill to about $2,900 via ABR, H.265/AV1, CDN TTL, P2P and ARM compute

Figure 3. Architecture matters more than provider shopping: stacked levers bring a $10k egress bill under $3k (illustrative).

Worked example 1: 9-person SFU video conference

Scenario. 9 participants per meeting, 480p per upstream, SFU fan-out (each participant sees the other 8). Meetings run 4h/day × 30 days = 120 hours/month.

Egress math. 9 × 8 = 72 downstream legs. At ~0.68 GB/hr each for 480p, that totals ~49 GB/hr, or ~5.88 TB/month over 120 hours.

Approximate monthly cost. AWS: SFU media leaves EC2 directly (CDNs don’t carry WebRTC) and is billed at $0.09/GB after the 100 GB free tier, so ~$520 egress + $50 app + $40 SFU usage = ~$610. Hetzner dedicated (traffic is free on the 1 Gbit port): two AX42 boxes for app and SFU at €99 net each = ~€200, plus €98 one-off setup. DigitalOcean: ~$55 egress + $50 app + $70 SFU = ~$175. Recording storage (one composite 720p file, 1.35 GB/hr × 120 = 162 GB/month) adds ~$4–$10 depending on provider. If conferencing is the product, our video conferencing development team sizes the SFU fleet against exactly this math before the first sprint.

Worked example 2: 2-streamer, 50-viewer HLS webinar

Scenario. 2 speakers, 50 viewers watching HLS at 720p. 4h/day × 30 days.

Egress math. Each viewer pulls ~1.35 GB/hr, so 50 viewers come to 67.5 GB/hr. Over 120 hours that is ~8.1 TB/month, or closer to ~10 TB with a full ABR ladder.

Approximate monthly cost. AWS CloudFront pay-as-you-go: ~$765 egress (first 1 TB free) + $90 compute = ~$855. On CloudFront’s Pro flat-rate plan ($15/month, 50 TB allowance) the same traffic costs ~$105 all-in: 50 viewers × 120 hours of 6-second segments plus playlist reloads come to ~7.2M requests, within the plan’s 10M. Bunny CDN: ~$100 egress + $50 origin = ~$150. Mux: 360,000 delivered minutes minus the 100K free ones × $0.0008 = ~$210 for delivery, plus live-input and storage charges.

Worked example 3: 10,000-student VOD catalogue

Scenario. E-learning VOD library, 10,000 MAU, each watching 6 hours a month, 720p by default with a full ABR ladder. After the ABR uplift, monthly egress comes to ~90 TB.

Approximate monthly cost. AWS CloudFront pay-as-you-go ~$6,400 egress (1 TB free, then tiered $0.085 → $0.06/GB). At 90 TB you’re past the 50 TB Pro and Business flat-rate allowances, so going flat-rate means Premium: its 125 TB level costs $2,250 a month, about a third of pay-as-you-go. Since May 2026 Premium has been configurable, from $1,000 for 50 TB to $10,000 for 600 TB. Bunny CDN ~$450 (Volume network) to ~$900 (Standard, EU/NA). Mux: 3.6M delivered minutes at 720p ≈ $2,800 plus ~$145 storage. Cloudflare Stream: ~$3,600 delivery ($1 per 1,000 minutes) plus ~$300 for 60,000 stored minutes ($5 per 1,000). Storing a 1,000-hour catalogue with the full ABR ladder takes ~6 TB, which is $120–$140/month on S3 Standard and $30–$60 on Bunny/Cloudflare R2.

Per-student-hour cost. AWS pay-as-you-go: ~$0.11/hr. Bunny: ~$0.008–$0.015/hr. Over a year, that 7–14× gap adds up to tens of thousands of dollars. For a closer look at VOD specifically, see our video streaming app development cost guide and how to implement video streaming in an e-learning app.

Video streaming server cost by stack: a lean Hetzner/Bunny setup runs 3-7x cheaper than AWS pay-as-you-go on three workloads

Figure 4. The same three workloads on two stacks, at October 2026 list prices. On pay-as-you-go rates the lean option comes out 3–7x cheaper.

Build vs buy: when managed video services win

Managed video platforms (Mux, Cloudflare Stream, Bunny Stream, api.video, Vimeo OTT, Dacast, Kaltura) bundle encoding, storage, CDN and often analytics into one contract. At low-to-mid scale they cost less than DIY once you count engineering time. At high scale, a well-built DIY stack on Hetzner + Bunny is 2–5× cheaper. November 2026 list prices for reference: Mux charges $0.0008 per delivered minute at 720p after 100K free minutes a month, and basic-quality encoding is free; Cloudflare Stream charges $1 per 1,000 minutes delivered and $5 per 1,000 minutes stored, with free encoding. Since 15 October 2026 Cloudflare has also billed WebRTC live delivery at the same $1 per 1,000 minutes.

Go with managed video (Mux or Cloudflare Stream) when: you’re under ~50 TB/month, have no dedicated DevOps, and one engineer-week costs more than a year of the vendor premium.

Go with a DIY Hetzner + Bunny stack when: you’re past ~50 TB/month egress, you have ops capacity, and a 2–5× infra saving covers a quarter’s engineering salary in the first month.

Scale Usually cheapest Usually fastest
MVP (< 1 TB / mo) Mux free tier / Cloudflare Stream Mux / Cloudflare Stream
Growth (1–50 TB / mo) Bunny Stream / Cloudflare Stream Mux / Bunny Stream
Scale (50–500 TB / mo) Hetzner + Bunny DIY Bunny Stream
Enterprise (> 500 TB / mo) Negotiated CDN commit + custom origin Kaltura / JW Player enterprise

Video hosting pricing models: per GB, per minute or flat plan

Video hosting pricing, like video streaming pricing in general, comes in four models: per GB delivered (CDNs), as a flat-rate CDN plan (CloudFront), per minute delivered and stored (video APIs such as Mux, Cloudflare Stream and api.video), or as a flat SaaS plan with unmetered bandwidth (Vimeo). For 1,000 viewer-hours of 720p a month, November 2026 list prices range from $15 to about $190. The gap between models grows with volume, so price each one at your year-two traffic rather than at launch.

Pricing model Who bills this way (Nov 2026) List price 1,000 viewer-hours at 720p
Per GB delivered (CDN) CloudFront pay-as-you-go, Bunny CDN, Fastly $0.085 / $0.01 / $0.12 per GB (NA/EU) ~$60 / ~$17 / ~$190
Flat-rate CDN plan CloudFront Pro, Business, Premium $15 / $200 for 50 TB; Premium $1,000–$10,000 for 50–600 TB $15 (Pro)
Per minute delivered and stored Mux, Cloudflare Stream, api.video $0.0008/min (Mux 720p); $1 per 1,000 min; from $0.0017/min ~$48 / $60 / ~$102
Per GB, encoding included Bunny Stream Storage from $0.01/GB, delivery from $0.005/GB ~$9–$17
Flat SaaS plan Vimeo Creator, Core, Professional, Studio $9 / $24.50 / $71 / $400 a month, billed annually Plan price only

The 1,000-viewer-hour column assumes ~1.7 TB of egress (60,000 delivered minutes) and ignores free tiers, except CloudFront’s first free TB. Per-minute pricing is the easiest to forecast, but at 720p it comes to $28–$60 per TB delivered, compared with $5–$10 on Bunny. api.video lists storage from $0.00285 per stored minute, and Bunny Stream encodes for free at standard quality but charges $0.025–$0.15 per output minute for Premium Encoding.

Vimeo replaced its Starter, Standard and Advanced plans with Creator, Core, Professional and Studio, announced on 28 October 2026; existing accounts switch at renewal. Flat plans work for marketing and course video with a predictable audience. If viewing in your product grows with the user base, a per-GB or per-minute model keeps the bill tied to usage. To run your own traffic through the formula, use our CDN cost calculator walkthrough.

Five pitfalls that sink video-infrastructure budgets

1. Serving video straight from S3. With no CDN and no cache, every byte a viewer pulls is billed at the raw S3 rate and grows linearly with DAU. This one mistake leaves many early AWS bills 5–10× higher than they should be.

2. No ABR, single-bitrate streams. Everyone gets 1080p whether they need it or not. Mobile users hit buffering, the bill balloons, and engagement still drops.

3. Infinite retention. Recordings are never archived or deleted. A year in, you’re storing 10 TB of footage nobody has watched since week 1.

4. Always-on reserved capacity. SFU and transcoder nodes run 24/7 at peak capacity. With autoscaling + spot you pay only for what you use.

5. No unit economics. Teams track monthly totals but not cost-per-user-hour, so they can’t tell whether growth helps or hurts until cash runway becomes the problem. Track per-user cost from day one.

A decision framework — pick your stack in five questions

Q1. How much monthly egress in year 2? < 1 TB → Mux or Cloudflare Stream (the free tier covers you). 1–50 TB → Bunny Stream or Cloudflare Stream. 50+ TB → DIY Hetzner + Bunny, or a negotiated CloudFront commit.

Q2. What’s the interactive pattern? 1:1 → P2P WebRTC keeps server cost lowest. Small many-to-many rooms → SFU. Large “stadium” broadcasts → HLS/LL-HLS with CDN.

Q3. Region and compliance? EU with data residency rules → Hetzner + Bunny or OVH. US-first → AWS / Cloudflare. Mostly APAC → GCP or Cloudflare with regional POPs. Courtroom / medical / defence → on-prem or dedicated cloud with signed audits.

Q4. Team shape. No dedicated DevOps → Mux or Cloudflare Stream. Full ops team → DIY with a quarterly cost audit. In between → Bunny Stream or hybrid.

Q5. Monetisation model. Per-seat SaaS → model cost-per-seat-hour. Pay-per-view / AVOD → model cost-per-viewer-hour. Flat-rate enterprise seat licences → model cost-per-peak-concurrency.

Need help stress-testing the numbers?

We’ll co-build your cost model, pressure-test assumptions against our production experience, and deliver a conservative monthly estimate with per-user-hour economics. Agent Engineering acceleration means the numbers land fast.

Book a 30-min call → WhatsApp → Email us →

Cost plan per product stage: MVP, launch, growth, scale

MVP (0–100 active users). Keep the bill under $100/month. Use Mux or Cloudflare Stream for VOD and a single DigitalOcean Droplet for signalling, and don’t pay for observability beyond free tiers. Skip the over-engineering; track unit cost and nothing more.

Launch (100–1,000 active users). The bill is usually $200–$1,500/month. Start splitting roles (API, SFU, origin). Add ABR if you haven’t yet. Check cost-per-user-hour weekly.

Growth (1k to 10k active users). Bill $1.5k to $15k/month. Review the four lines every month and switch providers when a line can be cut in half. If you’re on AWS and growing fast, look at a hybrid stack (Hetzner compute + Bunny CDN).

Scale (10k+ active users). The bill is $15k+, but unit cost should be falling. Negotiate CDN commits, look into direct peering, tighten cold-tier storage and revisit the codec mix every two quarters. At this stage an in-house SRE/infra engineer pays for themselves.

Monthly video platform server cost by stage: MVP under $100, launch $200-1.5k, growth $1.5k-15k, scale $15k+

Figure 5. What to run and what to budget at each product stage in 2026.

Mini case: Worldcast Live — 10,000 concurrent HD viewers at predictable cost

Situation. Worldcast Live needed a high-definition live concert streaming platform for 10,000+ concurrent viewers with sub-second latency. Event nights brought unpredictable peaks, and between events costs had to stay under strict control.

What we built. A tiered stack. Dedicated ingest nodes on Hetzner handled transcoding cost-efficiently, a Bunny CDN with regional POPs carried egress, and CloudFront served only the regions where Bunny performance needed back-up. Signalling autoscaled on Kubernetes, and VOD replay used long-TTL segments. Where concert networks permitted, we added P2P relay on the viewer side. Full-ladder ABR and H.265/AV1 dual encoding gave 50% egress savings vs baseline H.264.

Outcome. Cost per concurrent viewer-hour held at a predictable rate through 10× traffic spikes. Between events the infra bill drops to baseline autoscale, and peak nights never blew the budget because the architecture bent instead of breaking. Want a similar predictability model for your live product? Book 30 minutes and we’ll walk through the numbers.

KPIs: the four numbers to watch on the infra dashboard

1. Cost per active user-hour. Total infra spend ÷ total active user-hours. This is the north-star unit metric. Target: it falls, or at least holds flat, as the product scales.

2. Egress GB per DAU. Watch for creep from unoptimised client-side logic, duplicate fetches, accidental 4K on mobile or missing CDN caching.

3. Cache hit ratio. At 90+%, your CDN is doing its job. Below 80%, manifests or segments aren’t cached long enough.

4. Storage growth vs usage growth. If storage grows faster than DAU, your retention policy is broken. Audit it quarterly.

When you should not run your own video infra

Fewer than 1 TB/month egress, no dedicated DevOps, MVP stage, or video that isn’t core to the business: use Mux, Cloudflare Stream, Bunny Stream or Vimeo. The engineering hours you’d spend tuning an SFU or a CDN pipeline are better spent on product. Revisit DIY when you cross ~10 TB/month, start hitting managed-vendor price shelves, or when compliance or residency rules force on-prem.

Reach for a cost audit when: your video bill grows faster than daily active users, or a single line (egress or transcode) tops 70% of spend and nobody has re-architected it in two quarters. Our independent video architecture review starts from this four-line breakdown and ends with a plain list of what to keep and what to change.

FAQ

What’s a realistic monthly bill for an MVP video app?

A 1:1 video MVP on DigitalOcean or Hetzner runs $50–$300/month with WebRTC and TURN. A simple VOD product under 1 TB egress on Mux or Cloudflare Stream lands at $20–$100/month. Costs climb seriously only once you cross 1 TB/month egress or 500 concurrent WebRTC legs.

AWS or Hetzner for a video startup?

Hetzner for anything egress-heavy in Europe. AWS when you need managed services (MediaConvert, MediaLive, Transcribe), enterprise contracts, or global breadth beyond Bunny/Cloudflare’s footprint. Many production stacks are hybrid: Hetzner for bulk origin and SFU, AWS for the managed pieces that would take weeks of engineering to build yourself.

Do I need a CDN for a WebRTC conferencing product?

For real-time WebRTC streams, no: CDNs don’t cache RTP traffic. For recordings (VOD playback after the call), yes. Put a CDN in front of S3/R2 from day one.

How much does transcoding really cost?

AWS MediaConvert’s Basic tier (US East (Ohio), first 100,000 normalized minutes) works out to about $0.0075/minute for SD, $0.015 for HD and $0.03 for 4K. FFmpeg on a spot GPU runs ~$0.002–$0.005/minute for HD at scale. Mux and Cloudflare Stream don’t charge for basic encoding at all, which is fine for most MVPs. Transcoding is almost never the dominant cost. Egress is.

Does H.265 or AV1 really save money?

Yes, once egress crosses ~10 TB/month. H.265 cuts bitrate 25–50% vs H.264, and AV1 cuts another 20–30% on top. Encoding costs more (AV1 especially, at ~10× H.264 encode time), but past a few TB the egress savings far outweigh it.

When should I move off managed vendors?

Usually around 50–100 TB/month egress, when the vendor’s per-GB rate exceeds what a DIY Hetzner + Bunny stack would cost. Below that, stay managed: the savings don’t justify the engineering spend.

How do I predict the bill for a product that hasn’t launched?

Model the four lines with conservative assumptions: 70% of DAU actively streaming, average session of 15–20 minutes, a 1.3× uplift for full ABR, and a 90-day retention policy. As a sanity check, run the model at 10× your expected launch DAU. Once live, review it monthly against real metrics.

How much does a live streaming server cost per month?

Plan on roughly $100–$300 a month of compute for a small live stack (ingest and transcode box plus origin), then egress on top: about 1.7 GB per viewer-hour at 720p with ABR. At November 2026 list prices, 10,000 live viewer-hours cost ~$170 on Bunny, $200 on CloudFront’s Business flat-rate plan (live playlist reloads push the request count to ~12M, past Pro’s 10M) and ~$1,325 on CloudFront pay-as-you-go.

What does video CDN pricing look like in 2026?

Pay-as-you-go list rates in North America and Europe: CloudFront $0.085/GB after 1 TB free, Fastly $0.12/GB after 100 GB free, Bunny $0.01/GB Standard or $0.005/GB on its Volume network. The newer option is flat-rate. CloudFront Pro costs $15 a month for 50 TB and 10M requests, and Business costs $200 for 50 TB and 125M requests, with no overage charges (November 2026).

How much does a worldwide CDN cost for 50 TB a month?

At November 2026 list prices, 50 TB a month costs about $250 on Bunny’s Volume network, about $875 on Bunny Standard with a 70/20/10 split across North America and Europe, Asia and South America, and about $3,965 on CloudFront pay-as-you-go at North America and Europe rates. CloudFront’s Business flat-rate plan covers it for $200. 50 TB of 720p video is roughly 29,000 viewer-hours and 17M or more HLS requests, too many for Pro’s 10M.

How much does video hosting cost per month?

For 1,000 viewer-hours of 720p a month, video hosting costs about $17–$102 at November 2026 list prices: ~$17 on Bunny CDN, $48 on Mux before its 100K free minutes, $60 on Cloudflare Stream and ~$102 on api.video. Vimeo sells flat plans instead, from $9 to $400 a month billed annually, and doesn’t meter bandwidth.

How is storage priced on video platforms?

Object storage bills per GB-month (S3 Standard ~$0.023, Cloudflare R2 $0.015, DigitalOcean Spaces $5 for 250 GiB), while managed video platforms bill per minute stored (Cloudflare Stream $5 per 1,000 minutes, Mux $0.0024 per minute at 720p). An hour of 720p with a full ABR ladder takes about 5–7 GB, so storage usually stays at 2–10% of the bill.

Is ARM cheaper for video streaming?

For the compute line, usually yes. ARM instances such as AWS Graviton or Hetzner’s Ampere servers run ~30–40% cheaper for comparable throughput on modern video pipelines. Egress doesn’t change, though, so ARM trims only the 5–20% compute share of the bill.

Can Fora Soft build my cost model and validate the stack?

Yes, that’s a common first engagement. We take your product spec, build a conservative four-line model, and recommend a provider mix with per-user-hour economics. Typical turnaround is 3–5 working days. Book a call to kick it off.

Implementation

How to Implement Video Streaming in an E-learning App

The playbook that complements this cost model on the engineering side.

Cloud

AWS vs DigitalOcean vs Hetzner

Deeper comparison of the three providers that dominate startup video infra.

Architecture

P2P vs MCU vs SFU for Video Conferencing

The architecture that defines how much compute you burn per participant.

Edge

Edge Computing in Live Streaming

How edge tricks cut latency and egress at the same time.

Case study

Worldcast Live — 10,000-concurrent HD streaming

How we held cost-per-viewer-hour flat through 10× peaks.

Ready to model your real video-platform cost?

Video-platform infrastructure cost looks scary until you split it into four lines. Egress is the one that matters. It’s usually 70–95% of the bill, and the levers that shrink it (ABR, H.265/AV1, CDN choice, P2P, long TTLs) are architecture decisions, not provider negotiations. Storage is small but still needs policing with retention policies and cold tiers. Compute stays moderate once you autoscale and pick the right SFU pattern for your cohort size. Third-party services are a modest tax for everything you don’t want to build from scratch.

Pick providers with the four lines in mind: AWS for managed breadth (and, under 50 TB a month, its flat-rate CloudFront plans), Hetzner for bulk EU compute and free dedicated egress, Cloudflare or Bunny for world-class CDN economics, DigitalOcean for predictability. Model per user-hour rather than monthly totals, and review quarterly. Do that and your infra bill stops being a surprise and becomes the quiet margin that keeps your video business viable. If you want a team that has shipped video infra across conferencing, streaming and OTT for 20+ years, book a call. We’ll co-model your stack with Agent Engineering acceleration and tell you honestly where to save and where not to.

Book cover: Video Streaming, Volume 3 of the Fora Soft Video Engineering Handbook

The book · Volume 3 of 7

Video Streaming: A Complete Guide to Video Delivery: From ABR and CDNs to Players, WebRTC, DRM, and the Economics of Streaming

Your cost model shows that egress dominates, and the levers that move it live in the delivery layer. Volume 3 of our handbook covers ABR ladders, CDN caching, WebRTC topologies and the economics of streaming, including how each choice changes the bill per viewer-hour. Read it before you sign a CDN contract or size an SFU fleet. Written by Nikolay Sapunov, CEO at Fora Soft.

Read it on Amazon →Kindle & paperback · Fora Soft Video Engineering Handbook

Let’s build your video cost model together

20+ years of real-time video across 250+ products. Book 30 minutes and walk away with a four-line cost model, a provider recommendation and realistic per-user-hour economics for your product.

Book a 30-min call → WhatsApp → Email us →

  • Processes
    Services