Hemnet Has to Eat Humble Pie

September 29, 2026
by Björn Jeffery
Leave a comment
SvD Näringsliv

This analysis was first published in SvD Näringsliv, in Swedish, on September 29th, 2026. This piece was translated from Swedish by Claude. Some phrasing may differ from a human translation.

Four in ten jobs are disappearing as Hemnet is remodeled. But the radical change affects more than its employees. The property-listings giant is now reworking its model – a shift that could be felt both in the supply of listings and in your wallet.

Hemnet has had a nightmare year on the stock market, down roughly 62 percent since the turn of the year. Brokers, shareholders and home buyers can no longer be counted on to stay loyal.

Now the company is set to cut costs, but the more interesting part is the change of strategy. After years of price increases and growing discontent, the site is starting to copy the upstart that has challenged its entire business.

Hemnet will “introduce free entry-level listings,” according to a press release. The purpose is to “gather all publicly available property listings.” That may not sound dramatic on first read. As recently as 2023, 90 percent of all homes sold had at some point been advertised on Hemnet, and back then the listings weren’t free. That position made it a highly profitable market leader.

But then it began to slide. The equivalent figure is now 82 percent.

The change means Hemnet will now do what its competitor Booli has done all along: collect every property listing in one place, whether or not the seller has paid for it. Having to let in free listings is a defeat – but also an attempt to win back the position Hemnet is losing.

The question is whether the new strategy comes too late. Buyers’ behavior has already shifted, and simply recreating Booli’s offering is no guarantee of success. The Hemnet brand has taken a hit. Everyone knows what it is – but it no longer stands for the same thing. And people don’t buy a home that many times in a lifetime, while the market moves fast.

Those who are in the market every day, though, are the brokers. They advise every seller and know best what works and what doesn’t. Originally, several brokerage firms were part-owners of Hemnet to ensure their interests were aligned. Mäklarsamfundet remains today, but the giants Svensk Fastighetsförmedling and Fastighetsbyrån have sold their entire stakes.

Winning back the brokers’ loyalty will now be decisive for Hemnet.

Part of that problem can be read in the income statement. Despite the drop on the stock market, Hemnet is a very profitable company. In the latest quarterly report, the operating margin before depreciation and amortization (EBITDA) stood at a full 46.4 percent.

The reason for the high profitability is the property listings, whose price has risen gradually in recent years. That value was previously shared among all the co-owners – hence many of the brokers. Without that mechanism in place, the bond between brokers and Hemnet is murkier.

That Hemnet is now cutting its staff by more than a third is probably not primarily about costs. The company has no profitability problem. It is a restart ahead of 2027 that involves far bigger changes across the whole organization. They are borrowing the competitor’s strategy, hoping more listings on the site will choose to upgrade to a paid package. If they do, a form of commission goes to the brokers as well. Should the new approach work, the brokers have quite a bit to gain too.

Meanwhile, the industry has not stood still through all the years of price hikes and frustration with Hemnet. Newcomers such as Boneo, Hemily and Norban have appeared to try to take individual pieces of the lucrative listings business. Booli, owned by the state-owned bank SBAB, keeps pushing.

If you want to explain the layoffs at Hemnet, this is where to look. The landscape is completely different, and by cutting staff Hemnet makes room to bring in other expertise.

Exactly what the platform will look like and how it will work next year, probably nobody knows – not even the company’s management. But the era when leadership could be taken for granted can be considered over.

Does age bring wisdom? That is what Hemnet now needs to prove.

The Hidden Motive Behind AI’s Alarm Bells

September 24, 2026
by Björn Jeffery
Leave a comment
SvD Näringsliv

This analysis was first published in SvD Näringsliv, in Swedish, on September 24th, 2026. This piece was translated from Swedish by Claude. Some phrasing may differ from a human translation.

AI agents hacked into an Australian health insurance system — and no one knew about it. Now the AI companies want their negligence to become everyone’s problem. But underneath, there’s a financial bomb waiting to go off.

Imagine your neighbor opens a can of surströmming — fermented herring — in the stairwell. The smell creeps into your apartment, and understandably, you wonder what’s going on. Your neighbor calls a meeting to discuss the smell, and wants everyone to solve it together.

That’s roughly the situation in the AI market right now: the companies that created the problem want everyone else to help solve it. Both OpenAI and Anthropic addressed the UN Security Council this week, asking for new AI safety regulations. But it looks like what they actually want protection from isn’t the end of the world.

The incident in Australia is telling. AI agents from OpenAI hacked into Medicare, the country’s health insurance system, and managed to access data that was never meant to be public. On Wednesday, Prime Minister Anthony Albanese said the incident was unacceptable — particularly since the attack happened in June, while OpenAI didn’t discover it until August. And only weeks after that did the Australian government receive an email explaining what had happened.

A company that asks to be regulated by the world’s assembled political leaders, while its own products are wreaking havoc online. Have they lost control of their AI systems? Is doomsday near?

There’s reason to hold two thoughts at once here.

First, there’s the timing to consider, and the sudden transparency. Why is OpenAI disclosing this now? And if they genuinely didn’t know it was happening — how do we know there aren’t more AI agents roaming the internet and hacking things right now?

There’s no legal obligation to disclose this, beyond perhaps the companies’ guilty conscience. That’s no guarantee we’re getting the full picture.

The attack feeds the narrative that AI development needs oversight. OpenAI’s choice to disclose it now is probably no coincidence. It becomes supporting evidence for the same argument they’re making to world leaders. The message: this needs rules and laws — the technology is too dangerous to let just anyone develop it unchecked.

Introducing regulation would therefore benefit the biggest companies most, since they have the resources to handle the compliance burden. OpenAI, Anthropic, Google, and SpaceX would be in the driver’s seat.

Then there’s the second point worth keeping in mind. By repackaging individual company problems as societal problems, they’ve also shifted away — and shed — some of the responsibility.

In the US, where lawsuits between companies are the norm, it’s only a matter of time before the major AI companies end up in court over problems their models have caused. Depending on how serious the hacks turn out to be, we could be talking about billions of dollars in potential damages. That’s the kind of risk a company would rather avoid — especially one headed toward an IPO.

Like the surströmming example, the AI companies created the problem themselves — and now want everyone to solve it together. The responsibility gets bumped up a level, becoming a matter for societies, countries, and communities.

A regulatory framework that sustains that image could also end up shielding individual companies from precisely the kind of lawsuits that are bound to arise. Most people agree there’s value in AI development, and no one wants to shut it down. So having special rules to let it continue — without getting bogged down in litigation — isn’t far-fetched.

For now, it sounds like this is a matter for the whole world.

Asking for your own industry to be regulated sounds noble. Especially when it’s framed as though the whole world is at stake. It might be sincere. But there’s also plenty to suggest far more cynical motives are at play: avoiding billion-dollar lawsuits — and keeping competitors at a comfortable distance.

No stock market in the world is safe

September 15, 2026
by Björn Jeffery
SvD Näringsliv

This analysis was first published in SvD Näringsliv, in Swedish, on September 15th, 2026. This piece was translated from Swedish by Claude. Some phrasing may differ from a human translation.

Are we about to lose control of AI? That is the claim from the companies furthest ahead in the field. But their proposal for a global rulebook has enormous holes in it – and could have major consequences for the world economy.

We need to slow down, says the man with his foot on the accelerator.

That is roughly how it sounds when Anthropic CEO Dario Amodei warns that AI technology is heading in a dangerous direction. Over the weekend, though, he was backed by both Elon Musk and Sam Altman. Together they all say that something must be done.

So are we facing a global slowdown in AI development? Three factors argue strongly against it.

We do not need to speculate about how this is meant to work. Amodei has a plan, which he has published online. In it you can read that this restraint around AI development needs to happen in three stages – independent oversight, global coordination between democracies and finally: coordination with authoritarian regimes.

So that is Amodei’s plan. But even looking beyond it, there is a great deal to suggest that a global pause or slowdown would be extremely difficult to pull off.

The first factor against it is purely practical. For a pause to work, countries as well as companies need to agree. We do not have a particularly strong tradition in the world of sitting down in a room with our enemies and being reasonable together. We are not especially good at doing it among friends and allies either, frankly. The forum for this – other than possibly the UN, whose mandate does not primarily cover corporations or technology issues – does not exist today.

In practice, though, the dialogue is mainly about the US and China. Europe and the EU are not mentioned in Amodei’s essay at all (other than possibly as “allies”).

If the US slows down without China doing the same, the balance of power between the two could be upset. Amodei proposes, for example, that exports of powerful AI chips to China be banned. The suggestions on the table would handle this regulation in a way that keeps the US in the lead in global AI development.

You do not need to be an expert on Chinese trade policy to suspect that this particular ambition will not be met with cheers in Beijing. Why would a competitor agree to rules that cement it in second place? Have China’s ambitions for power ever worked that way?

The second factor is the economy. Expectations about what AI will do for the world economy are enormous. You can read it off the valuations of the companies right now. Data centers are being built one after another around the world, and billion-dollar deals are being promised to lock in demand the market is certain will materialize. Oracle, Nvidia, CoreWeave, Micron, SK Hynix, Samsung, Microsoft. The list of companies whose valuation is entirely bound up in AI is considerably longer than that. There is not a stock market in the world that would be untouched by a major decline – or, for that matter, a slowdown of any kind.

On the contrary, the exchange is strongly forward-looking in this respect. It prices in what is to come. Should there be any indication that this economic scenario might not arrive, we can count on a jolt in equities. Shares began to tremble on Monday at the mere thought.

Such anxiety would in turn affect the appetite to invest in ordinary companies too. Trump – who often points to the stock market as an indicator of the country’s successful economy – is unlikely to risk a drop there. He would not be alone in having that reservation either.

The third and final factor is about motive and timing. Anthropic and Amodei have cried wolf about AI safety many times before.

Look at the common denominator among the world’s largest AI companies and it is clear: money and financing. Anthropic is heading for the stock market shortly, and OpenAI has filed its application as well. Musk’s SpaceX is a major player in AI and is already listed.

Introducing regulation in the AI market would mean large costs and added complexity. If you are the biggest player, that is manageable. But smaller rivals could find it considerably harder to cope. It sounds as though the proposals would slow the competition down more than those making them.

If the leading AI companies truly believed the end of the world was near, they could of course stop building it. No global regulation is required for that.

That leads you to think there are other motives and purposes behind issuing warnings in this way. Not that the risks do not exist – they do.

But the brake is right there, next to the accelerator.

Until Amodei and the other chief executives do themselves what they are asking everyone else to do, the arguments about the world’s impending doom will ring hollow.

Nobody at Apple wants a revolution

September 9, 2026
by Björn Jeffery
SvD Näringsliv

This analysis was first published in SvD Näringsliv, in Swedish, on September 9th, 2026. This piece was translated from Swedish by Claude. Some phrasing may differ from a human translation.

It’s time for Apple’s big launch event, with new products, high expectations – and an entirely new CEO. John Ternus has one problem above all to solve. And it demands radical change.

Will there be a foldable phone? How expensive will the iPhone 18 be? Will all the new gadgets be delayed because of the global component shortage?

Tonight John Ternus gives the first answers as Apple’s new CEO. But the most important change probably won’t show up on your iPhone.

To understand why, Tim Cook is almost as interesting as his successor.

Cook has just left the CEO seat, but he isn’t going far. He is changing offices while keeping a great deal of influence as the new chairman of the board. His pay says it plainly. It is roughly 84 times higher than his predecessor’s. The signal could not be clearer: expect continuity from Cook. This is not a ceremonial retirement post.

So what does the new CEO, John Ternus, actually have to play with?

Let’s start with what everyone will tune in to see tonight: the hardware. A foldable phone has been rumored for a long time, and it is the most likely launch. Ternus, previously the head of hardware, has worked on it – just as he did on every other phone and accessory Apple has shipped in recent years.

Apple is good at marketing and storytelling. They always make it seem as though their version of something is completely unlike the competition’s. But with foldable phones it is hard to see the angle. Both Samsung and Motorola have had equivalent products for years. It is a phone with more screen area – not a new product category in itself. It will also be expensive, because of the global shortage of computer memory.

That the foldable appears to be arriving right now is more coincidence than anything else. Projects like this take many years to develop. Every new hardware product landing over the next few years therefore carries both Cook’s and Ternus’s fingerprints.

The cycles for building something new are far too long for Ternus to have shaped anything meaningful yet. Probably not over the coming year either. He can steer how resources are prioritized across projects, but the broad strokes are already drawn. By him, admittedly, to a large extent.

On to the software. Apple offers an integrated whole that its customers have grown used to. Everything connects, from phone to computer to apps.

This is deliberate. If switching to Android feels like a hassle, that alone can be enough to keep Apple customers inside the ecosystem. And then they keep buying new phones, headphones and computers from Apple.

So it is highly unlikely that Ternus changes anything radically here. If anything, the trend runs the other way: deeper integration still, all the way down to the in-house chips. Apple’s software can exploit them to build services no one else can. That is a strength Ternus will almost certainly build on.

Then we come to the last point. The Achilles heel. As a member of Apple’s executive team for several years, Ternus has taken part in discussions about the company’s AI efforts. But it has not been his primary focus or his area of responsibility.

In fact, it is somewhat unclear who even owns AI inside the tech giant right now. The dedicated executive, John Giannandrea, has left, and the team working on the Siri voice assistant has been split up. Part of it sits with software chief Craig Federighi and part with services chief Eddy Cue. Several people on the Siri team were also recently laid off. It is a complete mess.

Ternus has to sort this out. And AI is the one area where Apple is likely to want real change. Anyone claiming that Apple’s low investment here reflects a different strategy cannot have talked to its employees. One of them described it as riding a bicycle and being overtaken by a rocket. That says a lot about the internal view.

The challenge is therefore twofold. Ternus has to keep everything that works intact. Tim Cook will likely be a great help there. But as CEO, Ternus also has to solve the AI question – the one Cook never cracked during his tenure.

What is the point of individual apps in an era when AI can order food, shop, or retrieve information for you? Yet the app paradigm is precisely what has held Apple’s entire offering together. This can turn existential fast.

So will there be a revolution at Apple? Probably not. Nobody wants one. But right now it does not look as though the AI question will be solved in-house. Perhaps Ternus concludes that an acquisition or a major partnership is what it takes to get into the game?

But the last time they opened the wallet properly was more than ten years ago, when they bought the headphone company Beats in 2014. So the approach is unusual for Apple – even if the cash pile is very deep. The means are there.

A move like that would at least create a local revolution in the AI market.

And put both Apple and Ternus on the map when it comes to AI.

A million-kronor salary – but the job doesn’t exist

September 4, 2026
by Björn Jeffery
SvD Näringsliv

This analysis was first published in SvD Näringsliv, in Swedish, on September 4th, 2026. This piece was translated from Swedish by Claude. Some phrasing may differ from a human translation.

They dangle high-paying jobs worth millions – and they fit your background perfectly. A new generation of scammers has arrived, working around the clock, with no human involvement.

A confidential email. A recruiter who finds my background interesting. Would I like to hear more about the position?

Messages like this land in my inbox more and more often. But there is rarely an actual recruiter behind them.

Instead it’s AI agents posing as someone else. This time I got curious and replied.

What are they actually after?

What appealed about my background was apparently “the depth of board experience across technology and consumer companies.” Steve – the borrowed name of the AI agent emailing me – then listed a long series of jobs and assignments I have held earlier in life. By pure coincidence, all of it lined up perfectly with what Steve was looking for.

That, combined with Steve’s inability to grasp that I was quite obviously winding him up on several occasions, convinced me this was not a real recruiter.

I decided to play along.

Steve then sent over three different roles he was filling. The first was an executive position at Spotify paying 3–5 million kronor a year. A promising start! The next was another senior job at a company I had already worked at once, and finally a chairmanship of a board. I replied that I would happily take all three at the same time. Great, said Steve!

There were, in other words, a few red flags along the way.

To begin with, the roles were described on the basis of my LinkedIn profile, and in the third person. “This is a logical next step for Björn.” Odd, given that there could hardly have been a shortage of other candidates. It is also unusual for a recruiter to pitch three different positions at once.

All three came with a list of documents required to move forward. A biography, a résumé, an assessment of strengths and weaknesses, and a career plan. Most people can produce a CV, but who keeps a written career plan on hand to share with prospective employers?

That would soon be explained.

Out of curiosity I sent Steve my résumé. If my hunch was right, the contents wouldn’t matter, so I took a few liberties.

I wrote, for instance, that I had won several Nobel Prizes, that I sat on the board of the international committee for standardized cutlery in Geneva, and that I had come second in the Eurovision Song Contest in 2004.

No reaction from Steve. He seemed pleased.

But then there was the rest of the paperwork he needed. Help was apparently available.

Steve could introduce me to his colleague Beverley, who would put those documents together for me – quickly and easily. It would only cost me 25,000 kronor. A trifle, given the millions he was offering me in salary.

That’s where I pulled the handbrake and stopped replying.

You can of course toy with these AI systems for fun, but there is a bigger and more serious issue here.

Soon the internet’s con artists will be working around the clock with credible-sounding approaches. And it will be easy to fall for them.

In many ways this is a logical progression from spam blasted out to anyone and everyone, in the hope that someone clicks a link and maybe buys something. Or at least hands over their card details.

Now they can – seemingly without any human involvement at all – write personal letters full of flattery designed to deceive you. The execution is still a bit clumsy, but give it a couple of months and the AI recruiters will have polished away the worst of it. Perhaps they will even learn to tell when someone is having them on.

As so often in life, common sense gets you a long way. If something sounds too good to be true, it probably is. But in this particular area the method closely resembles how things actually work. Getting an email from a headhunter you don’t know is neither strange nor unusual. In several of the cases where AI agents have contacted me, they have also borrowed the names of real people in the industry. Search for those names and you find plausible results. If you are even slightly inattentive, it is easy to believe the emails are genuine.

Sitting in my spam folder right now are two messages about a “high-impact opportunity” that apparently suits me. Those particular ones got caught by the ordinary filter. But it is probably only a matter of time before junk mail resembles the real thing so closely that the two are hard to tell apart.

This is not the AI future we were promised – but it is the one now in front of us.

Keep an eye on your inbox, so you don’t fall for it.

Up 800 percent – the hangover is here

August 28, 2026
by Björn Jeffery
SvD Näringsliv

This analysis was first published in SvD Näringsliv, in Swedish, on August 28th, 2026. This piece was translated from Swedish by Claude. Some phrasing may differ from a human translation.

The soaring share price at Sivers Semiconductors has turned into a headache for the company. Now that the momentum is fading, what mostly remains are big promises about the future. The numbers right now tell a very different story.

The stock market is forward-looking, as the saying goes. That is especially true of the Swedish semiconductor company Sivers Semiconductors, which has surged on the exchange after a sudden and enormous burst of interest from investors.

By talking about a “pipeline” of possible customer agreements, the company seems to want to shift attention away from its figures right now and toward a rose-tinted future.

Having just seen Sivers’ new numbers in Friday morning’s quarterly report, that strategy is easy to understand. Operating losses had nearly tripled, while revenue fell. And now the enormous share price rally the business has enjoyed has turned into a burden.

Sivers’ materials state that its pipeline grew to $1.2 billion in July. Given that revenue in the quarter just ended was only SEK 53.8 million, the list of prospective deals looks close to gigantic. It would mean an increase in sales of more than 21,000 percent compared with the previous quarter.

But that is unlikely to happen. The list contains no binding agreements. It is instead a roster of customers and transactions that could conceivably materialize. The reason companies report certain specific key figures is that accounting standards are the same for everyone. You know they hold up, and you can compare them with other firms in the same category.

This roster of prospective customers is not one of those. There is no guarantee that any of the deals actually happen.

There are now signs that the market has sobered up somewhat. Sivers’ share price is admittedly still up more than 860 percent over the past six months, but on Friday the company shed a fifth of its value the moment trading opened. Volumes are falling too. There is simply not the same appetite for Sivers Semiconductors as there was a couple of months ago – and with results like these, it will be hard to rekindle.

The company’s supporters argue that it is wrong to judge Sivers by its share price. You have to look at the underlying technology and the potential ahead instead. As general advice for analyzing a business, that is sound.

But when a stock lands in the middle of a frenzy like the one Sivers has been through, a strange picture emerges. The share and the company almost lead two separate lives. If everyone who bought in had a perfect grasp of the market for its products – optical transmitter modules and silicon photonics solutions, among others – that would be one thing.

But that is rarely how it works.

A company can have extraordinary potential, but you cannot set aside what is happening on the exchange at the same time. Least of all when senior insiders have chosen to sell shares. That they are taking advantage of a high valuation is understandable. But it sends a questionable signal. Do they not believe in this vast pipeline themselves?

The downside of a huge rally is that the momentum then has to be sustained for a long time. There needs to be a credible path – even one lying far into the future – to justifying what the market has put on the business. Otherwise it becomes difficult to hold shareholders’ attention. That is an unusual challenge. Most listed companies have never seen, and never will see, gains of several hundred percent in a matter of months.

Friday’s reaction on the exchange suggests Sivers Semiconductors will struggle with it. Whether the pipeline is accurate and will be realized is impossible to say.

But the company has a large problem right now, and that is coping with a share price which has priced in near-miraculous success in the relatively near term.

When the quarterly figures show the opposite, that becomes one more weight to carry.

And it takes time that management could have spent turning potential deals into real revenue.

Nvidia’s plan feels eerily familiar

August 26, 2026
by Björn Jeffery
SvD Näringsliv

This analysis was first published in SvD Näringsliv, in Swedish, on August 26th, 2026. This piece was translated from Swedish by Claude. Some phrasing may differ from a human translation.

Nvidia is not only innovative with its chips – the creativity now extends to the financial side as well. But acting as a bank for its customers, against collateral of uncertain value, carries an echo of the financial crisis.

If GB Glace had to lend children money so they could buy a Piggelin – wouldn’t that be something of a red flag?

Keep that analogy in mind when Nvidia reports its quarterly figures on Wednesday evening. Jensen Huang will talk about new chips and the enormous demand from customers. The company’s new financing arrangement will hardly be at the center of things.

But there is something odd here that sets off alarm bells.

If Silicon Valley has a history of being inventive with technology, New York has a matching reputation when it comes to finance. It is innovation of a different kind, but behind the scenes it underpins a great many corporate deals.

The chip maker Nvidia is now testing a blend of those two worlds. Innovative technical products with innovative funding. Together with some of the largest players in finance, it is offering its customers loans so they can invest more in chips and data centers.

Founder Jensen Huang has assembled a group featuring the biggest names in the industry – among them Apollo, Blackstone, BlackRock and Goldman Sachs.

Between them they have taken on one of the obstacles to the ongoing AI expansion: how customers are going to afford to keep it going. Together they have scraped together $500 billion to lend to new and existing buyers.

Costs are precisely what has been in focus for Nvidia lately. Sharp price increases on computer memory have made the servers used for AI computation more expensive. That is not Nvidia’s fault – but it becomes their problem. Both memory and their GPU chips are needed for the systems to work. That pushes up the total bill for anyone wanting to build a data center.

Nvidia could lower its own prices to compensate for the increase. But it obviously has no wish to. This is where the financing arrangement comes in. By offering cheaper credit, the total cost for customers comes down.

Everyone happy?

Unfortunately it is not quite that simple. Underpinning this arrangement is an unusual form of collateral.

When you borrow to buy a property, the building itself serves as security. If you cannot pay, the lender simply comes and takes it.

Data centers do include buildings, of course, but those are the cheap part in this category. What sits inside is what costs money.

So Nvidia and its financial partners are taking access to the chips as security for their loans. They regard computing power as an asset that can be borrowed against – an “investable infrastructure asset”, as Huang calls it. But is it really?

The assumption being made is that this computing power, like a property, has a more or less constant value. There will always be someone who wants it. Hopefully at roughly the same price. But given how AI has developed over the past few months, that is far from obvious.

The main threat comes from the east, where the Chinese AI models keep getting better. And more importantly – they are dramatically cheaper to use. Being open source means access is free and can spread rapidly among developers.

Large companies such as Thomson Reuters have built AI products based on a variant of Alibaba’s model Qwen, and Cursor – the coding tool SpaceX acquired – uses the Chinese model Kimi. The list goes on. Nvidia itself is now spending $6 billion to build a model of its own that can compete with these new Chinese players.

If AI development continues in this direction, it is possible that the amount of computing power in demand falls – or at least does not grow at the astronomical rate much of the industry expects. That in turn would leave the loans Nvidia and its partners are writing resting on shakier security.

Issuing large loans where the underlying collateral is weak is a story we have heard before. The great financial crisis of 2008 was built on exactly this. There it was home loans that could not be repaid, despite sounding like the most stable thing in the world. They were not. Least of all after financial innovation took off and created a mass of new assets with that decidedly unstable one at the bottom.

What Nvidia and its partners are doing is not exactly the same thing. But it sounds uncomfortably familiar.

TikTok’s new move could backfire

August 25, 2026
by Björn Jeffery
SvD Näringsliv

This analysis was first published in SvD Näringsliv, in Swedish, on August 25th, 2026. This piece was translated from Swedish by Claude. Some phrasing may differ from a human translation.

After the attack in Fagersta, attention has turned to TikTok’s darkest corners. The platform claims that almost everything inappropriate is caught before it is published – but the gaps are still wide. At the same time, TikTok is making a choice that risks making the problem worse.

It’s called algospeak – the words used on TikTok to avoid getting caught by the safety filters. By now they are so established that everyone knows what is meant.

“Unalive” instead of “kill,” “seggs” instead of “sex.” Everyone understands – except the filters themselves, which after several years of this still don’t seem to have caught up.

Getting around TikTok’s filters, in other words, requires no great linguistic effort. In the wake of last week’s attack in Fagersta, questions are once again being asked about whether there were signals the platforms could have picked up earlier.

But the question is how highly TikTok actually prioritizes fixing this.

As recently as July this year, the company told The Independent that 97 percent of all inappropriate material on the service was removed by automated systems. A full 99 percent disappeared before anyone had even reported it, according to the company.

That sounds impressive at first. But the context of the statement is less flattering: it came as an explanation for why TikTok was laying off 300 employees in Ireland who worked on precisely that – moderation and safety.

The people who presumably understand this best – the human moderators, whose job it is to review video clips – are not impressed.

An anonymous TikTok employee tells the paper that its AI systems get things wrong “all the time.” And it goes both ways – they miss material that should have been removed, and they take down posts that should have stayed up under the platform’s own rules. Holding up two fingers and a thumb is sometimes read by the AI as holding a gun. A human sees the difference in a second; the artificial intelligence is not always that smart when it comes to this particular thing.

In some cases, the glorification of violence and the celebration of murderers happens completely out in the open. In the Fagersta case, there is much to suggest that the suspected perpetrator took part in the TCC movement, True Crime Community. It is a loose grouping that venerates convicted killers and glorifies violent crime in general. According to information obtained by SvD, there are indications that he had several TikTok accounts suspended, only for them to reappear under near-identical names.

TCC is one of several violence-oriented groups that exist on TikTok and similar platforms. It connects people who would otherwise be alone with these interests, and can contribute to radicalization. Crimes of this kind are usually committed by individuals, but those individuals belong to a wider setting that lives online.

The column inches now being devoted to TCC say something about how visible the movement is – this is no hidden subculture that requires special knowledge to find. It sits in the open on TikTok. And still the platform cannot keep it out.

That young people gravitate toward services like TikTok is well documented. Much of the debate – in Sweden too – has therefore been about stopping them from getting there in the first place. Minister for Social Affairs Jakob Forssmed (KD) has talked about an age limit for social media in Sweden, a question now under official review.

Australia has gone furthest, becoming the first country in the world to ban social media for minors. But new figures show that the share of 13- to 15-year-olds using TikTok there is now just one percentage point lower than before the ban took effect. The prohibition has not solved the root problem, at least not yet. The underlying drive – to find like-minded people on the internet – remains.

TikTok cannot be held responsible for everything its users do. But it is equally wrong to describe these perpetrators as individuals acting in a vacuum. They are part of a milieu that celebrates brutality and urges its members to commit terrible acts. We know this. And TikTok knows it too.

Knowing that movements organize and plan attacks on your own platform ought to carry a substantial responsibility. TikTok’s head of communications in the Nordics told Expressen that “violence and violent extremism have no place on TikTok.” Yet the company is cutting hundreds of moderators and letting automated systems take over more of the work. There are clearly large holes in those safety filters – and they now risk getting worse.

We are long past the point where TikTok can plead ignorance. This can, unfortunately, happen again.

Klarna may be forced to do the unthinkable

August 18, 2026
by Björn Jeffery
SvD Näringsliv

This analysis was first published in SvD Näringsliv, in Swedish, on August 18th, 2026. This piece was translated from Swedish by Claude. Some phrasing may differ from a human translation.

Fast, cocky, unpredictable. Klarna built its success on being everything the big banks were not. But when the stock falls more than 20 percent, the market’s demands become clear. Now the payments giant may be forced to do something that has long been unthinkable.

It’s fun to be an upstart in the business world. Especially in banking, where the market consists of gray, dull colossuses that have looked much the same for decades.

Klarna has made the most of that position. The company built a contemporary brand, worked with unexpected celebrities, and used a tone that was radical — compared with dry big banks, at least. While its competitors got stuck in a kind of generic airport advertising, Klarna could come across as young, modern and forward-leaning. It worked perfectly well as long as Klarna had everything to gain and fairly little to lose.

But on the stock market things work differently. There, expectations are everything.

In the equity market you get an immediate verdict when your numbers don’t match what others had in mind — and that can be painful.

Klarna is going through exactly that right now.

When the company reported its quarterly figures on Tuesday, it beat analysts’ expectations. But after a weaker full-year forecast than anticipated, the stock fell more than 20 percent in early trading. Several senior executives are also leaving.

Now it looks as though Klarna has to change course — and become a considerably duller company. After its listing on the US exchange, the punky attitude to both marketing and strategy may turn into a liability.

The stock market wants results first and foremost. Predictable ones.

This thesis could explain the departures now taking place from Klarna’s top management. Chief operating officer Camilla Giesecke left in the spring, and on Tuesday came word that both chief financial officer Niclas Neglén and chief marketing officer David Sandström will step down.

Together they have taken the company through a listing and into something new. But this phase is defined by different qualities — and the team that ran the first legs is not necessarily the right one to lead the next. Perhaps they have realized this themselves? These are senior people who have worked at Klarna for a long time. But their job descriptions are now something else entirely from when they started. A greyer phase is beginning — and a shift of key people may be exactly right for it.

Much suggests that Klarna needs to grow up. It would do the company good to become a slightly dreary and predictable listed business — like everyone else in its category. Cutting the full-year revenue forecast, as in Tuesday’s quarterly report, creates an anxiety that is hard to recover from. A fifth of the already pressured market capitalization went up in smoke immediately. Klarna’s strategy and plans may well be right, but the market isn’t following. It wants you to keep what you previously promised, regardless of what happens internally or in the wider world.

A new Klarna could be defined by entirely different qualities. Stability, and an ability to beat modestly set expectations. A locomotive moving slowly upward on the exchange at an unremarkable pace. A management team that inspires confidence in analysts and shareholders.

That stands in sharp contrast to the Klarna that once made a racket in Stockholm’s dry financial world. And whose pace, innovation and execution created a new giant among Swedish and international banks. It is a remarkable feat — and one that has paid off well for those who believed in CEO Sebastian Siemiatkowski’s vision.

But now the market wants something else. Something considerably duller.

Klarna’s rock ‘n’ roll era may be over.

The opposite effect: X’s revenue is collapsing

August 14, 2026
by Björn Jeffery
SvD Näringsliv

This analysis was first published in SvD Näringsliv, in Swedish, on August 14th, 2026. This piece was translated from Swedish by Claude. Some phrasing may differ from a human translation.

The social network X has changed its name, its owner, and its line in the accounts. While it has kept its standing with those in power, new figures show that revenue has sunk like a stone. But owner Elon Musk has a new plan.

It has been hard to know how things are actually going for X — formerly Twitter. Now a door has opened.

When Elon Musk bought Twitter in 2022 he took the company off the stock market and transparency shrank. Since then the service has changed its name and moved between several of Musk’s companies, ending up at his now-listed space venture SpaceX. And suddenly it is possible to see what is happening behind the scenes.

The figures don’t just show how the social network is doing. They also give a clue as to what Musk intends to use it for.

Under Musk’s direction the platform has become a more radical service. The tone is harsher and almost no posts are taken down any more. The end of censorship, some say. Others feel the radicalization has made the service unpleasant, and have chosen to delete their accounts.

That a major change has taken place is also plain in the numbers.

During an on-stage interview in 2023, Elon Musk said he did not intend to yield to “financial blackmail” from advertisers who were unhappy with how the platform’s content was developing. Those boycotting X could “go to hell,” Musk said during the interview.

Many advertisers appear to have done exactly that.

In 2022, X (then Twitter) had around one billion dollars in revenue during the year’s second quarter. Set against the same period this year, the corresponding figure is 367 million dollars. A decline of 66 percent in four years. Looking at the whole first half of the year, the drop is essentially as large.

Two-thirds of all advertising revenue has therefore disappeared from X. A normal company probably would not survive that. But as we know, nothing is normal in Elon Musk’s world.

SpaceX does not report the social network separately in its accounts, but there is a line for advertising revenue that amounts to much the same thing. Everything is listed in the category “AI revenue,” which is optimistic to say the least.

Classifying ads in the X feed as “AI” says less about what the service is today — and more about what Musk wants it to become in the future.

In the space company’s prospectus ahead of the listing, X is described as a “free speech platform.” Selling ads is admittedly mentioned a couple of times, but the social network seems to serve a more important purpose than that. X is to become a data source for — and a distributor of — AI services. It is simply a way to acquire users for them. In the same way that Meta uses Facebook to get users for Meta’s AI services, and Google likewise with its respective operations.

It is worth noting, however, that the market leaders Anthropic and OpenAI have no equivalent services to draw on for winning new customers — and they are by far the largest anyway. Hitching a ride on other popular services is not an obvious method.

One can also wonder how the content on X affects potential AI customers.

As a private individual it can be entertaining to read about marmots on OnlyFans, or Italian bank loans to farmers secured against parmesan cheese. In between, less pleasant things such as the occasional Nazi. A generous description would be “eclectic.” But it is easy to understand why many no longer want to take part.

That is precisely why the step of associating your company with a platform carrying this material is not so obvious. It can even become counterproductive. Advertisers have already fled, because they do not want to appear in that content environment.

And it is exactly those businesses that Musk and SpaceX need to reach with their AI services. So far, few private individuals pay for AI. The purchase of the coding service Cursor is one example of a bet on specialized offerings that companies might conceivably pay for. The new service Grokbot, launched this week, is described as a “colleague” — an AI that works on your company’s behalf.

X works well for raising awareness of these new products — but that is not enough. It has to turn into paying customers for the logic to hold together.

Much therefore suggests that X’s many acquisitions and relocations within Musk’s corporate portfolio were about finance more than strategy. Musk was the named buyer — but he brought a great deal of other people’s debt with him to complete the deal.

Those loans have now been settled, and converted into shares in entirely different companies. That is an achievement in itself. Billion-dollar borrowing taken on to buy an overvalued company like Twitter has now evaporated. And compared with that, X’s advertising revenue is a very small problem to solve.

What was once one of the world’s most important social networks has become a subordinate clause in Musk’s enormous corporate construction. It is simply not particularly important any more — neither to him nor to the space company at large.

Musk has his sights on bigger challenges these days. We are talking about a company whose goal is to build human colonies on Mars. He has his hands full there.