A PRI is a physical TDM circuit carrying a fixed number of voice channels — 23 on a standard T1-based PRI — dedicated to voice and nothing else. SIP trunking carries calls as IP traffic over a data connection, so capacity is a licensing and bandwidth question rather than a hardware one. SIP is the better fit for most organizations still on PRI: it scales in software, survives site failure through rerouting, and usually costs less per concurrent call. PRI retains a real advantage where the PBX cannot be upgraded economically, where the site’s data connection is not yet good enough to carry voice reliably, or where analog-dependent equipment dominates the building.
PRI has been the enterprise voice standard for long enough that a great many phone systems are still connected to one. Some carriers have published timelines for withdrawing TDM services and others have not, and the position differs by region and by carrier, so the first thing worth doing is asking yours directly rather than assuming a deadline exists or that none does. That answer changes whether this is a scheduling exercise or an optimization.
The important thing to understand is that the two differ in what they are, not merely in how they are priced. A PRI is a dedicated circuit whose capacity is fixed in hardware. A SIP trunk is a service delivered over a data connection whose capacity is defined in software and bounded by the quality of that connection. That last clause is where migrations go wrong, and it is worth more attention than the per-channel comparison that usually dominates the conversation.
| Criterion | SIP Trunking | PRI |
|---|---|---|
| What it is | Voice service delivered as IP over a data connection | A dedicated TDM circuit carrying fixed voice channels |
| Capacity unit | Concurrent call paths, set in software | 23 channels per PRI, set by the circuit |
| Adding capacity | A configuration and licensing change, usually same-day | Another physical circuit and another install |
| Granularity | Add or remove call paths individually | Add in blocks of 23 whether you need them or not |
| Physical delivery | Rides your existing data circuit | Its own copper or fiber circuit into the building |
| Voice quality depends on | Jitter, packet loss, and how voice is prioritized on the link | The circuit itself; quality is inherent to TDM |
| Site failure behaviour | Calls reroute to another site or to mobiles in minutes | Calls fail with the circuit unless forwarding is pre-arranged |
| Number portability | Numbers move with the trunk; ranges are straightforward | Portable, but the range is delivered over that circuit until it is |
| Equipment required | An IP-capable PBX or a session border controller | A PBX with a PRI card and a channel bank if analog is involved |
| Fax and alarm lines | Workable but needs deliberate handling — T.38 or analog paths | Native; this is what legacy equipment expects |
| Cost model | Per concurrent call path plus usage, no separate circuit | Per circuit regardless of how many channels are used |
| Long-term availability | The direction the industry is moving | Varies by carrier and area — ask yours for its position in writing |
| 911 handling | Address registration per location, and per endpoint if any are nomadic | Tied to the physical circuit, so the address is inherently fixed |
A PRI gives you 23 channels. If you need 25, you buy a second PRI and pay for 46. If your call volume is seasonal — a processing operation at harvest, a clinic during an enrolment period, a retailer in December — you either carry the second circuit all year or you accept busy signals at your busiest moment. SIP trunking makes the same change a configuration item: add call paths for the season, remove them afterwards, pay for what you provisioned. For any organization whose call volume is not perfectly flat, this alone usually settles the comparison.
A PRI is immune to what the rest of the building is doing because it is a separate circuit. SIP trunking is not: calls now share a path with backups, hosted applications, and video. Voice is unusually sensitive to jitter and packet loss and comparatively indifferent to bandwidth — a call needs very little capacity, but it needs that capacity to arrive on a regular cadence. This is why organizations that move to SIP over a shared, contended internet connection often report that quality got worse, and conclude that SIP is the problem. The connection is the problem. Over a dedicated circuit where voice can be prioritized deliberately, SIP quality is generally indistinguishable from TDM.
When a PRI’s circuit fails, calls to those numbers fail with it, unless forwarding has been arranged in advance with the carrier — and that arrangement is often discovered to be untested at the moment it is needed. SIP trunks decouple the number from the physical location. If a site is unreachable, calls can be redirected to another site, to mobiles, or to a hunt group elsewhere, and that reconfiguration is measured in minutes rather than in a carrier ticket. For organizations where an unanswered phone is a direct commercial loss, this is frequently worth more than the recurring cost difference.
A PRI ties calls to a physical circuit, so the address emergency services receive is inherently correct — it is a property of the wiring rather than of a database. SIP breaks that coupling, and the address becomes something you register and maintain. In practice that means three things. Register the correct service address for every location, not one head-office address for the whole trunk. Decide what happens to endpoints that move — softphones on laptops and home-worker handsets are the well-known hazard, because a call can present an address the caller is nowhere near. And test it, by placing a supervised test call and confirming what the dispatcher sees, rather than assuming the configuration is right. Requirements around notifying on-site personnel and dispatchable location have tightened in recent years; treat the provider’s answer on how it handles them as a selection criterion rather than a detail to settle after cutover.
The most common cause of a stalled migration is the equipment nobody inventoried. Fax machines, lift emergency phones, fire alarm dialers, gate intercoms, postage meters, and some medical devices expect an analog line and behave badly on IP. None of it is insurmountable — T.38 handles most fax traffic, and analog terminal adapters or retained POTS lines cover the rest — but each device needs a decision before the cutover rather than after. Walk the building and list what plugs into a phone jack. That list, not the PBX, is usually what determines how complex the project is.
SIP trunking needs a PBX that speaks SIP, either natively or through a session border controller that translates and provides a security boundary. A modern IP PBX connects directly. An older TDM PBX with a working PRI card may need a gateway, and at that point the sensible question is whether to keep it at all — if the system is at end of support, replacing the trunk and the PBX together often costs less than bridging between them. That is also the point at which hosted PBX becomes worth evaluating against keeping a system on site.
The structural difference is that a PRI charges for a circuit and SIP charges for capacity. A PRI carries the circuit cost whether you use one channel or all 23, plus usage. SIP has no separate circuit — it rides a connection you already have and pay for — and charges per concurrent call path plus usage. Organizations well matched to a full PRI see a smaller difference than the marketing suggests; organizations carrying two PRIs to get 25 channels, or one PRI to use eight, see a large one. Number ranges, inbound and outbound usage profiles, and how many sites you are consolidating move the total more than the per-path rate.
If your PBX is stable, still supported, connected to a PRI that works, and your call volume is flat and close to 23 channels, there is no urgency — the honest advice is to plan the migration on your own timetable rather than treat it as a fault to fix. Equally, a site with a poor or contended data connection should fix the connection before moving voice onto it, not after. The migrations that go badly are almost always the ones where voice moved onto a link that was never specified to carry it.
We sell SIP trunking and hosted PBX, which should be read into this. For most organizations still on PRI the move is straightforwardly worthwhile — the flexibility and the failure behaviour are better, and the cost usually is too. The caveat we would insist on is the one that gets skipped: do not move voice onto a connection that was not specified to carry it. If the site runs on a contended shared service, fix that first. Where the underlying circuit is dedicated and voice is prioritized deliberately, we have not seen a quality argument for keeping a PRI. Where it is not, we would rather sequence the project properly than hand you a worse phone system.
Six questions that settle this choice faster than a feature table. The answer to each one tells you something the specifications do not.
Peaks favour SIP, where capacity flexes. Flat volume close to a full circuit narrows the gap considerably.
A dedicated circuit where voice can be prioritized is the right foundation. A contended shared connection is where SIP migrations get their bad reputation.
A modern IP PBX connects directly. An end-of-support TDM system may be cheaper to replace than to bridge.
Fax, lift phones, alarm dialers, and gates each need a decision before cutover. Inventory them first.
If you cannot answer this for your current PRI, that gap is itself an argument for SIP, where rerouting is a configuration change.
Ask directly. A published withdrawal date turns this from an optimisation into a scheduling exercise.
A PRI gives 23 simultaneous calls, so 23 paths is the like-for-like figure — but most organizations discover they do not need it. Look at your actual concurrent-call peak from carrier reporting rather than the channel count you have been paying for. It is common to find a full PRI covering a peak of ten, and equally common to find a business that has been quietly hitting its ceiling at month end.
Only if the connection underneath is wrong. Voice needs very little bandwidth but is highly sensitive to jitter and packet loss, so it degrades on a contended connection that tests fast. On a dedicated circuit with voice prioritized, quality is generally indistinguishable from TDM. Ask what the connection is before accepting that SIP itself is the variable.
Yes — numbers port to a SIP provider, including main numbers and DID ranges. Porting takes coordination and a firm cutover date, and it is the part of the project with the least flexibility once it is scheduled, so it is worth agreeing the sequence in writing before anything else is committed.
It stops being automatic and becomes something you configure. On a PRI the address is a property of the circuit; on SIP it is a registered service address that has to be right for each location and kept current when sites move. Endpoints that travel — softphones, home-worker handsets — need an explicit decision, because the failure mode is a call presenting an address the caller is not at. Place a supervised test call from each site before cutover and confirm what the dispatcher actually receives, and ask the provider how it handles dispatchable location and on-site notification.
They need a plan, not a leap of faith. T.38 handles most fax traffic reliably; alarm dialers, lift phones, and some medical devices are better served by analog terminal adapters or by retaining a small number of POTS lines. Inventory everything plugged into a phone jack before the cutover — that list is usually what determines the project’s complexity.
Describe what the connection has to carry and where. An engineer will tell you which service actually fits — including when the cheaper option is the right one.
Service availability depends on location, network proximity, capacity, and engineering review. Share an address and we will confirm what can be delivered there.