The Switchboard Does Not Need the Line

Anatel closed 2025 with 20,049,435 fixed lines in service. That is 12.9 per cent fewer than a year earlier, the lowest count since 1998, and 54.5 per cent below the 44.1 million peak reached in 2014. On that number alone, Brazilian business telephony looks like a business in liquidation.

Over the same eight years, the number of cloud switchboard seats in Brazil rises from 1.62 million to 4.05 million. Not 12.9 per cent down. Two and a half times up, at 12.11 per cent a year, while the value of the market goes from USD 177.33 million to USD 416.11 million.

Both numbers are correct, and the fact that they are compatible is the whole story. The fixed line was never the switchboard’s real dependency. It only looked like one.

The trunk became software

A PABX used to be a box in a cupboard with copper trunks arriving at it. The trunks were the scarce thing: a business had a fixed allocation of them, and the switchboard’s job was to share that allocation among extensions. When the box became a service, the trunk became a SIP registration — a credential, pointed at an address, riding the fibre the broadband provider had already provisioned for the internet connection the business was already paying for. The switchboard stopped needing a line the moment the trunk stopped being copper.

That is not a detail of network architecture. It changes which market this is. The measured decline of the fixed line is not a headwind the cloud market has to overcome; it is the condition that made the cloud market possible. Every line that dies is a trunk that becomes software, and software trunks cost the provider almost nothing to add. The same migration is visible across the wider enterprise stack, where moving a fixed capability into a subscription is the defining move of the decade in global digital transformation.

A market mostly outside its own base

What is left of the fixed base is unusually thin where it matters. Of Anatel’s 20.05 million remaining lines, 52.1 per cent are residential. That leaves 9.60 million business lines for a country with 23.9 million active small-business CNPJs at the close of 2025, a base that grew 9.7 per cent in a year after 5.1 million new registrations.

Divide one by the other and only 40.2 per cent of Brazilian small firms have a business fixed line at all. Which means roughly three in five of this market’s prospective buyers are not replacing a switchboard. They are buying a first one — or, more precisely, they are buying a business phone number for the first time in a form they can answer from a mobile handset.

This is why seat counts can triple inside a shrinking category without contradiction. The growth is not coming from displacing incumbents. It is coming from firms that were never in the fixed-line base to begin with, because a five-person company incorporated nineteen months ago was never going to order copper.

The price floor is a software licence

The published price of a Brazilian cloud extension now ranges from free to R$56.20, and the free end is not a promotional rate. 3CX licenses ten users at no charge and sells the SIP trunk separately. Zadarma gives five users away and sells Brazilian numbers from USD 3 a month. Nvoip is free to three agents. This is a market whose marginal price is zero because the software is open source and the number is the only genuinely scarce input.

The cheapest published Brazilian seat is R$19.90 per extension from Vozio, on a ten-extension minimum, so R$199 a month in practice. The market average for a bundled five-seat deal is R$56.20 per extension. That is a 2.82-fold spread for a product whose core function — receive a call, route a call — is identical in every case.

The spread is not explained by quality. It is explained by what is inside the price and how long the buyer will commit for:

• R$56.20 buys a bundle with an unlimited minute allowance, which is why Vivo Voz Negocio lists at R$50 a month and Claro Conecta Cloud at R$49.06 per line. On a 36-month term.
• R$19.90 buys a seat and nothing else. The minutes are a separate line on the bill, and so, at Nvoip, is the number, billed at R$29 to R$39 a month on top of R$29 to R$39 per agent.

So the market has not segmented on features. It has segmented on contract tolerance. A firm willing to sign for three years gets an incumbent bundle priced as a phone bill. A firm that will not gets a self-serve seat and assembles the rest itself. Those are two different products sold to two different buyers who happen to need the same function.

The arithmetic of a small ticket

Divide the market by its own seats and the revenue comes to USD 9.12 per seat per month in 2025, falling to USD 8.56 by 2033 — a 6.14 per cent decline while the seat count doubles. In reais, the seat price slides from R$46.80 to R$44.00.

Spread the whole market across the 23.9 million CNPJs and it is worth USD 7.42 per small business per year. That is the single most honest measure of how early this is. It also tells you where the money actually sits: not in the seat, which is being competed towards zero, but in the number and the minute, which are licensed, billed separately and priced with far less transparency. The number is also the component detaching fastest from hardware, as the eSIM transition is already demonstrating on the mobile side of the same subscriber identity.

Who is actually winning

Not the five operators. Claro holds 30.5 per cent of the fixed-line base, Vivo 25.0, Oi 19.1, Algar 4.0 and TIP Nacional 3.5 — 82.1 per cent between them, and all five sell a cloud PABX. Their product is competent and their price is not aggressive, because it cannot be. A R$200 to R$560 monthly ticket will not fund an enterprise sales motion, and an incumbent that matches a R$19.90 seat is competing against the line rental the cloud PABX exists to defend. Algar, with the smallest legacy base and the strongest regional small-business franchise, is the one incumbent with a rational path to fighting on the pure-plays’ terms rather than defending a bundle.

The channel that is winning is not a vendor at all. At 1 April 2025 Anatel had authorised 22,093 providers to offer fixed broadband — 11,892 under outorga, 10,201 under the dispensa regime — of which 5,951 report access data monthly and carry 95.4 per cent of all accesses. Every one of those has a fibre drop into the customer’s premises, an invoice already being paid, and no second product to sell. Business telephony is the only recurring line item it can attach without sending a technician, on a marginal cost per tenant that approaches zero. The metro fibre estates described in the Brazil data centre market are the same physical plant — the ISP that lands a building for connectivity has already paid for the ability to sell a switchboard into it.

The pure-plays, meanwhile, are not out-engineering anyone. They are winning the firms that will self-serve and the ISPs that will resell. Grupo C&M — the operator Vocom together with the Brazilian distribution of GoTo — disclosed 6,000 active clients in Brazil in September 2025 against a stated target of 7,000. Six thousand clients against 23.9 million CNPJs. That number is the clearest evidence available that nobody has won yet; no participant in this market even publishes a Brazilian seat count, which is itself informative.

Voice is the second channel

The reason the ordering of the product matters is that in Brazil the phone call is not the first thing a customer reaches for. 79.3 per cent of Brazilian companies already use WhatsApp in the business, 90.8 per cent use it for customer service, and 77.6 per cent expect it to remain their main channel. A cloud switchboard sold into that base is an integration sale into a messaging-first stack, and the vendor that built the messaging integrations first is selling a different product from the vendor that built a voice platform and attached WhatsApp afterwards.

Nvoip’s own description of itself makes the point in six words: plataforma omnichannel de telefonia e WhatsApp. Telephony and WhatsApp, in that order, from a company that started with neither. The incumbents started with the voice network — a real asset, and the wrong one to have started with.

What would make this wrong

Three things, and the first is in the same quarter as this article.

Voice agents eat seats. The productivity case for a cloud seat assumes a human answering a call. If an AI receptionist handles the first sixty seconds at a marginal cost near zero, a ten-person firm needs fewer seats, not more, and the seat — the unit this entire forecast is denominated in — becomes a worse proxy for the market than the number. The direction of travel in the AI market is towards exactly that substitution, and it would compress the volume line faster than it compresses the value line.

One assumption carries the model. The forecast assumes cloud seats reach 16.9 per cent of business lines in 2025 and 53.3 per cent by 2033. At 12 per cent the 2025 market is USD 126 million; at 22 per cent it is USD 231 million. That is a 1.83-fold spread on a single ratio, and it is stated on the page rather than buried, because it is the number most worth arguing with.

The exploited trunk is a liability. A SIP registration that anyone can point anywhere is a toll-fraud exposure, and an unsecured multi-tenant platform concentrates every customer’s call detail records in one place. In a market where the switchboard is sold as a low-cost add-on by 22,093 small providers, the security function is what gets skipped, which is the problem security orchestration and automated response exists to triage and which the DIY tier will not buy.

The durable position

There is one asset here that a free licence cannot replicate, and it is not the platform. It is the numbering authorisation — the outorga that lets a provider issue a Brazilian number and hold it. Seats can go to zero. The number is billed at R$29 to R$39 a month, separately, by companies selling the seat at R$19.90, and it is the line item nobody has yet found a way to commoditise.

Which returns the argument to where it started. In markets where voice is genuinely critical — mining, utilities, field operations — the buyer still pays for a specialist system rather than a commodity extension, and pays for it in a different market entirely, as the Argentine mining push-to-talk market and the wider critical communication segment both show. What is happening in Brazil is the opposite: telephony becoming an ordinary software line item attached to a bill that already exists.

The line is dying. The switchboard is the one piece of the old network that never needed it, and the firms winning it are the ones that were never in the old network at all. Full sizing, the three cases and the channel analysis are in the Brazil cloud PABX market report.