
Building an IoT platform in 2026 typically costs $80,000 to $1,000,000+ to ship a real V1. A single-product device fleet runs $80k to $200k. A multi-tenant platform that hosts other companies' devices runs $250k to $1M. Industrial or regulated IoT (medical, energy, automotive) starts at $1M and grows from there, mostly on certification, not engineering.
Most cost guides hand you a single number and a generic feature list. This one names the actual stack picks at each tier, and budgets the line item nobody else admits to: device certification routinely costs more than the software.
An IoT platform is not one product. It is roughly nine moving parts, and each one has its own cost curve.
A note on platform risk: Google Cloud IoT Core was decommissioned in August 2023, forcing thousands of platforms to migrate mid-flight. If you build on a managed IoT-specific service, treat deprecation exposure as real. Founders today default to AWS IoT Core, Azure IoT Hub, or self-hosted MQTT with portable message schemas.
Every honest IoT cost conversation starts here. The differences are huge.
Tier 1: single-product device fleet ($80k to $200k). You make one product (a smart sensor, a connected thermostat, a fitness wearable). You sell it to end users. The platform serves your devices only.
Tier 2: multi-tenant platform ($250k to $1M). You host other companies' devices. Think a SaaS where customers bring their own hardware and you handle the broker, ingest, dashboards, and APIs. This is roughly a 4x cost jump because tenancy, billing, RBAC, and data isolation all become first-class.
Tier 3: industrial or regulated IoT ($1M+). Medical devices, smart-grid endpoints, automotive, industrial control. Engineering is the small part. Certification, audits, and field service own the budget.
If you're not sure which tier you're in, the test is simple: do you ship the hardware, host other people's devices, or get audited by a regulator?
Hiring shape moves the bill more than tooling does. Here's what each path looks like for a 6-month Tier 1 build:
| Approach | Cost | Timeline | Pros | Cons |
|---|---|---|---|---|
| US full-time IoT engineer | $160k-$240k/yr + 30% benefits | 6-12 weeks to hire | Deep ownership, IP retention, on the cap table | Long hire cycle, single point of failure, hard to fire |
| IoT dev agency (US/EU) | $120-$250/hr; $200k-$1M project | 4-8 weeks to start | Multi-skill team, accountable contract | High markup, often weak in firmware specifics, locks you to their stack |
| Freelancer (Upwork) | $25-$120/hr | 1-2 weeks to start | Cheap entry, fast | Wildly variable quality, few firmware + cloud generalists, ghost risk |
| Toptal | $60-$200/hr | 2-3 weeks | Vetted, established brand | Premium markup, monthly minimums, slower to replace |
| Cadence | $500-$2,000/wk | 48-hour trial then ship | AI-native baseline, weekly billing, replace any week, voice-vetted | Less suited to enterprise procurement that wants a master services agreement |
Cadence sits at the bottom by design. Founders book engineers by the week against a four-tier rate card: junior $500/wk for cleanup and integrations with good docs, mid $1,000/wk for end-to-end shipping of standard features, senior $1,500/wk for architecture and edge-case work, and lead $2,000/wk for systems design and fractional-CTO scope. Every engineer on the platform is AI-native by default, vetted on Cursor, Claude, and Copilot fluency in a founder-led voice interview before they unlock bookings. The pool sits at 12,800+ engineers with a 27-hour median time to first commit.
This is the section the other top-10 results skip. Vendor-agnostic advice is useless when you're sizing a budget.
| Broker | Hosting | Cost | When to pick |
|---|---|---|---|
| Mosquitto | Self-host | Free | Tier 1, under 1,000 devices, you have a Linux box |
| HiveMQ Cloud | Managed | Free up to 100 connections; from $0.04/connection/mo | Tier 1 to early Tier 2, you want zero broker ops |
| EMQX Cloud Serverless | Managed | From $0.18/hr metered | Tier 2, bursty load, multi-region |
| AWS IoT Core | Managed | $1.00 per million minutes connected + $1.00 per million messages | Tier 2, already on AWS, want IAM-native auth |
| Azure IoT Hub | Managed | From $10/month for 400k messages | Tier 2 to Tier 3, you live in the Microsoft stack |
Rule of thumb: managed broker until you have 50,000+ connected devices. Past that, self-hosted EMQX or HiveMQ on your own VMs cuts the bill 40-70%, but you own broker uptime.
If you're building greenfield, default to AWS IoT Core for Tier 1 and 2. The combination of IAM auth, device shadows, and Lambda triggers is hard to beat for the first 50,000 devices. Move to a self-hosted EMQX cluster only when the AWS bill crosses what a DevOps engineer costs.
Do not roll your own. Memfault is $0.50 to $2 per device per month. Mender is open source with a paid tier. Balena bundles OTA with device management. Three weeks of in-house OTA work costs more than five years of Memfault for a 5,000-device fleet.
For a deeper read on managed-cloud trade-offs at this layer, our writeup on migrating from Heroku to AWS walks through the same managed-vs-self-host decision tree at the application layer.
For Tier 1 consumer IoT, certification is real but bounded. For Tier 3 industrial, it's the dominant cost. Here are the rough numbers per SKU:
For an industrial IoT platform with three SKUs targeting US, EU, and medical-adjacent markets, certification routinely costs 2x to 5x the engineering build. Founders who skip this in their pitch deck either haven't sold to a hospital yet or are about to learn an expensive lesson.
Two adjacent reads: building a healthcare app covers the HIPAA / SOC 2 layer in detail, and our Stripe integration cost breakdown is a useful reference for the billing surface most multi-tenant IoT platforms eventually need.
Five rules that have saved founders six figures more than once.
If you want help running the numbers on which approach fits your scope, book a 48-hour trial with a senior engineer on Cadence and have them sketch the architecture and a phased budget before you commit to a stack.
A clean three-step plan for founders who already have a hardware prototype and need to get to paying customers.
Step 1: lock the stack on paper before you write code. Pick the broker, the time-series store, the cloud bridge, and the OTA vendor. This decision drives 70% of your monthly bill. A senior engineer can do this in three days.
Step 2: ship a working fleet of 10 devices end-to-end. Firmware, broker, ingest, dashboard, OTA. Real devices in real conditions. This is the build that tells you what's actually broken in your assumptions. Budget 6 to 8 weeks with two engineers.
Step 3: hire incrementally with weekly billing. Once the 10-device fleet is live, you'll know exactly which role you need next: more firmware depth, a dedicated frontend, a DevOps engineer to harden the broker. Hire that exact role, not a "full-stack IoT engineer."
For the third step, the on-demand path is faster than recruiting. Cadence will shortlist four engineers in 2 minutes against your spec, and the 48-hour free trial means you can drop anyone who doesn't fit before the first invoice. If you'd rather see what other founders shipped first, the recruiting platform cost breakdown is a useful comparison for what a full-cycle hire costs in time and dollars.
Try it on your own scope. Tell Cadence what you're building. Get a 4-engineer shortlist with rates in 2 minutes, run a 48-hour free trial, and ship. Replace anyone the next week if the fit is wrong. Start your shortlist.
A single-product device fleet ships in 4 to 6 months with two engineers. A multi-tenant platform takes 9 to 12 months with 4 to 6 engineers. Industrial IoT with certification cycles adds 6 to 18 months on top of that.
AWS IoT Core is cheaper until roughly 50,000 connected devices. Past that, self-hosted EMQX or HiveMQ on your own VMs typically cuts the bill 40 to 70%, but you take on broker uptime yourself. For most early-stage platforms, the managed bill never reaches the threshold where a self-host migration pays back.
Google decommissioned IoT Core in August 2023, forcing thousands of platforms to migrate. The lesson is platform risk. If you build on a managed IoT-specific service, your deprecation exposure is real. Many founders now default to AWS IoT Core, Azure IoT Hub, or self-hosted MQTT with portable message schemas so they can move brokers without rewriting firmware.
FCC for a US radio runs $5,000 to $30,000 per SKU. CE marking is €3,000 to €15,000. UL listing for safety-critical devices is $10,000 to $50,000+. HIPAA controls and audits start at $20,000 and climb to $100,000+. For regulated industrial IoT, certification often costs 2x to 5x the engineering build.
You can prototype on a Raspberry Pi or ESP32 with a managed broker like HiveMQ Cloud and a no-code dashboard like Tinybird or Grafana Cloud. To ship to paying customers you need at least one firmware-fluent engineer and one cloud engineer. Booking a senior on Cadence for two weeks is a faster way to find out what the production-grade build actually requires than reading another cost article.