Quick answer: VoIP is the outcome, meaning voice carried over data networks instead of copper. SIP is the signaling protocol that sets up, changes, and ends those sessions. They are not competing options. Asking whether to choose one over the other is a bit like asking whether to choose email or SMTP, since almost every modern deployment uses both together. The decision hiding behind this search is nearly always SIP trunking versus hosted VoIP, and that one is worth thinking about properly.
I would rather spend this piece on the real choice than pad out a differences table for two things that are not alternatives. Some of what follows will contradict the comparison articles you have already read, which is deliberate.
What Each Term Actually Covers
VoIP names an outcome
Voice over Internet Protocol describes the technology that delivers calls across data networks rather than the switched copper network. VoIP is a category, not a product, and it covers everything from a consumer app to carrier-grade equipment serving thousands of seats.
Nothing in the term specifies how a session gets established, which codec carries the audio, or who owns the equipment. Those are implementation choices sitting underneath the label. VoIP is best read as the answer to “what happens” rather than “how”.
Worth clearing up one persistent muddle here: the claim that VoIP only supports voice is not really accurate either. Most deployments carry video, messaging, and presence over the same infrastructure, and the name has simply outlived its narrow origins.
The internet part of the name is doing less work than people assume, too. Plenty of business voice never touches the public internet at all, travelling instead across private circuits or a carrier’s own network. What the term really means is that the audio is packetized and routed rather than carried on a dedicated pair of wires, which is a statement about method rather than about which network the packets cross.
SIP names a method
The Session Initiation Protocol was published by the IETF as RFC 3261 in 2002. It handles location, invitation, negotiation, and teardown. When somebody dials, SIP finds the far end, proposes a session, agrees the parameters, and gets out of the way while the media flows over a separate transport.
Two things follow from that, and both matter for the decision further down.
- SIP carries no audio itself. Media travels over RTP, a distinct protocol that the SIP exchange merely arranges.
- SIP is not voice-specific. The specification talks about sessions with participants, and those sessions can be anything: a phone conversation, a video meeting, a messaging exchange, an instant message thread.
That second point explains why the protocol survived. It was designed as a general session mechanism and the industry bolted voice onto it, rather than the other way around.
If you want a picture of what happens when you dial, it goes roughly like this. Your device sends an invitation naming who it wants to reach and what kinds of media it can handle. Servers along the way work out where that person currently is. The far end answers with what it can handle. Both sides settle on a codec, the audio starts flowing directly between them over the internet or whichever network sits between, and the signaling goes quiet until somebody hangs up. Every call you make on any modern setup follows that shape, whether the call lasts ten seconds or an hour.
Why This Comparison Is Framed Wrong
Search results for this query are full of tables putting the two side by side as though you might pick either. I understand why: the search volume exists, and somebody has to answer it. But the framing quietly teaches people something untrue.
Here is the more useful version. Roughly every business voice deployment sold today runs on SIP. Your hosted provider uses it. Your desk handsets register with it. Your softphone speaks it. The VoIP SIP relationship is not a fork in the road; one is the general category, the other is the near-universal method of arranging it.
So when somebody asks whether they should get SIP or VoIP, what they usually mean is one of these three questions:
- Should I keep my existing phone system and change how it reaches the outside world?
- Should I replace the whole thing with something hosted?
- What is this SIP trunk quote I have been sent, and is it a good deal?
Those are answerable. The abstract comparison mostly is not, and I think the volume of content answering it anyway explains why so many buyers arrive at procurement holding a confident but incorrect mental model.
Side-by-Side, Since You Came For One
| Aspect | VoIP | SIP |
| What it is | A category of delivery | A signaling protocol |
| Defined by | Common usage | RFC 3261, published 2002 |
| Carries audio | Yes, through the media stream | No, it arranges the media stream |
| Scope | Voice, and usually more | Any session type with participants |
| Can exist without the other | In theory, using other protocols | Only as a way of doing VoIP |
| What you buy | A service or a system | A trunk, or a capability inside one |
| Relevant question | Which delivery model? | Which provider and how many channels? |
The row that matters is the fifth one. Other signaling methods exist and a handful remain in use, but the practical answer in 2026 is that these two travel together.
Everything your team needs in one platform
The Decision People Actually Mean
Now the useful part: keep what you have and change the connection, or replace everything.
When SIP trunking fits
A SIP trunk replaces your old ISDN or analogue circuits with a data connection into your existing switch. The phone system stays. Extensions, routing rules, and handsets carry on unchanged. Only the path to the outside world changes.
This suits you when:
- You have an on-premise PBX with life left in it and no appetite for replacing it yet
- Someone in-house understands that equipment and enjoys running it
- You need specific integrations already built against the current setup
- Regulatory or contractual terms tie you to particular equipment
- Your volume is high enough that per-channel pricing beats per-seat pricing
SIP trunking is also the least disruptive option available. Number ranges stay, dial plans stay, staff notice nothing. That is genuinely valuable if you are working to a deadline.
One caveat I would raise anyway. Choosing trunking to avoid a replacement decision often just postpones it by two or three years, and the equipment tends to be older and harder to support by then. That is a legitimate choice if you have a reason, and a poor one if the reason is simply that nobody wants to run the project. Ask yourself honestly which applies before signing a three-year term on the trunk.
When hosted VoIP fits
Hosted VoIP removes the switch entirely. Your provider runs everything; your handsets and softphones register directly to it. Nothing sits in a cupboard.
This suits you when:
- Your current equipment is due for replacement anyway
- Staff work from several locations, or from home
- You want new capabilities without a hardware project every time
- Nobody in-house wants to own a switch
- You would rather have a predictable per-seat figure than manage capacity
A hosted arrangement also changes who answers the phone when something breaks, which sounds trivial and is not. With trunking, a fault could sit in your switch, in the trunk, or in the network, and establishing which takes time nobody has during an outage. Hosted arrangements collapse that to one number to ring.
Most operations under a few hundred seats end up here, and I would say that is the correct default in 2026 rather than a fashion. The maintenance burden of on-site equipment rarely justifies itself at that size. Our comparison of cloud and on-premises deployment covers that trade in more depth.
There is a middle option nobody mentions: hosted providers that also accept trunks from a separate carrier. You get the software as a service and keep your own wholesale voice pricing. Worth asking about, because it occasionally saves a great deal on international routes.
What SIP Does That Voice Alone Does Not
Since the protocol was built for sessions generally, it does things a voice-only framing hides.
Portability. A SIP trunking arrangement is arguably the least proprietary interface in this industry. Providers differ on price, quality, and support, but the interface is standardized enough that moving is a configuration exercise rather than a rebuild. Compare that with almost any other vendor relationship you have.
Multiple media in one session. Adding video to an established session is a renegotiation, not a new connection. Messaging works the same way. This is why the same signaling handles what marketing calls unified communications.
Presence and registration. Knowing which device a person is currently reachable on is a SIP function, which is why your desk handset and softphone can ring together without any additional system coordinating them.
Interoperability across vendors. Handsets from one manufacturer register to a system from another. That sounds unremarkable until you remember how proprietary this industry used to be, and how much of your other software still is.
Our explainer on voice protocols goes deeper into the signaling and media split for anyone who wants it.
The Deadline Changing the Question in 2026
Something is forcing this decision for a lot of organizations right now, and it deserves more attention than the comparison tables give it.
Openreach is retiring the UK’s public switched telephone network on 31 January 2027. As of mid-2026 the operator estimated around 1.5 million lines still on legacy copper, including roughly 350,000 business premises, and has stated the date will not move again. Similar programs are running or completed across other markets.
The trap is treating it as a like-for-like swap of landline phone calls onto an internet-powered alternative. It is not, because those copper lines carry more than conversations: payment terminals, lift alarms, door entry, security panels, and other equipment somebody installed years ago and nobody has inventoried since. Openreach has been explicit that these dependencies are what catch organizations out.
If this applies to you, audit every circuit before choosing between trunking and hosted anything. The voice part is straightforward. The forgotten alarm line in a stairwell is what generates the emergency in January.
Cost, and Where Estimates Go Wrong
| Cost element | Trunking approach | Hosted approach |
| Up-front | Low if the switch stays, high if it needs replacing | Handsets or nothing at all |
| Recurring | Per channel, plus usage | Per seat, usually bundled minutes |
| Maintenance | Yours, including hardware and updates | The provider’s |
| Scaling | Order channels, sometimes with notice | Add or remove seats |
| Hidden cost | Support skills you must retain | Add-on tiers for advanced features |
| Predictability | Good, until something breaks | High, unless usage varies wildly |
Two estimating mistakes turn up repeatedly.
The first is forgetting the cost of the person who maintains on-site equipment. It rarely appears in a comparison because it is a salary rather than an invoice, but a switch nobody can service is a liability rather than an asset.
The second is sizing channels by headcount. Concurrency is what matters, and a fifty-person operation almost never needs fifty simultaneous paths. Look at your busiest hour last quarter and add headroom to that instead. Guessing high is one of the more common ways a trunking quote loses to a hosted one on paper when it should have won.
A third, less obvious: comparing bundled minutes against metered usage without modelling your actual destinations. Bundles usually cover domestic traffic generously and international traffic poorly. If a meaningful share of your outbound goes abroad, price that separately before deciding, because the headline figure will mislead you in whichever direction the provider’s bundle happens to favor.
What to Ask a Provider
Bring these to any evaluation. They separate the serious options quickly.
- Which signaling and media protocols do you support, and can we bring our own devices?
- Is number porting included, and how long does it take in our regions?
- What happens to an inbound call if our connection fails, and how fast does failover happen?
- Are emergency services correctly registered for each site address?
- Do you charge per channel, per seat, or per call, and what triggers a tier change?
- Which features sit behind an upgrade, and which are included?
- What are the notice period and exit terms, and do we keep our numbers?
- Where does media terminate for our locations, and does that affect quality?
The failover question is the one people forget until the day it matters. A trunk into a single connection with no secondary route is a single point of failure for every conversation your business depends on, and the fix is cheap if you arrange it before rather than after.
Ask for the answers in writing, ideally against your own scenarios rather than in general terms. Sales conversations produce optimistic verbal answers about redundancy that turn out, on inspection, to describe an option nobody included in the quote. A short document naming what happens on a specific bad day is worth more than an hour of reassurance, and any provider worth using will produce one without complaint.
Frequently Asked Questions
Can you have VoIP without SIP?
Technically yes. Alternative signaling exists, including proprietary methods used by some consumer applications and older protocols such as H.323 that still run in a few legacy environments. In practice, business deployments sold today are overwhelmingly SIP-based, so the theoretical possibility rarely affects a purchasing decision. If a provider proposes something else, ask why, and ask specifically what happens to your ability to move elsewhere later without replacing every handset you own.
Are IP phones and SIP phones the same thing?
Mostly, and the terms get used interchangeably. IP phones are capable of running various signaling protocols, while the SIP label indicates that specific standard. Since nearly all current handsets use it, the distinction matters only with older equipment or vendor-locked devices. When buying, confirm the handset supports open registration to any compliant provider rather than only the manufacturer’s own service, since that single detail determines whether the device can move with you later or becomes scrap.
How many channels does a small business actually need?
Base it on concurrent conversations, not headcount. A common starting ratio is one channel for every three to four staff who handle calls regularly, adjusted for your traffic pattern. Sales floors making outbound calls need proportionally more; offices where phones ring occasionally need fewer. Review your busiest hour over the past quarter, add roughly 20% headroom, and revisit after a month of real data rather than committing to a guess.
Does the choice affect emergency calling?
Yes, and it deserves attention during setup rather than afterward. Because sessions travel over data networks, the physical location is not inferred from a copper pair, so each site address must be registered correctly with your provider. Requirements differ by country and some jurisdictions impose specific obligations. Remote staff complicate this further. Confirm how your provider handles address registration, what happens when somebody works elsewhere, and how the arrangement is tested.
What actually happens during a migration?
Numbers are ported, which takes anywhere from days to several weeks depending on the range and the losing provider. Meanwhile you configure the new setup, register devices, and rebuild routing. Sensible projects run both paths in parallel and cut over gradually rather than overnight. The steps that cause trouble are rarely the voice ones: fax lines, alarm circuits, and payment terminals attached to the old service need their own plan.
Is quality worse than a traditional line?
Not inherently, though it depends on your connection rather than on the technology. Copper delivered consistent, unremarkable audio. Data networks can deliver considerably better quality with wideband codecs, or considerably worse when jitter and loss go unmanaged. The variable you control is network behavior under load, not the protocol. Prioritize voice traffic, measure latency while the link is busy, and quality generally stops being a topic anyone raises at the management meeting.
Working out which arrangement fits your operation?
Voiso runs a cloud platform with global number coverage, regional media placement, and published infrastructure detail, so you can see how sessions travel before committing. Talk to the Voiso sales team about your concurrency, your regions, and what still sits on legacy circuits, and ask them the failover question above.
Sources
- Rosenberg, J. et al. (2002). SIP: Session Initiation Protocol. IETF RFC 3261: https://www.rfc-editor.org/info/rfc3261/
- Openreach, Openreach warns businesses just six months left before the UK’s old phone network is switched off (31 January 2027 retirement; approximately 1.5 million lines and 350,000 business premises remaining): https://www.openreach.com/news/openreach-warns-businesses-just-six-months-left-before-the-uks-old-phone-network-is-switched-off/