- Narvinger takes charge as Ericsson’s Networks business returns to growth
- Network APIs offer a route into the emerging industrial AI economy
- Geopolitical fragmentation and uncertain returns complicate the opportunity
STOCKHOLM — To Sweden, then, to meet the new man at Ericsson. Per Narvinger took over as CEO on Oct. 1, and I’m the first journalist to interview him in person since his ascension. It’s a week of firsts: FNTV is also the first media outlet allowed a tour of Ericsson’s future campus in Stockholm’s Hagastaden district.
So, the company has a new boss and a new home on the way. The connection turns out to run rather deeper than a change of address.
But first, the boss.
Narvinger is a tall, handsome fellow (he is Swedish, after all) with a boyish charm at odds with his shock of white hair. An Ericsson veteran of 30 years, he started on the engineering side of the house and arrives in the top job with a reputation for calm pragmatism. After interviewing him, I can confirm that neither quality seems to have deserted him on promotion.
This is quite an achievement. He now runs a business with roughly $24 billion in annual sales and responsibility for building much of the infrastructure for the next great digital industrial revolution. You might expect some visible signs of strain. But no, Narvinger convincingly gives the impression that this is all perfectly manageable.
He arrives in the job at a time of mixed fortunes for Ericsson. On the good news side of the ledger, the company is celebrating its 150th anniversary and remains one of five incumbent mobile network vendors, alongside Nokia, ZTE, Huawei and Samsung. That buys it a seat at the table of the new digital industrial revolution, which will run over fiber, 5G and, eventually, 6G.
The financial results have improved, too. Third-quarter organic sales rose 7%, with the Networks business Narvinger previously ran growing 9% and delivering an adjusted gross margin of 50.3%, above guidance. Investors greeted his first results as CEO by sending the shares up more than 10% in early trading.
Ericsson also placed a (very) early bet on network APIs: a way for the AI applications of the future to ask the network for what they need. Building the network is one thing. Having a hand in how machines use it could prove rather lucrative. Eventually.
On the other side of the ledger, there is plenty to worry about. Ericsson has little exposure to optical networking, leaving it outside the AI infrastructure bounty currently being enjoyed by companies such as Ciena, Cisco, Huawei and Nokia. Then there is geopolitical fragmentation, uncertain investment returns and the familiar difficulty of turning promising technology into services somebody will pay for.
Narvinger is clearly comfortable discussing all of it.
“It’s a very fun period of time,” he told me, “particularly for a company with substantial technology assets.”
“It’s chaotic though,” I insisted. “Do you like a bit of chaos, Per?”
“Whether you like it or not, I think that’s the reality, and I don’t see it changing,” he replied. “So I think it’s better to embrace it and try to figure out how to make the most of it.”
That combination of technological ambition and practical restraint ran through six significant findings from our conversation.
Best effort has an expiration date
Telecom has spent much of the mobile broadband era improving its ability to deliver video. Narvinger believes the services coming next will require a different proposition.
“We have been in what we call a best effort paradigm,” he said. Networks, however, are becoming “much more critical in society. We must be able to guarantee quality for all these new use cases.”
Some will come from industry and enterprise. Others will come from consumers. “We don’t think consumers will accept the best effort paradigm forever.”
That is a substantial change in the sales pitch. The question becomes whether the network can deliver a required outcome when the application needs it, rather than just how much traffic it can theoretically carry or how quickly.
Narvinger illustrated the point with the example of personal AI agents. People will become accustomed to having an agent available wherever they go. Losing access will feel very different from waiting for a video to load.
“If you don’t have your agent with you, you will probably feel a little bit lonely in the future,” he said. And nobody wants that.
The enterprise/carrier boundary is fading
My view of the current comms market is that the traditional border between enterprise networks and carrier networks is disappearing. I asked Per if he agreed, and whether that helped explain his decision to dismantle Enterprise Wireless Solutions as a separate business area.
“We absolutely believe that seamless connectivity will become even more important,” he replied. “We wanted to make sure that we put our entire R&D machine behind this.”
His October restructuring puts that thinking to work inside Ericsson. Private 5G and neutral-host solutions are moving into Networks, while Wireless WAN will remain a standalone business reporting to Americas chief Yossi Cohen. The boundary between carrier and enterprise networks is disappearing from the market; Narvinger is removing it from the organizational chart.
The identity of the customer deploying the technology, he explained, is less important than the experience it delivers.
“But making sure that the experience is seamless will be key in the future,” he said.
That might sound unremarkable, but coming from a company that has spent a century organizing itself around carriers’ needs, it certainly got my attention. An enterprise is no longer a special case to be handled somewhere else in the building. Its network requirements belong at the heart of Ericsson’s business.
The commercial logic follows: Ericsson can put its engineering resources behind a problem without first deciding which customer category owns it. The same technology can serve a carrier, a factory, a power plant, a transportation system — or all of the above. Read that way, the restructuring looks rather more ambitious than a retreat from enterprise: it changes who Ericsson’s core business is for.
Distribution: the carriers’ AI advantage
In the U.S., the fashionable picture of AI infrastructure is of enormous data centers — the bigger the better. Narvinger sees something equally essential in the infrastructure already spread across the landscape: mobile networks, with sites close to the people and machines that need computing.
“Mobile networks are probably one of the most distributed architectures out there in any industry,” he said. “Sometimes you will need compute closer to where you are.”
Other workloads can run in the data center. Ericsson’s strategy accommodates both; its opportunity lies in connecting them.
Carriers already have sites and aggregation points throughout their networks. As inference spreads outward, far beyond the data center, those locations can process AI workloads while also carrying traffic.
“[Carriers] have a lot of relevant assets,” Narvinger said.
Ericsson’s future now extends beyond the traditional perimeter of the carrier network and onto the vast hinterland of heavy industry. When AI controls machinery, where the computing happens and how quickly it responds become production decisions. The network is part of the operating machinery.
Ericsson’s private 5G and network API portfolio gives it a way into that market: connecting the machines, supporting the applications that control them and giving those applications access to the network capabilities they need.
Connectivity itself can also do more. “We basically can transform our mobile networks to also operate as a sensing network,” Narvinger said.
These are the words of a man who spends considerable time in Ericsson’s R&D division. Sensing has been a largely unappreciated part of the telecom story, usually bundled into the miscellaneous drawer marked IoT. But it is essential to AI’s expansion into the physical world — giving it electronic eyes and ears and, thus, agency.
Sensors feed AI information about its surroundings: movement, temperature, position, the condition of a machine. Network sensing extends that idea by turning communications infrastructure itself into an instrument of observation. The network can detect what is happening around it as well as carry information about it.
The possibilities are exciting and slightly terrifying. Think less Internet of Things, more ubiquitous Internet of Trillions.
Narvinger offered no revenue forecast for sensing. But he was clear that the infrastructure Ericsson has spent decades building is acquiring a wider purpose.
“We see a lot of assets that we have built that today are used for mobile networks,” he said. “They will have a wider use in a digital stack.”
For operators still wondering where they fit in the AI economy, that is a useful starting point.
The industry has not finished deploying real 5G
Narvinger acknowledged that many customers have yet to see a large volume of new 5G use cases. His explanation included an uncomfortable distinction between deploying the label and deploying the capabilities.
“It took us some time to get to what we would call true 5G or 5G standalone,” he said.
Non-standalone deployments arrived quickly. He described them as “basically a little bit more advanced 4G version.” (I prefer to call them “faux 5G.”)
“It’s not until you get to 5G standalone that you can really support some of these new use cases.”
That awareness is essential when operators judge whether 5G delivered on its promises. Services dependent on capabilities that were never fully deployed cannot provide a particularly useful benchmark of those capabilities, a point that I have argued for several years.
Narvinger said new use cases were starting to appear in some standalone networks, while acknowledging that the industry had not fully exploited the technology.
“Very rarely do you take a major technology step and then no one figures out what to do with it,” he said. “For me, it’s just a matter of time.”
That is confidence rather than evidence of an imminent commercial breakthrough. The key distinction is that he places further work on 5G ahead of assuming the next generation, 6G, will fix everything.
Network APIs need a pragmatic timetable
The pragmatism became particularly apparent when we discussed enterprise profitability and network APIs.
“It’s fair to say that that has taken longer than we expected,” Narvinger said of network API business development.
That admission should be stored alongside the larger architectural argument. A network capable of delivering differentiated performance needs ways for applications to request and use those capabilities. Establishing the technology does not establish the business on the same schedule. It also does not invalidate the business model.
Ericsson still sees exposing network capabilities as a way to create value. It’s right about this, but timing is everything. Meanwhile, the enterprise business needs to improve its financial performance. Narvinger spoke about reaching break-even sooner, without offering a date.
“We have a bad habit in our industry to start talking quite early about what will come,” he said.
An understatement for the ages.
Per offered a similarly useful correction when I suggested Ericsson’s longer perspective distinguished it from companies driven by quarterly results.
“I do have to say, though, that we also have quarterly earnings calls,” he replied.
Its hardware development imposes another timetable. Ericsson already knows quite a lot about products it expects to deliver four or five years from now. Patience has an engineering basis, even if investors would prefer a faster result.
Sovereignty needs trust
The largest question reaches beyond Ericsson’s own equipment. A company can trust its network supplier while remaining dependent on other companies for cloud services, AI models and the applications directing its operations.
That is the problem I describe as operational sovereignty. Who ultimately controls the infrastructure, and what happens when the interests of its suppliers diverge from those of its owner?
Narvinger’s answer began with the limits of technological independence. Only so many complete technology stacks can exist in the world.
“I truly believe you need allies,” he said. “Different allies will provide different parts of that technology stack.”
A sound financial strategy spreads risk. A sound sovereignty strategy does the same. Counterintuitive, perhaps: retaining control depends on choosing which mix of suppliers to depend on.
That calculation runs through the entire stack. “If you trust the entire stack, you also need to trust every part of it and every supplier in that.”
Who supplies the model? Who operates the data center? Who controls the complete service? Trust in the company connecting it all cannot answer for everyone else. Ericsson’s Nordic heritage gives it an advantage here, as it does its neighbor, Nokia. But even a trusted supplier cannot vouch for the whole chain.
“I don’t think we have figured out exactly how that will work in practice,” he said.
A refreshingly candid admission from someone whose company intends to supply a significant part of the answer.
“In general, trust will become more important. That’s my absolute conviction.”
He connected that conviction to AI and the difficulty of verifying its results. Trust acquires greater significance when an answer influences how someone, or something, acts.
Ericsson’s challenge is to make that confidence meaningful across the wider infrastructure its customers depend on. Its opportunity is to become a supplier they want beside them while working through those dependencies.
I suggested he was building a business around communications infrastructure interwoven with trust infrastructure.
“That’s a good summary,” he said.
Well, I guess I got one thing right.
A new home. Still Sweden.
Ericsson’s future campus in Hagastaden brings me back to where we began. It is a spectacular location beside the water, where visitors will be able to arrive by boat like, er, Vikings.
The journey from its existing HQ in Kista is short. The statement Ericsson is making travels much further.
After speculation that the company might move its headquarters to the U.S. and draw closer to the cultural center of Big Tech, Ericsson is putting down fresh roots in Stockholm. The appointment of Narvinger, a Swedish-born (björn?) Ericsson lifer with 30 years inside the company, gives that decision a human counterpart.
Together, Ericsson’s new campus, the strategy and the man suggest a commitment to an industrial tradition in which engineering carries obligations. Build things that work. Make them last. Accept responsibility for the people and institutions that depend on them.
At the other end of that spectrum sit the U.S. hyperscalers, whose pursuit of scale, market control and financial returns increasingly determines the direction of technological development. Making money is essential. Treating every other consideration as subordinate to it is a choice.
As AI moves into factories, utilities, transport and the essential machinery of society, that choice becomes geopolitical. We need companies capable of building this infrastructure whose ambitions include the prosperity and resilience of the societies using it. Ericsson’s Swedish roots, Narvinger’s engineering instincts and the strategy he is setting out offer an encouraging counterweight to the hyperscaler worldview.
The next great digital economy needs more than competing technologies. It needs competing ideas about what technology is for — a digital dialectic. The companies building the next digital economy will help decide whom it serves and how. It is a good thing for the world that they won’t all be American hyperscale companies.
Learn more about telecom’s role in the AI economy on Fierce Network
5G’s true potential: Why analysts are missing the mark on industrial digitalization
Ericsson, Aduna and the gradual rise of 5G APIs
Can you turn it off? The new test of tech sovereignty
Introducing: The Unified Infrastructure Stack
Stephen M. Saunders MBE is a communications analyst and USPTO-registered inventor examining how digital infrastructure — 5G, cloud and AI — is reshaping industry, power and society, as well as underpinning the emerging, ubiquitous global digital economy. As anchor of FNTV and a longtime industry insider, he focuses less on growth narratives and more on execution, risk and how hyperscale technology is distorting markets, governance and society at scale.
Opinion pieces from industry experts, analysts or our editorial staff do not represent the opinions of Fierce Network.