Out of the corner of shame — into the ‘strategic reserve’

SvD Näringsliv

This analysis was first published in SvD Näringsliv, in Swedish, on March 3rd, 2025. This piece was translated from Swedish by Claude. Some phrasing may differ from a human translation.

While Musk searches through the machinery of state looking for waste, Trump has found a new way to spend Americans’ tax money — on cryptocurrencies. This creates a new problem for those who believe in bitcoin.

“Many Americans don’t realise that the US government is one of the largest holders of bitcoin. Does anyone know this?”

Donald Trump is standing before an energised crowd at a bitcoin conference in Nashville in July 2024. He has just announced that he intends to fire Gary Gensler, a critic of cryptocurrencies at the country’s financial regulator. Jubilation erupts. He floats the idea of a strategic cryptocurrency reserve.

Fast forward to last weekend, and the shift is here. Trump announces he is going ahead with the strategic reserve. Yet the bitcoin enthusiasts are not happy. Why?

The reason the US already owns so much bitcoin is not strategic at all — it is simply the result of seizures and forfeitures in various criminal cases that have made the country a major holder of the cryptocurrency. The message that the US government would now accumulate even more bitcoin is one that lands well with bitcoin fans. It is the other cryptocurrencies they take issue with.

Over the weekend Trump posted on his social network Truth Social that he had asked a working group to proceed with the idea of a strategic reserve for cryptocurrencies. The currencies named were Solana, XRP, and Cardano. Trump then quickly followed up by adding that Bitcoin and Ethereum would also be included.

Outsiders tend to lump all of this together under the label “cryptocurrencies,” but within the industry the perspective is entirely different. Bitcoin — long claimed to be capable of functioning as an actual currency — has in recent years been repositioned as a kind of digital gold: an asset, rather than a currency. At the other end of the scale is the category more dismissively known as “shitcoins” — speculative projects that are neither asset nor currency, but something more akin to a lottery ticket with crypto technology at its core.

The underlying technology can therefore be used in different ways for different purposes. But the association between them contributes to the negative image many people have of cryptocurrencies as a whole. Enthusiasts had therefore hoped that Trump’s promised reserve would consist exclusively of Bitcoin — to distinguish it from the rest. Instead, it turned out to be the exact opposite.

How were the chosen currencies selected? Take XRP as an example. By the equivalent of market capitalisation it is one of the largest on the market, but the company behind it — Ripple — was investigated by the US financial regulator and paid over 1.3 billion kronor in fines. One possible explanation: Ripple spent over half a billion kronor on lobbying ahead of the presidential election.

In terms of money, the crypto industry as a whole was the second-largest lobbying force in the election cycle (after oil and gas companies). The industry saw an opportunity with Donald Trump as president — and now they are being paid for that investment.

A great deal of money has been invested in these cryptocurrencies and in the companies that work with them. Having the US government as a buyer is enormously positive for the entire ecosystem. This was visible immediately, as the price of several jumped by more than 20 percent in a single day. It amounts to a form of legitimisation for an industry that has had to stand in the corner for many years.

The biggest question of all is, of course, what the actual purpose of this reserve is. The executive order Trump signed claims that “the digital asset industry plays a crucial role in innovation and economic development in the United States.” There is very little to support that claim. If Cardano disappeared tomorrow, how many Americans would even notice? Nor is it a central part of the country’s innovation. None of the largest US tech companies deal in cryptocurrency at all — with the exception of Tesla, which holds a small amount of bitcoin.

The purpose most closely resembles a favour to an industry that gave Trump a great deal of support. And while Elon Musk tries to identify waste in the American state, tax money will now flow directly to buying cryptocurrencies — assets largely owned by a small number of people and companies. Beyond those specific beneficiaries, few can reasonably see any value in this reserve — strategic or otherwise.

Relief — but the threat from China remains

SvD Näringsliv

This analysis was first published in SvD Näringsliv, in Swedish, on February 27th, 2025. This piece was translated from Swedish by Claude. Some phrasing may differ from a human translation.

Nvidia’s strong results show that AI investment continues to grow. But the pressure mounts with each quarter, while concern over China’s DeepSeek lingers. And the key question remains.

This was probably not how Jensen Huang, Nvidia’s CEO, had envisioned his career.

In 1993 he founded a company called Nvision together with a few friends. They were going to make graphics chips for computers. The name had to be changed quickly, however, when it turned out to belong to a manufacturer of toilet paper. The new name — Nvidia — was borrowed from the Latin word “invidia,” meaning envy.

An enviable situation is, however, not exactly what either Nvidia or Huang finds itself in right now.

After spending some 20 years in relative obscurity on the stock market, interest in the company has now exploded. When Nvidia reported its quarterly results late on Wednesday evening, Swedish time, the world’s stock markets sat on tenterhooks with a single question in mind: do we have an AI bubble? And it was Huang’s job to give them the answer.

For virtually any other company, the numbers would have been compelling. Nvidia grew its revenues by 78 percent compared with the previous year, and beat market expectations on both revenue and profit. Over the full year, revenues grew by an astonishing 114 percent — a barely comprehensible achievement at these levels.

For Nvidia, however, that apparently was only to be expected. The stock market reacted with a yawn. The share price in after-hours trading barely moved.

The results were strong — but not strong enough to make markets forget the anxiety that has characterised the AI market for a couple of months now. Nvidia’s explosive share price growth has stabilised — admittedly at a very high level, but stabilised nonetheless. When we are talking about the world’s second-most highly valued company by market capitalisation, every such signal carries weight.

One source of that concern is spelled DeepSeek — the Chinese AI model that has taken the world by storm in recent months. The interest in DeepSeek has not been because it is better than other AI models, but because of how surprisingly capable it is given the limited resources used to create it. China faces trade restrictions on what type of chips it is permitted to buy. To work around them, it has had to develop differently.

This potential technical innovation could be seen as a positive development for AI, since it demands fewer resources. If you are Nvidia — essentially the only seller of those resources — it could instead become a problem.

“DeepSeek made us understand that Nvidia is not invincible,” said Shana Sissel, chief investment officer at fund Banrion Capital, to Bloomberg.

The quote says it all. No company is really invincible — but looking at Nvidia’s share price since January 2023, one might be forgiven for thinking otherwise. The increase is around 560 percent. The market capitalisation stands at over 32,000 billion kronor — roughly three times the entire Stockholm Stock Exchange combined.

Another potential concern is the trade barriers and tariffs that may be introduced in the United States. It is already forbidden for Nvidia to sell its best chips to China. But the company has developed an entire product line sitting right at the edge of what is permitted — and sells a great deal of it. China is a very important market for Nvidia, and if further tariffs or restrictions are imposed the impact could be severe. The US is also investigating whether China has managed to circumvent the ban in some way — for example by using data centres in Singapore, another of Nvidia’s major markets.

Nvidia’s quarterly results make clear that these clouds of concern have not significantly dampened AI investment appetite — not yet, at any rate. Even Nvidia’s forecasts for the coming quarter came in higher than analysts had expected.

In practice this means that tech companies continue to place large orders for Nvidia’s chips and services. And even if markets have wondered whether new, cheaper ways of building AI services might exist, nobody seems quite ready to cut back on investment pace just yet.

The world’s stock markets breathed a collective sigh of relief. Nvidia and Jensen Huang delivered yet another strong quarter. But the pressure is unlikely to ease. In three months’ time, the market will ask the same question it is asking now: can Nvidia really do it again?

Swedish media’s worst-kept secret

SvD Näringsliv

This analysis was first published in SvD Näringsliv, in Swedish, on February 25th, 2025. This piece was translated from Swedish by Claude. Some phrasing may differ from a human translation.

After the catastrophic Viaplay deal, media group Schibsted is getting its revenge. It is buying TV4 from Telia for 6.55 billion kronor. Schibsted Media’s transformation can now begin in earnest.

It has been Swedish media’s worst-kept secret.

Just over a year ago, Schibsted Media’s CEO Siv Juvik Tveitnes said the following to Dagens Media:

“If we are going to remain relevant — especially among younger users — we need to broaden out and invest more in both sport and entertainment.”

Breadth. Sport. Entertainment. In other contexts she mentioned moving images and geographical expansion to Finland. Given the shape of the Nordic media market, it was virtually impossible to achieve this vision without buying TV4 and Finnish MTV.

The negotiations were most likely primarily about what price Telia could accept. 6.55 billion kronor was the midpoint at which Telia CEO Patrik Hofbauer could justify refocusing the company on communications services once again.

Six years ago, Telia bought what was then Bonnier Broadcasting — TV4 and MTV — for around 10 billion kronor. That deal was two CEOs ago — an eternity in these contexts. Johan Dennelind signed the agreement and Alison Kirkby inherited it when she took over. It was clear even then that she was not convinced Telia should be in the content business — she had separated the two sides at her previous telecoms job. Now Hofbauer — who inherited the Schibsted negotiations when he took over a year ago — can finally close the TV chapter and move on.

For Telia’s part, the deal is fairly undramatic. The share price barely moved on the announcement. For Schibsted, it is a minor revolution — albeit a widely anticipated one, as noted.

Schibsted is a Norwegian media group that primarily operates newspapers, including Svenska Dagbladet and Aftonbladet in Sweden and titles such as Aftenposten and VG in Norway. Unlike virtually every other media company in the world, however, it became most known for the business that had nothing to do with media. By building successful marketplaces such as Blocket and Finn, the group became unevenly balanced: the marketplaces provided growth and at times high profitability; the media operations often provided the opposite. Schibsted — then listed on the Oslo Stock Exchange — had a difficult job explaining to the market why anyone should ever invest money in the media side.

The result was a split. The international marketplaces were spun off, and Schibsted subsequently sold a stake in them in a major transaction worth 24 billion Norwegian kronor. The Nordic marketplaces remained listed, while the major shareholder the Tinius Foundation bought the media side out from the exchange. That is where we are now.

Seen from this perspective, today’s TV4 deal is entirely logical. Is it the best financial investment available in corporate Sweden today? No. But the situation does not leave many openings. The owner is a foundation whose mandate is to run media operations. They want to stay within the Nordics. How many other major deals are there to be done? Essentially none.

The commercial media market in the Nordics consists of essentially a handful of players — Bonnier, Egmont, Aller, Sanoma, and Schibsted — and as a result very few companies change hands. Schibsted already tried to buy Viaplay once, last year, in a deal that ended with losing around 380 million kronor. If you want to grow significantly as a media owner in the Nordics, the alternatives are very limited.

The Tinius Foundation has now broadened its media ownership beyond newspapers. A different type of media company is beginning to take shape. Tinius has also opened its wallet significantly to enter the game. But to stay in it, they will in all likelihood need to do so again.

The challenge is that while the competing media owners are Nordic, the competition for advertising and viewers is very much global. A Schibsted-owned TV4 faces not only Bonnier but also players such as Netflix, Disney+ and Max. Viaplay has new owners who — despite their crashed share price — keep pressing on. How a local player meets these global tech giants is something nobody has really figured out yet.

Disclosure: Schibsted Media today owns Aftonbladet, Svenska Dagbladet, Omni and Podme in Sweden.

This is the definitive end

SvD Näringsliv

This analysis was first published in SvD Näringsliv, in Swedish, on February 18th, 2025. This piece was translated from Swedish by Claude. Some phrasing may differ from a human translation.

Following SvD’s report that Scania is buying a factory from Northvolt, the next question immediately arises: who — or what — might buy the rest?

It was painted as a corporate empire in the green transition — before the first factory had even opened.

Northvolt rapidly launched operations in Sweden, Germany, Canada, and Poland. Joint ventures, so-called “joint ventures,” could be found in both Norway and Portugal. It looked ambitious and large-scale. The momentum behind Northvolt was enormous — and why not strike while the iron was hot?

Then came the crisis and the restructuring. The large corporate group with tentacles in different parts of the world may now instead become a liability. What type of buyer wants to deal with all these operations simultaneously? Especially when so few of them are functioning at their intended capacity.

Scania is now acquiring a factory in Poland, according to SvD, for a fraction of the roughly 1.6 billion kronor invested there. That solution — applied on a broad front — is looking increasingly like a realistic path forward. It appears that no white knight is coming to lift Northvolt out of its restructuring in one piece.

In other words: a large Northvolt could end up becoming many small ones.

Finding major shareholders willing to step in and take on greater responsibility has been a challenge for Northvolt throughout the crisis. Before the restructuring, Volkswagen was the single largest shareholder with around 22 percent of the shares. But the German automotive giant had its own problems to work through before a rescue of Northvolt could receive sufficient focus.

In December, Volkswagen agreed with German trade unions not to close the factories it had previously announced it would, but 35,000 jobs will still disappear by 2030 — a saving of 168 billion kronor. In such circumstances, prioritising further money to save Northvolt was not possible.

Being a part-owner of something is also not the same as wanting to run the entire operation. For Volkswagen to be a good customer and place purchase orders as a form of guarantee is one thing. Being a permanent industrial owner of a half-finished production facility making battery cells and systems is quite another.

Northvolt may now be broken apart by individual industrial players, financial consortia, or suppliers and industry peers who see an opportunity to turn the operations around.

The three tracks — which can proceed in parallel — look like this:

The first is industrial players, exactly like Scania. They are often already parties to the process, have been following Northvolt for a long time, and know what they need and what they don’t. If they can secure part of their supply chain cheaply, it could be an attractive option. Another example following the same track is Volvo Cars, which in January bought out Northvolt from the joint venture Novo Energy in Gothenburg. And like Scania, it would not be inconceivable for truck manufacturer AB Volvo to be interested.

Track two is financial consortia. They tend to emerge opportunistically when deals are to be made. Private equity firms are certainly doing the sums on what price individual parts of Northvolt would need to be at to make a good investment. An educated guess is that Vargas — another major Northvolt shareholder — could become a player here. EQT — with its newly arrived CEO — is another potential option.

The third track would be if a direct competitor to Northvolt bought a factory, or alternatively one of their many suppliers. Among the competitors one finds names such as BYD and CATL. And the most important supplier is Wuxi Lead. What all three have in common is that they are Chinese.

For that solution to be possible, there is a further complicating factor. Not only do willing buyers need to be found — they must also be approved by the government agency ISP, the Inspektionen för strategiska produkter (the authority for strategic products). They have previously blocked Chinese ownership of similar projects in Sweden.

Last autumn I posed the question of who would want to save Northvolt from its crisis. The conclusion at that point was that it was likely cheaper to let the company go into restructuring than to put money in at that moment. Now we are exactly there.

The Scania deal suggests that there could be a solution in which the company is broken up into smaller pieces. A handful of battery factories, spread across the world, could then live on under new owners. But it would also mean something else: the definitive end of the corporate empire Northvolt, as we know it today.

Update: American company Lyten, which announced the purchase of Northvolt’s assets in August, was subsequently reported by SvD to be experiencing difficulties securing the financing for its acquisition — with the deal’s completion repeatedly pushed back from October to December to January, leaving the situation unresolved.

The high priest behind the DOGE philosophy

SvD Näringsliv

This analysis was first published in SvD Näringsliv, in Swedish, on February 16th, 2025. This piece was translated from Swedish by Claude. Some phrasing may differ from a human translation.

Elon Musk is turning Washington DC upside down with DOGE. Behind the idea of running the US like a tech company lies a Silicon Valley philosophy — shaped by a handful of people.

In the spring of 2012, a particular document was circulated among entrepreneurs and investors in Silicon Valley. The document — “CS183: Startup” — spread like wildfire, and its strange abbreviation quickly became something everyone in the tech world was talking about.

The name sounded cryptic but had a simple explanation. “CS183: Startup” was a course at Stanford, the university outside Palo Alto at the heart of Silicon Valley. CS stood for “Computer Science.” Student Blake Masters took the course and had summarised the lectures. The teacher in question was already well known in the tech sphere: Peter Thiel, co-founder of PayPal and the first investor in Facebook.

Thiel was already known as a free thinker and a “contrarian” — someone who enjoys challenging conventional systems of thought and ways of working and living. But in the Stanford course, he articulated his philosophy a little more concretely: how can these ideas be applied when starting a new company?

Blake Masters’ notes subsequently became a bestselling book, “Zero to One,” which he co-wrote with Peter Thiel. Published in 2014, it has become something of a manual for new tech companies around the world, joining a relatively small and loosely connected body of ideas that Silicon Valley and the tech world around it draw upon.

Now these ideas have reached all the way into the White House, and are redrawing the political landscape of the United States at breathtaking speed.

The concept of “zero to one” — taken from Thiel’s book title — is an example of something that today is understood by entrepreneurs worldwide. It refers to creating something entirely new rather than improving something that already exists.

Likewise with ideas such as “product-market fit” (when a company has an offering the market genuinely wants), “MVP” (“minimum viable product” — creating the smallest possible product to test whether something works), or “blitzscaling” (scaling a company at extreme speed, often by spending enormous amounts of money).

The ideas come from Silicon Valley, but their spread is global. Startup companies at incubators from Stockholm to Sydney know exactly what is meant when these concepts are mentioned. The ideas now spread more like internet memes than as courses and studies — a vocabulary whose underlying thesis is at best available in a book, but may be something as small as a blog post or a social media update.

For somewhere so economically influential, Silicon Valley is not a place that particularly values political philosophy or grand ideas. Peter Thiel would probably agree with that — he moved away himself in 2018.

Entrepreneurial culture has instead built an informal base of ideas and methods that companies work from — a startup canon, if you will. And despite it not being formalised, rarely do new names emerge within it. In a culture that likes to present itself as different from the rest of the world, it is often nearly identical to its competitors — and former colleagues.

If the concepts are relatively few, the originators are even fewer. It is a handful of people who have coined virtually all of them, and whose writings are regularly pored over by tech entrepreneurs. When venture capitalist Paul Graham wrote about the concept of “founder mode” last autumn — being an extremely hands-on and detail-oriented leader — it was as if Silicon Valley itself changed.

The blog post was seen as endorsing a different type of leadership style in business. It was now acceptable not to include everyone in decisions, or to take into account the views of particular individuals or groups. Instead, the situation was urgent — and as a founder, you needed to take command.

Graham had captured the zeitgeist on a larger scale than he perhaps first realised. Two months later, Donald Trump won the US presidency, and the dismantling of DEI and diversity initiatives began in American corporate life. In the front of the pack were the very same tech companies that had previously established committees and working groups to promote it — Google, Meta, Amazon. Over recent months, those initiatives have been shut down rapidly. And now Trump has appointed Elon Musk and DOGE to clear all of this — and much else besides — from the American federal agencies.

Who, then, are these originators, whose views and opinions carry such influence?

At the top of the pyramid we find Peter Thiel himself — the high priest, one might say. He has credibility both as an entrepreneur and as an investor, and beyond that the courage to be clear in his opinions. In 2016 he was one of the speakers at the RNC, the Republican National Convention — the year Donald Trump first became president — and in his speech he noted the similarities:

“I build companies and I support people who build new things, from social networks to rockets. I’m not a politician. But neither is Donald Trump. He is a builder — and it is time to rebuild America.”

When Donald Trump was sworn in as president in 2025, the stage was full of tech executives. That was not how it looked in 2016. Thiel led the way.

The next figure is venture capitalist Marc Andreessen. Together with Ben Horowitz he started the fund known as A16Z — an abbreviation of their combined surname, with 16 letters between. The aforementioned “product-market fit” comes from Andreessen, but he is perhaps most famous for a 2011 op-ed in the Wall Street Journal in which he articulated what became a widely-quoted phrase: “software is eating the world.” Software would come to affect every sector and industry, and would thus become the most important force in the economy.

Today, 14 years later, it is hard to argue against Andreessen’s thesis. Of the world’s ten largest listed companies, seven are related to software — even if some, like Apple and Tesla, are also hardware-oriented. A further two of the ten make the chips that software subsequently runs on.

Two other names that appear frequently are venture capitalist Paul Graham and Sam Altman. Today, Altman has become a minor celebrity following the global success of the company he helped found — OpenAI — with ChatGPT. In Silicon Valley, Altman — or “Sama” as he is known online — was well known long before that. Both Graham and Altman came from the startup incubator Y Combinator, a powerful force in the tech world. Being accepted into Y Combinator is like winning the lottery in the startup ecosystem, with the companies there being virtually guaranteed continued investment from surrounding venture capitalists.

And finally — in descending order of philosophical significance to Silicon Valley — there is Elon Musk himself: founder of the quasi-agency DOGE, which is currently cleaning out — or closing down — American political institutions.

Musk is more of a respected entrepreneur than a thinker. He has prioritised decisive action over intricate plans, and made himself known for sleeping in his factories to avoid wasting any time. It is the methods, rather than the ideas, that have made him influential. Musk takes on the hardest problems with a radical optimism.

Tesla built its own global network of EV charging stations before all its competitors. SpaceX catches its own rockets and makes them reusable. Neuralink gives people the ability to control computers with their minds alone. Musk chooses incredibly hard problems and throws as many working hours as possible at solving them.

Now Elon Musk faces perhaps his hardest challenge yet — at least if you take him at his word.

In an interview at the World Economic Forum in Davos in January, Musk said that “bureaucracy is the second-to-last boss fight” in the world — to use gaming terminology. The very last, he conceded, was overcoming entropy — something he acknowledged that the laws of physics do not allow. Bureaucracy was therefore, in practical terms, the greatest challenge.

Self-confidence is certainly not lacking. The Silicon Valley toolbox is packed with concepts and ideas.

But a state is, as is well known, not a company. Can the same methods used to clean up among bureaucrats really be the same ones used to build the world’s most highly valued companies?

The answer to that we are witnessing right now in Washington DC. Silicon Valley has temporarily changed coasts. The great boss fight is underway.

Caught off guard — now the US will hit back

SvD Näringsliv

This analysis was first published in SvD Näringsliv, in Swedish, on January 30th, 2025. This piece was translated from Swedish by Claude. Some phrasing may differ from a human translation.

The much-discussed AI model DeepSeek has crashed the stock market — and pushed tech development into a geopolitical phase. Now Trump and Zuckerberg are desperately searching for a countermove that can stop China from seizing the throne.

“I think there are a number of original things they did that we are still digesting.”

When Meta presented its quarterly report on Wednesday evening, there was another company the analyst community would rather have talked about: DeepSeek — the Chinese AI company that had rocked the global economy over the past week. How will things go for Meta and its AI projects, given that the Chinese competitors appear to be so capable?

The same question hung over OpenAI’s major shareholder Microsoft, which also presented its quarterly report the same evening.

“I think DeepSeek has some genuine innovations,” said CEO Satya Nadella.

It is obvious that they do not quite know how DeepSeek managed to achieve its results. And they appear to have been caught completely off guard by how quickly they found themselves facing a new Chinese competitor.

The answers from the two tech leaders also hint at something even more interesting: AI development is entering a new phase. Now it is about geopolitics.

Until now, the AI market has been dominated by a handful of large American companies. Microsoft and OpenAI with ChatGPT have been pitted against Google’s Gemini and Meta’s Llama, with individual challengers like Anthropic also in the mix. Virtually all the major players have their home on the American West Coast and have competed with one another in full view.

Now the view of competition is shifting. Rather than pitting OpenAI against Google, it may increasingly become a national question — the US against China. Well-known venture capitalist Marc Andreessen called DeepSeek the AI world’s “Sputnik moment” — the point at which it suddenly became clear that there were more candidates ready to compete seriously.

In the same analyst call on Wednesday, Mark Zuckerberg made his position plain:

“There will be a new global standard for open-source AI. And for it to be advantageous for us nationally, it is important that it be an American standard.”

What was previously almost self-evident now needs to be stated clearly — because it is being called into question.

The issue has become particularly pressing in the wake of this week’s sharp fall in Nvidia, the chip company whose fortunes are closely tied to AI development. Meta plans to invest around 65 billion dollars — roughly 714 billion kronor — on AI in 2025 alone. Much of that money will flow directly to Nvidia.

The development is driven largely by anxiety about the Chinese AI model DeepSeek R1, which has performed extraordinarily well in comparative benchmarks. What spooked the markets, however, was not its performance — but the cost of developing it. That cost was said to be very low compared to its American equivalents, and moreover it was produced using inferior chips: a necessity for Chinese companies, since the best chips are not permitted to be exported to China.

Whether the claimed low cost is accurate is disputed. There are likely substantial hidden costs that have not been disclosed. Both DeepSeek and China as a whole have every incentive to present themselves as innovators in this space. But the uncertainty alone was enough to send Nvidia’s share price into sharp decline. In a single day, 6,440 billion kronor was wiped out — though the price partially recovered the following day.

Shortly afterwards, Nvidia’s share price was rattled again. This time it was the United States that drove the fall. Donald Trump is reportedly discussing tighter controls on what type of chips can be exported at all. China is Nvidia’s second-largest market, and further trade barriers of this kind could hit the company hard.

China is now regarded by many — for the first time — as a fully-fledged rival to the US in AI. China has long had AI development with several large models that have performed well — among others from tech giants Baidu and Alibaba. But none has broken through in the same way as DeepSeek. It is also possible that China’s hardware constraints have been compensated by innovation on the software side.

It is through this lens that one should understand Trump’s discussions about increased restrictions on Nvidia. This is a larger question than the profit-and-loss accounts of individual companies — it is now about the country as a whole. And the US is a country that does not like to lose.

Fear of DeepSeek shakes the stock market

SvD Näringsliv

This analysis was first published in SvD Näringsliv, in Swedish, on January 27th, 2025. This piece was translated from Swedish by Claude. Some phrasing may differ from a human translation.

Hundreds of billions have been invested in AI chips, making Nvidia the world’s most highly valued company. Now a small Chinese player is challenging the need for those investments — and the markets are beginning to shake.

On 27 March 2000, Cisco became the world’s most highly valued company. The American networking giant was riding high on dotcom waves and the internet economy promised growth and success. Cisco was selling the picks and shovels of the gold rush — networking equipment. As everyone around the world scrambled to get connected, it looked like a sure winner.

Everyone remembers how it ended. Less than a year later, Cisco’s share price had fallen by 80 percent and the crash was a fact. Now the markets are shaking and the anxiety is clear: are we facing a similar scenario once again?

The Chinese AI model DeepSeek R1 is giving markets a sense of déjà vu. Over the weekend, DeepSeek reached the top spot in the American App Store, overtaking rival ChatGPT. Venture capitalist Marc Andreessen called it the “Sputnik moment” of the AI world — the moment when the world understood that there were other players in the game.

The appeal for consumers is straightforward: DeepSeek is good and free. But it is how DeepSeek was built that is truly disruptive. It is said to have used far fewer resources than comparable competitors while still delivering equivalent results. Should that prove to be true, it could be bad news for companies like Nvidia, which has prospered handsomely from tech giants’ investments in chips and data centres. Will as many really be needed going forward?

Nvidia — today the world’s most highly valued company — carries expectations of growing, not shrinking, demand. When markets opened on Monday, the company’s share fell by around 13 percent — representing a loss of value of approximately 465 billion dollars, the largest single-day loss in stock market history.

Making matters worse, China faces strict restrictions on what type of chips it can import from Nvidia. DeepSeek therefore appears to have been developed using inferior chips compared to those available to Meta and Google.

A great many uncertainties remain. These range from suspicions that DeepSeek stole data and information from ChatGPT, to claims that the company behind it does in fact have access to the best chips but cannot admit it for understandable reasons. Meta has assigned extra staff to try to work out how it succeeded. But the uncertainty is sufficient to rattle the markets. Dutch company ASML, which makes machines for manufacturing chips, fell around nine percent when European exchanges opened.

If it turns out that the Chinese have found a more efficient way to develop AI models, it can be interpreted in two different ways.

The first is what underlies the anxiety described above. Do we really need more and better chips if we can achieve the same type of results with far fewer? Is this a potential paradigm shift in how AI models are trained? It could mean a reduced need for chips and data centres, and lower investment required to remain competitive. It would also mean that high-level AI development could take place in many more countries, moving beyond today’s heavy concentration in Silicon Valley.

The second perspective is what is known as Jevons paradox. Named after the English economist William Stanley Jevons, who showed in 1865 that increased efficiency in the use of coal did not reduce demand for the raw material — but rather increased it. Applied to AI development, this would mean that the need for capacity and chips will only grow as the stakes get larger and more players can participate. Microsoft CEO Satya Nadella wrote on Monday that he believes this is precisely a Jevons paradox moment.

Looking back at Cisco in 2000, there are further parallels to be found. The direction for the future proved to be correct — the internet economy would indeed become world-changing. But it took considerably longer than anyone had expected. And the immediate need for Cisco’s hardware did not materialise on cue. Only now — 25 years later — has Cisco’s share price caught up and is approaching the same heights it once reached. But more than anything else, Cisco’s dotcom valuation was based on near-infinite growth projections. Enthusiasm overrode reason.

The fact that a chip manufacturer like Nvidia is the world’s most highly valued company speaks to a similar enthusiasm today. It is not sufficient, necessarily, that AI is here to stay, or that it may come to revolutionise society and business life. Timing also matters. And as DeepSeek is reminding markets right now: there may be more ways to reach that vision than buying vast numbers of extraordinarily expensive chips.

Decision-makers know something we don’t

SvD Näringsliv

This analysis was first published in SvD Näringsliv, in Swedish, on January 20th, 2025. This piece was translated from Swedish by Claude. Some phrasing may differ from a human translation.

TikTok is banned — and permitted again a few hours later. The law says one thing and the incoming president says another. But both American and Chinese decision-makers have one thing in common: they know something we don’t.

“TikTok is definitely a threat to national security. We have seen classified evidence.”

Democrat Josh Gottheimer speaks with absolute certainty. The American congressman from New Jersey is one of the politicians behind the law banning Chinese ownership of TikTok, among other things. When the law came into force over the weekend, it should have felt like a victory. It didn’t quite work out that way. When the incoming president said he wanted to circumvent the law immediately, the feeling of a political win evaporated.

What we are left with instead is a long list of questions.

The first concerns the classified evidence. Gottheimer, along with several other politicians, has referred to information that American politicians have been given access to. It is said to have been so persuasive that both Democrats and Republicans united around the conclusion that TikTok poses a security threat. In the House of Representatives, the law passed by the clear margin of 352 to 65.

American politicians are, however, among the small minority of people who know what this evidence actually consists of. None of it has been shared with the public. It is unclear whether the evidence has even been shared with TikTok itself, so that it could respond.

The situation is therefore rather strange. There are 170 million monthly users of TikTok in the US, all wondering what dangerous activities TikTok may have been engaged in. There are a few hundred politicians essentially saying “trust us — we’re doing you a favour by banning TikTok.” As an outsider it is impossible to assess how serious this evidence is. But one can clearly observe that the method for convincing the American public has been sorely lacking.

The second question concerns TikTok’s own motives. The law does not ban TikTok as an app — it bans its Chinese ownership. Parent company ByteDance, a commercial enterprise with investors from around the world, therefore had every reason to sell the app during the period from last spring until now. So far, it has refused.

TikTok’s value has been estimated at around 40 to 50 billion dollars, or roughly 450 to 550 billion kronor. Regardless of whether they think the law is wrong or not, it is usually in commercial companies’ interest to do what maximises revenues and profit.

Unless other motives are at play, of course.

The Chinese state owns 1 percent of the shares in ByteDance. In the normal course of events, such a small stake would not be able to block a sale or significantly affect the company at all. The structure is known as “golden shares” — a name referring to the disproportionate influence they carry over a company’s operations. The method is common in China, and in January 2023 the Chinese state also purchased golden shares in subsidiaries of the Chinese tech giants Alibaba and Tencent.

Republican senator Lindsey Graham has written on X that it is precisely the golden shares that are preventing the TikTok situation from being resolved. He adds that he will propose legislation banning any company that holds these golden shares from trading on American stock exchanges — though this would not affect ByteDance, since it is not publicly listed.

The absence of business logic undermines TikTok and ByteDance’s arguments. If they genuinely have nothing to do with the Chinese state — then what company would turn down 500 billion kronor? Here too, they know something we don’t.

The third question concerns how clearly content on TikTok can be controlled, and to what extent this is already happening today. It is worth recalling that TikTok does not exist in China at all, since much of the information shared on it would not be legal there. ByteDance does, however, own a similar app, Douyin, which operates in China — though it prioritises content differently from TikTok and is described as having a profile that leans more towards educational material.

What is shown on TikTok is governed by an algorithm — a kind of data system that tries to predict which videos each user might enjoy. The feed is adjusted based on feedback. As SvD has previously reported, content is selected based on what you actually watch, not what you have expressed an interest in. Getting drawn into clips that encourage eating disorders, for example, can lead to ever more of them appearing — even if you try to avoid that type of content.

Beyond individual calibration, it is also possible to steer what content is shown from a central level. A study from the Network Contagion Research Institute at Rutgers University demonstrates that this is already happening today. The study was originally released in August 2024 and received a good deal of criticism — from TikTok itself in particular. In December an updated version was released with twice the amount of source data, and the study will also be published in a peer-reviewed scientific journal.

The study shows that when searching for topics considered controversial by the Chinese state — the protests at Tiananmen Square in 1989, for example — there is markedly less material on TikTok than on comparable social media platforms. One of the study’s authors, Joel Finkelstein, told The Free Press that “the scaled indoctrination is not hypothetical, it is real.” A search on Instagram about the Uyghur people showed negative sentiment towards China in more than 80 percent of all videos. The corresponding figure for TikTok was just 11 percent.

A TikTok spokesperson strongly criticised the study and considered it to have been designed to reach a predetermined conclusion. But the questions remain: is content on TikTok being steered to achieve a political goal? To what extent is this happening? And even if it is not happening today — could it happen in the future?

The fourth and final question concerns Donald Trump. During his previous term as president, he proposed banning TikTok. Now he is being portrayed as the saviour. TikTok is thanking him in the app itself before he has even taken office as president. Why does he suddenly want to save TikTok? And is it even possible for him to do so without breaking American law?

Most likely, Trump will push through a temporary three-month extension — a kind of reprieve while a commercial deal is being prepared. Given the questions above, the negotiations will be difficult. American authorities are demanding a deal; Chinese authorities are opposed to one. Both cannot get what they want. But by deferring the matter for a few months, Trump becomes a temporary winner and restores normalcy for the 170 million users here and now. The underlying problem remains, however: the law has been passed and the Supreme Court has ruled it constitutional.

Viewed from outside, the situation most resembles a spectacle — with states, politicians, and companies swinging back and forth in an unusually chaotic manner.

For those of us watching, several central pieces of the picture are missing. Are we in the middle of an information war between the US and China? The answer to that question appears to be classified. That is a shame. Everyone would benefit from understanding the motives at play — and who stands to win in this dispute.

Update: President Donald Trump subsequently approved a deal on TikTok in the United States through an executive order, with Trump telling reporters at the White House that Chinese President Xi Jinping had also agreed.

China’s budget AI is changing the market

SvD Näringsliv

This analysis was first published in SvD Näringsliv, in Swedish, on January 19th, 2025. This piece was translated from Swedish by Claude. Some phrasing may differ from a human translation.

Countless billions have been spent by tech giants to develop the best AI models. Now the next era is beginning — with budget versions from China that are completely changing the rules of the game.

“Have you got a Panadol?” Ever since paracetamol was launched in Sweden in 1958, the brand has remained synonymous with painkillers. Competitors like Ibuprofen have tried to challenge that image with the help of amusing advertising.

But it helps to be first to market, to have a well-known brand, or a reputation for being the best. All three at once is, of course, even better.

The young AI industry is in a similar position. Microsoft has invested over 145 billion kronor in OpenAI to stay at the absolute cutting edge — and to be both first and best. The rather square name ChatGPT has, somewhat unexpectedly, become the most well-known name for an AI product so far. But that could change quickly.

There is much to suggest that the market for AI products is now facing a major transformation. Like industries such as fashion, pharmaceuticals, and interior design, it will become broader — and substantially cheaper. And as in many other industries, it is China that is driving the development.

One new AI model has brought the question into sharp focus. Chinese DeepSeek V3 has only existed since December last year, but has already attracted a great deal of attention. The reason is not primarily that it is good — though in benchmarks it holds its own admirably against models from OpenAI, Meta, and Google. No, the reason for the attention is that it was extraordinarily cheap to build. It cost under six million dollars and took two months to put together — a pittance compared to the billions invested in today’s market leaders. Former Tesla AI chief Andrej Karpathy noted that DeepSeek should have needed at least eleven times more computing power to achieve these results.

As is often the case, the Chinese version closely resembles the original. DeepSeek looks and functions, in all essential respects, exactly like ChatGPT.

There may be a logical explanation for this.

When TechCrunch asked DeepSeek which AI model it had used, it answered that it was ChatGPT. No explanation for this obvious confusion has been given, but a fairly safe guess is that they copied material from their Western competitor in some way.

Regardless of the method — and however well that squares with copyright law — DeepSeek will have major implications for the AI market. A new category has emerged: so-called “fast followers.” Like H&M and Zara in the fashion world, they take inspiration from leading brands and produce their own version that sells quickly and cheaply. They are rarely the best, but as long as they are good enough, there are plenty of customers for this segment too.

The change may further polarise AI development. A fast follower has neither the ambition nor the resources to advance the category further. The major technical breakthroughs will therefore not come from here. On the contrary, more time and money than ever may need to be invested in creating a clear distinction between the leading AI models and everything else on the market. The most advanced version of ChatGPT currently costs around 2,200 kronor a month for users. Why would anyone pay that if a much cheaper version is almost as good?

DeepSeek itself is moreover built on open source code — like Meta’s Llama model — and is therefore free to use. But as with many free products, there is still a kind of cost. When you ask the Chinese AI model what happened at Tiananmen Square in 1989, it simply replies that “unfortunately this falls outside my current knowledge. Let’s talk about something else.” Chinese companies must comply with Chinese laws, even when users in Sweden are asking the questions.

Despite these obvious drawbacks, a clear and new direction for the AI market can be discerned. The barriers to entry just dropped significantly. What was once a business requiring billions in resources from the world’s largest companies has now made highly capable models available to the general public. Curious businesses can experiment at very low cost and achieve results that would have seemed incredible just a year ago. That is how fast this market is moving.

Keeping pace with AI development will still be a challenge. But now it is at least not your wallet that determines whether you can try.

Will Trump or MrBeast save TikTok?

SvD Näringsliv

This analysis was first published in SvD Näringsliv, in Swedish, on January 17th, 2025. This piece was translated from Swedish by Claude. Some phrasing may differ from a human translation.

The deadline expires this weekend. But will one of the world’s most popular apps really disappear? SvD’s tech analyst Björn Jeffery explains what happens when the espionage-accused TikTok is banned in the United States.

Yes — at present it looks that way. The US law was voted through in April 2023 and set a final deadline of 19 January, the day before Donald Trump is sworn in as president. The law does not actually regulate whether TikTok (or similar apps) may exist or not, but is aimed at its Chinese ownership. TikTok’s Chinese parent company ByteDance could therefore sell the app and thereby allow it to continue on the American market — but so far it has chosen not to do so.

There are still a few lifelines left for TikTok. The first and most important was the US Supreme Court, which considered whether the law violated the country’s freedom of expression — the so-called “first amendment.” TikTok appealed all the way to this level. On Friday, however, the definitive ruling came: the Supreme Court would not block the ban. The decision was expected, and the legal process is now formally concluded.

The second lifeline is whether some form of sale can be arranged. The law allows for a temporary three-month extension if such a deal is in progress but has not yet been completed. ByteDance has previously said a sale is out of the question, but that was when the outlook for TikTok looked brighter. Now that the Supreme Court route has not worked out, it is not impossible that they will apply for such an extension to explore a potential sale after all.

Several names have been mentioned as potential buyers or interested parties — among them Elon Musk, Bobby Kotick (the former CEO of gaming company Activision Blizzard), and Frank McCourt, a billionaire linked to the political movement Project Liberty. In recent days, even the YouTuber MrBeast has expressed interest in buying TikTok.

The third lifeline is the incoming president, Donald Trump. He is a self-styled dealmaker and has said he is interested in finding a solution — some form of partnership or sale. One can also imagine him issuing a so-called “executive order,” a mechanism that allows the president to prevent the law from being enforced in practice. TikTok’s CEO Shou Zi Chew has been invited to Trump’s inauguration, which can be seen as a positive signal.

Time is, however, extremely tight — and even if a new solution is found, there may be a gap during which TikTok disappears for users for a while.

The law has been known about for a long time, but many users assumed it would have no practical effect. The hope was that a sale or something similar would allow the service to continue running as before.

Over the past week it has become clearer that the end may be near. As a signal to American politicians, TikTok users have begun downloading and using the Chinese app Xiaohongshu instead. The app — previously virtually unknown in the US — reached the top spot in Apple’s App Store after the influx of TikTok users became so large. Xiaohongshu literally means “little red book,” which is itself a nod to Mao’s Little Red Book. If the original purpose of the law was to reduce the Chinese Communist Party’s influence on American citizens, it has at least temporarily backfired.

Setting aside the peculiar phenomenon of Xiaohongshu, TikTok’s users will quickly need to find a new, stable platform to use. Many use the service commercially and make their living by creating video clips, and businesses rely on it for marketing.

Most likely, the majority will switch to Instagram’s Reels — a product that is essentially a copy of TikTok. It has achieved considerable success, but has not fully broken through for the most prominent creators. Now there is an opportunity for Meta — Instagram’s owner — to restart with this influential group. Adam Mosseri, head of Instagram, has already announced that they are working on improvements to video editing and other tools used by creators.

Update: President Donald Trump subsequently approved a deal on TikTok in the United States through an executive order, with Trump telling reporters at the White House that Chinese President Xi Jinping had also agreed to the arrangement.