Aerix

How to Build a Private 5G Network

The eight steps we follow on every UK deployment, from the first site walk to the first SIM going live, and the places they tend to go wrong.

Most of what is written about private 5G describes the parts: a radio, a core, some spectrum, some SIMs. Far less is written about the order in which you put them together, and that is where projects actually stall. A radio can be delivered in a fortnight; the backhaul to feed it, the Ofcom licence to let it transmit and the landlord's consent to bolt it to a wall can each take longer than everything else combined.

We have built private 5G networks in some fairly unforgiving places, the Llanthony Valley in the Brecon Beacons being the one we usually cite: a steep Welsh valley that was living on 2 Mbps ADSL, now served by a handful of radios with backhaul brought in from outside. This guide sets out how to build a private 5G network the way we do it, in eight steps, with a candid note on where each one goes wrong. If you want the architecture first, see how private 5G works, which includes a full architecture diagram.

Step by step

How to build a private 5G network in eight steps

In the order we do them. Steps three and four run in parallel with the others, because they take longest.

  1. 1

    Decide what has to connect, and where

    Start with the devices, not the network. List what will connect (handhelds, cameras, sensors, vehicles, routers serving buildings), roughly how many of each, where they sit, and whether they move. A scanner on a forklift, a camera on a pole and a router in a static caravan are three quite different radio problems.

    Then write down what the network is for on its worst day. A network that carries card payments and staff radios can tolerate a short outage; one that carries production control or safety systems cannot, and that single answer drives most of the design choices that follow.

    Where it goes wrong: Designing for coverage area instead of device count. It is the number of devices, and how much traffic each sends, that usually produces the figure nobody budgeted for.

  2. 2

    Survey the site

    Walk the site with a spectrum analyser and a map. The survey establishes where radios can be mounted, whether there is power at each position, what obstructs the signal (hills, steel racking, stone walls, trees in leaf) and where devices will actually be used.

    On an open site one well-placed radio may cover several hundred metres; in a warehouse or a listed building the answer can be one small cell per bay. Indoor and outdoor customer devices behave very differently at the same spot, so the survey also decides which kind each building gets.

    Where it goes wrong: Surveying in winter and deploying in summer, or the reverse. Foliage and stock levels change the radio environment materially, and a design with no margin will find out.

  3. 3

    Secure the backhaul

    Every radio has to reach the internet, or your own network, and on most sites this is the item with the longest lead time. Fibre is best where it exists; fixed wireless or a microwave hop from the nearest fibre point is common on rural sites; satellite is the last resort but a perfectly workable one.

    Size the backhaul for the busy hour, not the average, and decide early whether you need a second route. If the core runs in the cloud, losing the backhaul takes the whole network down, not just the internet.

    Where it goes wrong: Leaving it until the radios arrive. On rural builds, backhaul routinely takes longer to arrange than everything else combined, so it should be ordered in week one.

  4. 4

    License the spectrum

    In the UK, private 5G networks transmit under an Ofcom Shared Access licence. Three bands are available: 1800 MHz, 2300 MHz and 3.8–4.2 GHz, the last being the usual choice for new 5G builds. Licences are issued per location, so the application needs the radio positions, antenna heights and power levels from the survey.

    Ofcom checks each application for interference with existing users nearby, which is why a licence is not guaranteed at the power or bandwidth first asked for. Build some flexibility into the design rather than assuming the maximum.

    Where it goes wrong: Buying radios or devices that do not support the band you are licensed for. 3.8–4.2 GHz is band n77, and plenty of equipment sold as "5G" supports only n78, which stops at 3.8 GHz.

  5. 5

    Choose the core, and where it runs

    The 5G core authenticates SIMs, sets up sessions and routes traffic. It is software, and comparatively cheap; the decision that matters is where it runs. A cloud-hosted core is simpler to operate, patch and monitor across many sites. An on-premise core, on a small server in a comms cabinet, keeps devices talking to local systems when the backhaul fails.

    For most community, hospitality and small-business sites we run the core in the cloud. For factories, ports and anything air-gapped we put it on site. Both are standalone 5G; there is no reason to build a new private network on a 4G core.

    Where it goes wrong: Choosing by vendor rather than by failure mode. Ask what should still work when the site loses its internet connection, and the answer usually chooses the core for you.

  6. 6

    Install the radios and power

    Mount the radios where the survey said, on existing structures wherever possible: a barn gable, a lighting column, a clubhouse wall. Small cells can often be powered over Ethernet from a nearby switch, which keeps each position to a single cable; larger outdoor radios need a mains feed.

    Whether a position needs planning permission depends on the size of the equipment, where it is mounted and who is installing it. Small units on existing buildings are often straightforward, but listed buildings, conservation areas, new poles and leased premises each add a consent to obtain. Start those conversations as soon as the positions are known.

    Where it goes wrong: Assuming the landlord will say yes. Landlord and planning consents are the most common reason a radio sits in its box for a month.

  7. 7

    Provision SIMs and devices

    Every device needs a SIM issued for your network identity rather than a public operator's, with its credentials loaded into your core. For a handful of devices that is a manual job; for hundreds it needs a proper provisioning process, so that a SIM, its device and its customer are all recorded in one place.

    Customer premises equipment (the routers that serve buildings) needs configuring for your network and, ideally, managing remotely so that firmware and settings can be changed without a site visit.

    Where it goes wrong: Device compatibility. Not every phone or modem will attach to a private standalone network, even on the right band; test the exact model before buying a fleet of them.

  8. 8

    Test, go live and keep watching

    Before handing the network over, walk it again with real devices: signal strength and speed at every location that matters, handover between radios for anything that moves, and a failover test of the backhaul. Then connect users in stages rather than all at once.

    A private network is a service, not a project with an end date. Radios need firmware updates, devices get replaced, and someone has to notice when a radio stops transmitting before the users do. Budget for monitoring and support from day one.

    Where it goes wrong: Treating go-live as the finish line. Most of the cost of a private network over its life is in running it, and most of the frustration comes from nobody owning that job.

Build or buy

Should you build it yourself?

Every step above can be done in-house, and for some organisations that is the right answer. We would say so, even as a managed service provider. Building your own private 5G network makes sense when (i) you have people who can run a cellular network as confidently as they run a Wi-Fi one, (ii) the network carries something where you need total control of the core, such as production systems or an air-gapped site, and (iii) you expect to keep it long enough for the equipment to pay for itself. A large port or a car plant with its own telecoms team is the obvious case.

Outside those conditions the equation changes quickly. The equipment is rarely the hard part; steps three, four and eight are, and they do not get easier with practice unless you are doing them repeatedly. That is the gap a managed service fills: we do the same eight steps, but we have done them on farms, holiday parks, city streets and valleys before, and we carry the monitoring afterwards. Ultimately however, the right choice depends on who will be answering the phone at 7am when a radio stops transmitting, and it is worth deciding that before buying anything.

Common questions

How long does it take to build a private 5G network?

For a single site with straightforward access, four to eight weeks from agreement to the first devices going live is typical. The long poles are rarely the radios; they are the backhaul, the Ofcom licence and any landlord or planning consents. An enterprise build procured through a vendor and systems integrator usually takes quarters rather than weeks, largely because of the procurement process around it.

Can I build a private 5G network myself?

Yes. The equipment and the Ofcom Shared Access licence are available to any organisation, and open-source 5G cores exist. What you need in-house is radio planning, core configuration, SIM provisioning and, above all, someone to run the network after it is built. Organisations with an existing telecoms team often do; most others find a managed service cheaper once the running costs are counted.

What do I need to build a private 5G network?

Eight things, in roughly this order: a clear list of what has to connect, a site survey, backhaul, an Ofcom Shared Access spectrum licence, a 5G core, radios with power and mounting positions, SIMs and devices provisioned for your network, and a plan for monitoring and support once it is live.

How much does it cost to build a private 5G network?

Bought outright in the UK, a single-site build typically runs from around £50,000 for a pilot to £500,000 and beyond for a large industrial site, with radios and integration work making up most of that figure. As a managed service the capex disappears and you pay monthly instead, from £35 a month for a connected site.

Do I need planning permission for a private 5G network?

Often not for small radios on existing buildings, but it depends on the size of the equipment, where it is mounted and who is installing it. Listed buildings, conservation areas, new masts and leased premises can each require planning permission or a landlord's consent, so it is worth identifying radio positions early and checking each one with the local planning authority.

Related guides

Before you start

Planning a build?

Tell us about the site and what has to connect. We will walk you through the eight steps as they apply to you, including the ones likely to hold you up.