Virgin Media O2, TalkTalk and Who Pays for the Next Generation of UK Cellular

What we actually know

On Saturday 31 October, Sky News reported, citing people close to the company, that Virgin Media O2 is preparing to cut thousands of jobs. Up to 4,000 of its roughly 16,000 staff could be affected, with non-customer-facing roles most at risk, alongside a review of what it spends on contractors and suppliers. The final number has not been settled. VMO2 declined to comment, and a cost-reduction target in money terms, rather than a headcount, is expected alongside its third-quarter results on Tuesday.

The £22 billion debt figure doing the rounds is real, but it is not new. It is the pile VMO2 has carried since the 2021 merger. What changed this year is that the bond market got nervous. Over the summer, VMO2’s roughly £1.1 billion of senior unsecured debt sold off hard, with one $925 million bond falling as low as 57 cents on the dollar. The owners, Liberty Global and Telefónica, responded by looking for around £600 million of cost savings, as the Financial Times reported in September.

The job cuts are the noise. Capex is the signal.

Headcount stories get the clicks. For anyone running connected devices, the number that matters is capital expenditure, because capex is what builds masts, upgrades cores and extends coverage. The owners’ £600 million target was always expected to come from a mix of job cuts and lower capital spending.

We have already had a taste of what that looks like. On 30 October the FT reported that VMO2 is nearing a deal to sell Openreach fibre to new customers outside its own footprint, and is pausing the upgrade of its own cable network to full fibre as part of the wider cost drive. That follows the CMA signalling it was not keen on nexfibre buying YouFibre, a deal linked to a £1.1 billion transfer to VMO2 that would have helped with the debt.

Put that next to O2 Satellite and a pattern emerges. In the sky, VMO2 rents Starlink. In the ground, it may soon rent Openreach. I don’t think that is daft. O2 Satellite made VMO2 the first operator in Europe to switch on satellite data direct to phones, and it lifted O2’s UK landmass coverage from 89% to 95% for a £3 a month bolt-on. That is a genuinely clever bit of capital-light coverage. But a company renting other people’s infrastructure at both ends is a company deciding where it will not spend its own money, and IoT buyers should be watching which places those turn out to be.

Three operator groups, three investment stories

Since VodafoneThree completed, the UK has had three mobile network groups. Right now they are telling three very different stories about money.

VodafoneThree is the mirror image of VMO2. It is spending, and that spending is a regulated commitment attached to the merger approval: an AI-ready 5G Standalone network reaching 99.95% population coverage by 2034. Whatever you think of the merger, it gives IoT buyers something rare in this industry, a published build plan with a regulator watching it.

Will the new technology earn its keep?

This is the question underneath all of it, and nobody in the industry likes answering it out loud. Every one of these programmes is a bet that the next generation of services will pay back the money spent building it. Look at the horizons involved. VodafoneThree’s plan runs over ten years. Its growth targets stretch to FY32. A mast built in 2027 has to earn its keep well into the 2030s, on a network that will be talking about 6G before it has finished paying for 5G.

And the honest history is not encouraging. Non-Standalone 5G, the version most of us have used since 2019, was largely sold as “4G but faster”. Customers took the speed and declined to pay much extra for it. The services that were meant to justify the spend, slicing, guaranteed performance, industrial 5G, all need Standalone, and Standalone is only now reaching scale. So the industry is being asked to fund the second half of 5G before the first half has obviously paid for itself.

Here is how I would score the next wave of services on whether they will genuinely pay their way. Treat the timescales as the industry’s own expectations, not promises.

Notice the pattern in the satellite row. The service that looks most likely to pay its way quickly is the one where somebody else built the expensive bit. That is exactly why it appeals to an operator under pressure to cut capex, and exactly why I don’t want it mistaken for a substitute for building the ground network.

The TalkTalk lesson: what customers expect, and how cheaply

Earlier this month we got a sharp reminder of what happens at the other end of this equation. On 5 October BT bought TalkTalk’s consumer business and its PlatformX wholesale arm out of administration, for around £400 million and on a debt-free basis. TalkTalk had been turning over roughly £1.2 billion a year, was loss-making, and had been carrying well over a billion pounds of debt. The government issued a Public Interest Intervention Notice, with Lisa Nandy warning that a disorderly collapse could put lives and public services at risk, and the CMA was given two weeks to report. It handed that report over on 19 October. VMO2’s chief executive, Lutz Schüler, has since called for a full in-depth investigation, arguing that two weeks is no way to review a deal of that size.

Yes, TalkTalk was a consumer broadband story. But it asks exactly the question this whole piece is circling. TalkTalk was built to make connectivity cheap, and it did. Customers took the cheap price, as customers do. Then the debt stayed while the customers drifted, and the business could not fund itself. The moment it wobbled, the same connectivity everyone had been haggling down to the last pound was suddenly described as critical national infrastructure that could not be allowed to fail.

That is the contradiction at the heart of UK telecoms. We expect connectivity to behave like a utility, always on, everywhere, resilient. We want to pay for it like a commodity. Operators then borrow to bridge the gap, and when the borrowing gets expensive, as it has for VMO2 this year, the bill turns up as job cuts and paused builds.

IoT is no different. I have sat in plenty of procurement meetings where the conversation about a ten-year device estate came down to pennies per SIM per month. Fair enough; nobody should overpay. But the same buyers expect 5G SA, RedCap, slicing and satellite fallback to be there when they want them, on a network that somebody else has paid to build. Somebody has to fund that. If it is not the customer, it is the bondholder, and bondholders have just shown what happens when they lose their nerve.

The kit is getting dearer too

If connectivity pricing is being squeezed from one side, hardware is being pushed from the other. The AI data centre boom has swallowed memory manufacturing capacity, and everything else with a DRAM or flash chip in it is paying the price. Gartner expects combined DRAM and SSD prices to rise by around 130% over 2026, and Bernstein reported memory contract prices up nearly 20% quarter on quarter in Q3, with another rise expected in Q4. In the US, the cable industry body NCTA says memory has gone from roughly 3% of a router’s manufacturing cost to more than 20% in a year. Industry forecasts have the shortage running into 2027.

Routers, gateways and modules are not immune. Teltonika partners have been told to expect price increases of up to 20% from 1 November, and I would be surprised if Teltonika is the last vendor to move. Add UK inflation, which rose to 3.5% in September from 3.1% in August, pushed up by fuel and goods prices, and the cost of putting a connected device in the field is rising at exactly the moment everyone is being told to cut.

And so is keeping it safe

Then there is security, which is becoming a bigger line on the bill every month. On 29 October, attackers drained around $6.3 billion from Binance wallets in roughly 20 minutes, the largest crypto theft on record. Blockchain analysts TRM Labs and Chainalysis have attributed it to TraderTraitor, a North Korean state-backed group. Binance says its initial investigation points to one engineer’s compromised laptop. One endpoint. That is all it took.

Crypto exchanges are not IoT estates, but the lesson travels. Every cellular router sitting in a cabinet with a public-facing management port is an endpoint, and most of them are looked after far less carefully than a Binance engineer’s laptop. The day after the Binance attack, Teltonika disclosed a critical vulnerability, CVE-2026-82104, affecting RutOS 7.22 to 7.24.2 and TswOS 1.9 to 1.10. It could let an unauthenticated attacker bypass security checks through the HTTPS port and potentially run arbitrary commands. The fix is in RutOS 7.24.3 and TswOS 1.10.1. Earlier in October, the RutOS 7.24.6 release had already patched a long list of CVEs, including two rated critical.

To be fair to Teltonika, this is what good looks like from a vendor. It holds IEC 62443-4-1 certification for its secure development lifecycle, publishes advisories, and ships fixes quickly. The problem is rarely the vendor. It is the estate that never gets updated: the router installed five years ago by a contractor who has since moved on, still running whatever firmware it shipped with, with its web interface open to the internet.

Fixing that costs money. Someone has to own patching, run a remote management platform, test updates before pushing them to thousands of devices, and occasionally send an engineer to a site where an update has gone wrong. Regulation is adding to the bill as well, with the UK’s PSTI rules already in force and the EU Cyber Resilience Act’s vulnerability reporting duties now applying to manufacturers. None of that shows up in a “pennies per SIM” procurement spreadsheet. All of it is real.

So put the whole picture together. Hardware is getting dearer. Security is getting dearer. Networks need billions to finish 5G Standalone. And customers, consumer and industrial alike, still expect all of it cheaper every year. Something in that equation has to give.

Why RedCap is where the squeeze bites

5G RedCap only works on a 5G Standalone core. No SA, no RedCap; the device falls back to LTE and waits. eRedCap, the Release 18 variant aimed squarely at replacing LTE Cat-1, needs exactly the same SA infrastructure, with modules arriving through 2027 and 2028.

So the operator under the most pressure to cut capital spending is also the one with the most 5G Standalone left to build. That is not a coincidence I enjoy. If VMO2’s SA rollout slows, an O2-only IoT estate waits longer for RedCap, longer for slicing, and longer for the 5G core features that make the newer device classes worth paying for. EE and Vodafone customers get there first. On our UK RedCap tracker that gap was already visible in May. Tuesday will tell us whether it widens.

Satellite gets cheaper while the ground gets harder

Direct-to-device satellite has gone from conference slideware to a consumer bolt-on in under a year. O2 Satellite went live in February on Starlink and added iPhone support in June. VodafoneThree started customer trials of its own service in summer 2026 through Satellite Connect Europe, the Vodafone and AST SpaceMobile joint venture formerly known as SatCo. Two operators, two constellations, two different technical bets.

For IoT, I would keep two things firmly separate, as I argued in my O2 Satellite analysis. Phone-grade direct-to-cell is a coverage fill-in for handsets and does not run on routers or modules today. Standardised 3GPP IoT NTN, NB-IoT and LTE-M over satellite, is the path that actually suits remote sensors, and it needs NTN-capable hardware. Both are covered in our NTN and satellite hub.

My worry is a different one. Cheap satellite coverage is a very attractive line on a slide when you are being asked to cut terrestrial capex. “95% landmass coverage” reads beautifully in a results deck. It does not tell you how a remote pumping station performs on a busy evening, or whether that site will ever see 5G SA. Satellite should sit on top of the terrestrial build, not stand in for it.

What I would do if I ran an IoT estate on O2

• Don’t panic-migrate. Nothing in this news affects how your SIMs work this month. Moving thousands of devices in a hurry costs far more than it saves.
• Do review single-network exposure. If your whole estate sits on one operator, this is a good week to ask whether it should. Multi-network SIMs or SGP.32 eSIM give you options without a truck roll. My eSIM vs plastic SIM breakdown covers the practical differences.
• Specify RedCap with LTE fallback, then check SA by operator. Buy hardware that will outlast this cycle, but don’t assume every network reaches 5G Standalone at your sites on the same timetable.
• Treat satellite as two separate decisions. IoT NTN hardware for low-data remote sensors now. Direct-to-cell for modules later; plan for it, don’t bet on it.
• Check your firmware this weekend. If you run Teltonika kit on RutOS 7.22 to 7.24.2, update to 7.24.3 or later, and close any management ports that don’t need to face the internet. See Teltonika’s Security Centre for the advisory.
• Budget for dearer hardware. Price rises and memory-driven lead times are likely to run into 2027. If you know you will need the kit, buying ahead of the next increase may be cheaper than waiting.
• Be realistic about price. The cheapest SIM tariff is cheapest right up until the network behind it stops investing. Ask suppliers what they are building, not just what they are charging.

The verdict

The £22 billion is real, and thousands of jobs are reportedly on the line. But the debt is not the story, and neither is the headcount. The story is that UK cellular has split into an operator spending heavily under a regulated commitment, an incumbent pressing on steadily while mopping up a failed rival, and an operator under pressure to spend less while it still has 5G Standalone to finish. Underneath all three sits the question nobody has answered yet: will the next generation of services pay for itself, when customers, consumer and industrial alike, have spent a decade being trained to expect more for less? Meanwhile the hardware costs more, keeping it secure costs more, and the attackers are getting better. TalkTalk showed what happens when the sums stop adding up. Check back on Tuesday. I will.