Telecoms.com Podcast Sep 21, 2026 1h 58m 1h 38m saved
With Jamie Davies, founder of The Content Store and former Deputy Editor of Telecoms.com
Jamie Davies has three explanations for why the AI labs have suddenly started asking to be restrained, and safety is not one of them.
The public account is that the people building these systems looked at what they were building and took fright. Davies starts somewhere else: a company's first duty is to make money for its shareholders and its investors, and asking for rules that cost time, money and resource does not obviously serve that.
"So, you have you have a responsibility to make money yet. You've got companies coming out making statements that seem to undermine that ability to make money."
Davies, founder of The Content Store and a former deputy editor of Telecoms.com, on the Telecoms.com Podcast, spent the six years since he left journalism running communications at Vodafone through its merger with Three and then at Nokia, which is where his reading of what companies say, and why, comes from.
The full episode is covered here so you can skip it. 118 minutes of audio, 20 minutes of reading.
Here are the 10 insights that matter.
Key Takeaways
Davies gives three commercial reasons for the AI industry's safety turn: protecting an IPO window, making a capex cut palatable, and regulatory capture
The programme's own preference is regulatory capture, because compliance cost always favors the incumbent over the new entrant
A warning that the technology could kill everyone doubles as marketing, one host argued, because nothing says powerful more cheaply
6G looks set to be built on refarmed 4G and 5G spectrum rather than new 6GHz bands, which are expensive and propagate badly
The technology that does it, MRSS, only works commercially if the operator keeps its existing radios, which locks in the 5G vendor
Open RAN did not produce multi-vendor networks and is unlikely to in 6G — the challengers that were meant to use it have gone
"It's like having a fifth gear in a car that you never use"
The same merger that was blocked as O2-Three was cleared as Vodafone-Three, and Davies says the market, not the argument, is what changed
Four operators in one area is overbuild, on his account, because the budget spent duplicating coverage is budget not spent extending it
Europe is broad and deep in network equipment and almost absent in models, cloud and devices
Deutsche Telekom's sovereign cloud is built on Nvidia chips, which the programme treats as the limit of European independence
Davies never uses AI to write and uses it as a sounding board and a research shortcut instead
Big Four firms have been caught shipping AI-written thought leadership, including work sourced from a blogger with about 117 followers
Telecoms operators that keep buying the cheap integrated option are choosing to become a utility, which is the argument Davies pressed hardest
1. Out Of Journalism, Into PR
Davies had not been on the podcast for six years, and the first stretch is what he did with them. He left Telecoms.com for Vodafone's communications team, recruited by a manager who did not want a conventional press officer.
He went from Vodafone to Nokia and then set up on his own.
What the new business actually sells
So it's called the content store. And my objective is always to say well how do you add an extra layer of storytelling around what you're doing as a tech company through editorial.
Jamie Davies
His complaint about in-house communications is that companies staff it with people who have never written for a reader, so the first question they ask is what the company wants to say rather than what anyone outside it would want to hear.
The reversal he says makes the difference
Well, that was that was fundamentally where we started when I joined Vodafone was, ultimately how do you translate the message that people outside the business care about as opposed to telling people why they should care about what you care about?
Jamie Davies
The hosts' version of the same point was that telecoms has a harder job than most industries at it, because what an operator sells is close to a commodity.
2. Why Vodafone-Three Passed
The Vodafone-Three merger went through while Davies was inside the company. The competition regulator had blocked a materially similar deal, O2 and Three, several years earlier, and the programme wanted to know what had changed.
Davies said two things did. One was the work, which he claims some credit for.
His own read on the campaign
I think did I think they did a better job. I think I think we did a better job.
Jamie Davies
The team segmented the argument by audience and built the message inwards from what each audience wanted, he said, rather than outwards from the deal. The second thing was the market itself: seven or eight years on, the cost of running a good network had risen while the pressure to hold prices down had not eased, which he said penalized anyone without scale.
The other half of the argument is about duplication. Asked how many networks a country should want in one place, the answer he arrives at is three, because the money spent building a fourth network over the same ground is money not spent covering somewhere else.
The overbuild trade, from a government's point of view
So from a big picture looking at it from a UK government you could probably say three concentrated in one area means they do have additional capacity to or additional budget to go elsewhere.
Telecoms.com Podcast
A host noted that the remedies attached to the deal are being policed closely, with the merged company measured every couple of years on how much spectrum it has actually put into use at each site. The risk he named is that those obligations do not flex: if equipment prices rise or the market turns, the building commitments stay and the margin absorbs it.
3. Selling A Commodity Pipe
The conversation turned to why telecoms advertising is so bad, and the diagnosis was that operators start from an engineering achievement and reverse-engineer a story onto it.
The example the programme kept returning to was an EE advertisement in which a man is shaved by a robot on a mountaintop. The reasoning behind it, one host suggested, was that low latency was supposed to be 5G's differentiator, so the brief became: illustrate low latency.
Davies' objection is that this starts in the wrong place, and he made the same point about BT's current campaign, which leans on heritage and the speaking clock. It builds trust, he said, without addressing the reason a customer would buy anything.
The counter-example he offered came from his own time at Vodafone. Someone brought him a network slice deployed at Glastonbury, where the site is an empty field for almost the whole year.
How you sell a network slice to somebody at a festival
Well, just position it in something that matters to them.
Jamie Davies
The narrative they built was about the bar queue rather than the network: a protected slice for payment authorization means less time queuing and more time in front of the stage.
The same principle, applied to consumer electronics, is why he thinks vacuum cleaners are sold on energy consumption when they should be sold on the electricity bill. A host's version was blunter: below a certain price point, buyers just respond to a bigger number.
4. How He Actually Uses AI
Asked whether he uses AI to write, Davies was categorical.
Where he draws the line
I never use it for writing.
Jamie Davies
What he uses it for is the thing a colleague across the desk used to do. Working alone, he types in an idea and asks the model to come back with small batches of questions that find the holes in it. The second use is research: he tells it what point he is trying to make, what kind of sources he will accept and which he will not, and gets back primary material that he says would have taken him two or three hours to find.
That is a stronger claim than the search-engine-on-steroids framing one host offered, and Davies accepted the distinction while refusing to extend it to the writing itself.
Why he says the output is unusable as prose
Now, as a as a as a writer, I think it is utterly awful. I think it lacks empathy. It lacks context.
Jamie Davies
Three pieces he had written recently were in three distinct styles, he said, and a model asked to do the same returns the same voice every time. A host's parallel complaint was about the summaries Google puts above search results: they read as though they are trying not to offend anyone, where the writers worth reading are the ones prepared to name a person and accuse them of something.
Davies' explanation was that a generic assistant has no idea who you are, and that the value arrives when a model is tuned to one user. The hosts' objection to that was immediate: personalization is how you get an echo chamber, and a system that agrees with everything is not a second opinion. Davies conceded the point on his own usage, saying he regularly has to tell the model to be more forthright and more objective.
5. The Big Four Got Caught
The failure mode both sides agreed on is publishing what the model produced without checking it.
What Davies says the output costs a company
And the biggest risk of this is number one, you sound generic and you create content which is just white noise
Jamie Davies
The second risk is the one with a cost attached. He said all four of the big accountancy and consulting firms have been caught using AI in published thought leadership, in one case citing a blog writer with something like 117 followers as a source, and that clients are paying thousands of pounds for it.
A host's own version came from asking a model how to restart a car that had not moved in fifteen years: the answer was comprehensive and step by step, and checking the sources showed that most of it came from one post on Reddit by one mechanic.
The part that does not get better with scale
Is that they produced it they didn't read it they didn't correct it they didn't fact check it they source check it
Jamie Davies
The hosts' side of it was less about risk than about time. One said it would take him longer to prompt a model, read the draft and rewrite it into his own voice than to start from a blank page.
6. Two Stacks, And Europe
The substantive part of the episode begins with a piece Davies published on LinkedIn called A Tale of Two Stacks, which argues that the world now has two technology stacks, one American and one Chinese, that can increasingly run without each other.
The idea is not new to him. He and the hosts had written about Harmony OS and the prospect of two operating-system ecosystems back around the time of the Huawei ban; that never amounted to much, but the chip work that followed did.
The question the piece asks is where Europe sits between them. Davies' answer is a chart plotting breadth, meaning how many companies a region has in a layer, against depth, meaning how competitive they are globally.
The one layer where Europe wins on both axes
So, for instance, you look at network vendors here, Europe is broad and deep.
Jamie Davies
Nokia and Ericsson are the reason, along with a small number of other suppliers, and the programme's addition was ASML, which sits at the far end of the depth axis and only halfway up the breadth axis because it is essentially one company with no peers.
At the other end, Europe has almost nothing in end-user devices, which the programme pointed out is now made in three countries. There is no European cloud of consequence, no European operating system, and on models there is Mistral and not much else, which even Davies would not claim is competitive at the frontier.
The structural problem he thinks that adds up to
So the biggest challenge Europe faces I think is that you've got a lot of highquality companies that are contributing to other stacks as opposed to the rest of the world contributing to a European stack
Jamie Davies
The worked example the programme brought was Deutsche Telekom's sovereign cloud and its Munich AI facility. The company has moved off reselling an American cloud and built its own stack underneath, and the one concession it makes is that the chips are all Nvidia's. Asked whether that is sovereign, the answer that comes back is that there is no alternative.
Davies read his own conclusion as a question rather than a prescription.
What he says Europe cannot have all of
Europe still wants leadership, independence, openness and unrestricted competition but it may not be able to hold all ambitions at once
Jamie Davies
His diagnosis is that Europe is holding the open-markets position on principle while the two largest players have abandoned it, America through tariffs and China through market access, and that the outcome of keeping the high ground on your own is that you lose.
7. 6G Runs On Old Spectrum
The second story on the agenda was the host's own, published six days before the recording under the title 6G is shaping up to be a big multi-vendor letdown. The technology at the centre of it is MRSS, multi-RAT spectrum sharing.
The set-up is about physics and money. The new spectrum earmarked for 6G starts at 6GHz, and the higher up the band ladder a signal goes the less far it travels and the worse it penetrates walls. The vendors' answer is more advanced radios, which means buying hardware.
Why the new bands are unlikely to carry the rollout
So, here's the implication. You're not going to do a big 6 GHz roll out because it's too expensive and it's not very useful spectrum. So, you're going to stick with you're going to use MRSS, but you're just going to implement it with the vendors that you've been using for 5G.
Telecoms.com Podcast
MRSS lets an operator split spectrum it already holds between 5G and 6G, which is why the programme treats it as far more important than DSS, the 4G-to-5G equivalent that never took off. Vodafone's head of research and development told the reporter that the whole point is not to replace the radios and not to replace the baseband equipment, and that MRSS has to work better than DSS did, because DSS damaged the experience for existing users.
Davies' memory of DSS was that in some markets it was a marketing device: refarm a low band, watch your 5G population coverage go through the roof, and deliver a 4G experience under a 5G label. The programme's parallel was a US operator branding 600MHz spectrum as 5G.
8. Open RAN Did Not Happen
The reason MRSS matters commercially is what it does to vendor choice. A Nokia white paper a couple of years ago floated the idea of using the 6G transition to bring in a second supplier: keep the existing radio unit, add a new distributed unit from someone else, and build an interface between the two.
That idea has gone nowhere. The Vodafone R&D chief's response to it, as relayed on the programme, was that introducing new hardware defeats the purpose of MRSS. Ericsson's comments for the story pointed instead to the open fronthaul interface, which allows a radio unit from one vendor to work with a distributed unit from another, and Nokia was downbeat on the original idea.
The programme's objection to the open fronthaul answer is that almost nobody buys equipment that way.
The line on open RAN compliance
It's like having a fifth gear in a car that you never use.
Telecoms.com Podcast
The evidence offered was the market rather than the standard: the analyst house Dell'Oro forecasts that the market will be single-vendor open RAN, the challengers that were supposed to take share have mostly gone, and AT&T's decision three years ago consolidated onto one supplier while being framed as an open RAN win. A telecoms analyst quoted on the show put the buyer's side of it as a fast-food analogy.
Why operators buy the whole thing from one supplier
You don't go you don't buy your burger at McDonald's, your fries at Five Guys, and your milkshake at Burger King, right?
Telecoms.com Podcast
Davies pushed back twice. His first objection was that the analogy is wrong, because a network lives for years: integrating a living room once and then living with best-of-breed components is closer to the real decision than assembling a single meal. His second was that open RAN was never only about hardware — disaggregation was always meant to open the software layer to specialist suppliers as well.
The programme's answer to that was that the incumbents control the software layer too, because the features a specialist application would sell can simply be built into the base station the operator is buying anyway.
Where the multi-vendor case stands on the evidence
6G I definitely think is going to be less open RAN than 5G even was
Telecoms.com Podcast
9. The Utility Endgame
Davies' strongest argument of the episode is what he thinks the vendor-consolidation path costs the operators themselves.
The choice he says operators are making without noticing
Like if you walk down a commoditized network route where you take the cheap option rather than the highly specialized option, you will never create highly specialized services for consumers to buy.
Jamie Davies
The alternative is a network with an intelligence layer in it, supplied by specialists, capable of hyper-local and customer-specific services. It may not work, he said, and the likely outcome is that it fails — but not trying has only one destination.
His word for it
you are just sleepwalking towards utilitization
Jamie Davies
The hosts' answer was that they have poured cold water on a decade of emerging-technology claims not out of hostility to trying things, but because the industry keeps rolling the dice without thinking the bet through — open RAN was sold on vendor diversity and never produced any.
Both sides agreed on where it ends up. UK consumers do not choose an electricity supplier on anything but price, and broadband is already close to that: the only variable is speed, and speed is in the price. Mobile is the one still in play.
10. Three Other Explanations
The last segment is the AI safety panic, and Davies' piece on it, published that day, is called three other explanations for AI's sudden mission to protect humanity.
His starting point is that a corporation owes its shareholders, and increasingly its private investors, a duty to make money, and that everything a company says in public should be read against that.
The contradiction the piece is built on
So, you have you have a responsibility to make money yet. You've got companies coming out making statements that seem to undermine that ability to make money.
Jamie Davies
Regulation costs time, money and resource, he said, and compresses a company's ability to experiment. His three alternative readings follow from it.
The first is the IPO calendar. Companies approaching a listing do not want the hype cycle to run out of stories before the float, because a share price that surges on momentum and then consolidates can consolidate below the issue price rather than above it.
The second is capital expenditure. If the revenue from end users is not matching what is being spent on GPUs, a company has to explain a slowdown to the market, and there are two available explanations.
Which of the two explanations a company would rather give
If you're going to sell this to the market, do you tell the market, "Sorry, the customers aren't buying this." or do you tell the market AI is terrifying. Let's pause for the moment and regroup in a couple of months.
Telecoms.com Podcast
One is a problem that can be fixed, in other words, and the other is not — and a pause that reduces GPU spending also produces better margins going into a listing.
The third is regulatory capture.
How the compliance burden lands on a competitor
Slam a load of a load of regulation and compliance around early developing AI models that you didn't have to deal with. Then all of a sudden it takes 2x plus years to get to where you are today.
Jamie Davies
That buys the incumbent time to build the ecosystem, the customer relationships and the infrastructure that a model on its own does not give you. The programme's own ranking put regulatory capture first, on the standard argument that compliance bureaucracy favors large companies over small ones and incumbents over entrants, with the capex explanation close behind.
There was a fourth reading on the table, offered by a host as the counter-intuitive one.
The marketing value of an extinction warning
if you say it's going to kill everyone in the whole world, that's a sort of perverse form of hype
Telecoms.com Podcast
Davies added the older objection that the line between scientific experimentation and reckless exploration was crossed years ago, and that the open question is how far past it the industry has travelled while unregulated.
The programme did not leave it there. One host said the substance of what the recent leavers are actually warning about is narrower and more credible than the headlines suggest.
What he thinks the real worry is
It's just that the technology that we currently have is getting to a point where it can escape the sandbox quite easily whether or not it was told to or not.
Telecoms.com Podcast
The disputed example was the Hugging Face incident and whether the systems involved acted on their own initiative or were instructed to; the programme could not settle it in the time left and said so.
Bonus Insights
The rebrand that lost a Google image search
The EY rebrand from Ernst & Young came up as an example of the discipline going wrong. One host recalled that for the first couple of months after the launch, putting those two letters into Google image search returned a magazine with an entirely different audience, and wondered aloud whether it was really an accident.
An AI biography of a telecoms figure that put him at Amazon
The hallucination example that landed was a model asked who a well-known telecoms engineer is, which got the career right and then placed him at Amazon.
Podcast length, and why this one runs two hours
Davies' advice was a hub-and-spoke model: keep the long conversation and cut specific segments out of it, so a listener can take fifteen minutes on one subject or the whole thing. The show already has a colleague working on clips. The hosts' defence of the long format was that a podcast is a recorded conversation rather than entertainment, and that pub conversations wander.
The drinking rule of podcast dynamics
One host's theory, delivered at length: after a couple of pints there is no difference, and after four the drinker becomes looser, more tangential and more inclined to interrupt. That works when everyone is at the same level and fails when one person is structured and the others are not.
Who they will not book
The host who does the bookings said the test for an incoming pitch is whether the guest has listened to the show. A chief executive of a large company would be a coup on paper and a bad episode in practice if they stayed on message.
A bottle of Pappy Van Winkle
A previous guest brought the show a bottle of Pappy Van Winkle, which a friend of the podcast later told them is close to impossible to get hold of. They tasted it on air, agreed not to finish it before she comes back, and declined to say what it costs.
Davies' bottom line is that the industry's two loudest stories are best read as commercial documents: 6G will be built with the equipment operators already have and the suppliers they already use, and the AI labs asking to be regulated are doing something a company does when a listing, a capital budget or a competitor is the actual problem.
Products, Companies & Tools Mentioned
Vodafone and Three UK (The merger Davies worked on from inside Vodafone's communications team. He says the market changed between the blocked O2-Three deal and this one, and that the campaign was built outwards from what each audience wanted)
The Content Store (His own agency, selling editorial storytelling to technology companies rather than conventional public relations)
Nokia and Ericsson (The two suppliers the programme says will build 6G. A Nokia white paper floated bringing a second vendor in at the MRSS transition; Ericsson's answer on multi-vendor is the open fronthaul interface)
ASML (The programme's example of European depth without breadth: one company, no peers, at the far end of the competitiveness axis)
Mistral (Named as Europe's best effort in models, and conceded not to be at the frontier)
Deutsche Telekom (Its sovereign cloud and Munich AI facility are the test case for European independence, and the chips in it are all Nvidia's)
Nvidia (The dependency that the programme says no European sovereign-cloud project has been able to design out)
Huawei (The chip launch that prompted Davies' two-stacks piece, and the ban that first raised the prospect of separate ecosystems)
Dell'Oro Group (Its forecast that the market will be single-vendor open RAN is the evidence the programme leans on hardest)
AT&T (Its consolidation onto one supplier three years ago, framed at the time as an open RAN deployment)
Mavenir and Parallel Wireless (Named among the open RAN challengers whose attempts to become RAN players did not work out)
BT and EE (The two campaigns the programme uses as examples of telecoms marketing that starts from the company rather than the customer)
OpenAI and Anthropic (The two companies whose approaching listings sit behind Davies' first explanation for the safety turn)
Hugging Face (The incident the programme treats as the live argument about whether AI systems acted on their own initiative)
Books & Resources Mentioned
6G is shaping up to be a big multivendor letdown – Iain Morris, Light Reading (The story the 6G segment is built on, published six days before the recording, carrying the Vodafone and Ericsson comments the programme quotes)
A Tale of Two Stacks – Jamie Davies (His LinkedIn piece on the American and Chinese technology stacks and where Europe sits between them, complete with the breadth-and-depth chart the programme walks through)
Three other explanations for AI's sudden mission to protect humanity – Jamie Davies (The LinkedIn piece published the day of the recording, setting out the IPO, capex and regulatory-capture readings)
Listen to the full episode
🔴 YouTube | 🔗 Episode page
Watch the full episode:
If this was worth your time, send it to someone closer to the industry than you are.
Get the latest market chatter as it happens: