KNX vs Loxone: which smart-building stack should you specify in 2026?

KNX and Loxone solve the same problem from opposite ends. KNX is decentralised by design: every push-button, every dimming actuator, every blind controller carries its own microcontroller and its own bus coupling unit (BCU). The bus is just a twisted pair carrying 9,600 bit/s telegrams between group addresses. Pull any device out and the rest of the installation keeps running. That is what makes KNX the default for offices, hotels, hospitals, and airports — a single failed sensor cannot black out a wing.

Loxone is unapologetically centralised. The Miniserver is the brain. Tree branches (up to 50 devices, 500 m per branch) and Air (wireless mesh) hang off Extensions that all report back to the Miniserver. All logic, scenes, schedules, and visualization live on the Miniserver itself. This is brilliant for installers: one device to program, one config file, one log to read at 11 pm when a client calls. It is less brilliant if the Miniserver dies or the firmware update goes badly — the entire house goes dark until you restore from backup.

Neither approach is wrong. KNX trades up-front complexity for long-term resilience and multi-vendor flexibility. Loxone trades single-vendor risk for radical simplicity and a unified user experience. If you are designing a building that has to be re-tendered to a new integrator in fifteen years, the decentralised, open, twisted-pair bus is the safer bet. If you are wiring a $400k villa for a client who will keep the same installer on retainer for a decade, the centralised approach is faster, cheaper, and prettier.

Loxone wins year one. KNX often wins year ten. Here is the honest math.

A small residential project (one Miniserver, 20-30 Tree devices, app visualization) lands around $15-20k installed in the UK or US, including labour. A comparable KNX project — say 25 actuator channels, 15 sensors, ETS programming, plus a third-party visu — typically runs $22-32k installed. The gap is real and comes from three places: ETS6 Pro licence (€1,000 one-off), longer commissioning hours, and the fact that KNX visualization is almost always a separate hardware/software stack (Gira G1, ComfortClick, 1Home, openHAB on a Pi).

Where Loxone loses ground is the maintenance and migration tail. When Loxone deprecates a Miniserver generation — as it did with Gen 1 — you eventually have to migrate. When the original installer stops trading, your client needs to find another Loxone Partner who will inherit a config they did not write. KNX hardware from 1998 still talks to KNX hardware from 2026 on the same bus, and any KNX Partner anywhere in the world can open the ETS project file. For a 10-year residential build that is a footnote. For a 20-year commercial fit-out it is the entire business case.

Rule of thumb: under €30k of hardware and a single installer, Loxone’s TCO is lower. Above €60k or multi-tenant, KNX pulls ahead.

This is where the two camps stop being polite. KNX evangelists call Loxone a "walled garden". Loxone evangelists call KNX "a 1990s spec held together by ETS licences". Both have a point.

KNX is genuinely open. The bus standard is ISO/IEC 14543-3. Datapoint types (DPTs) are publicly specified. Any of 500+ manufacturers — ABB, Gira, Jung, Schneider, Hager, MDT, Zennio, Theben, Berker, Siemens — can ship a device that plays nicely with any other. Your light actuator can be ABB and your weather station can be Theben and your room thermostat can be Zennio, and ETS6 binds them through group addresses. If MDT goes out of business tomorrow, your installation does not care.

Loxone is a single-vendor stack with selective openness at the edges. The Miniserver speaks KNX (via the KNX Extension), Modbus RTU/TCP, DMX, 1-Wire, EnOcean, and RS-232/485, so you can pull non-Loxone devices in. But the core — Tree, Air, Loxone Link, the app, the cloud services — is closed. If Loxone changes its commercial terms, deprecates a product line, or simply stops being a company, the depth of the hole varies from "annoying" (replace the Miniserver) to "rip and replace" (no more Tree firmware updates).

For a homeowner who values one slick app and one accountable vendor, lock-in is a feature, not a bug. For a facilities manager specifying a 15-year office refit, it is a yellow flag on the risk register.

The certification paths look similar on paper and diverge sharply in practice.

KNX Partner status requires attendance at a KNX-certified training centre and a passing exam. The KNX Basic Course is typically 30 hours over 5 days, runs around £750 / €900 in the UK or Germany, and ends with a written + practical ETS exam. You then buy ETS6 Professional at €1,000 (one-off, dongle or cloud licence) to commission real projects. KNX Advanced and KNX Tutor follow for those who want to teach. Globally there are around 115,000 certified KNX Partners. The skill is portable: a KNX Partner trained in Munich can open and finish a project file started in Dubai.

Loxone Partner status is more like a sales-channel programme than a craft qualification. You register on the Loxone Partner portal, work through the free Loxone Campus e-learning modules, and progress through tiers (Silver / Gold / Platinum) based on revenue and certified staff. There is no expensive software licence — Loxone Config is a free download. The learning curve is genuinely shorter; a competent electrician can be programming useful scenes after a few days. The downside is that the skill is non-portable: it is worthless outside the Loxone ecosystem.

For the EU/UK installer the practical advice is to hold both. Many of the most profitable integrators in 2026 do exactly that: Loxone for residential speed jobs, KNX for the commercial work that pays the mortgage.