This analysis was first published in SvD Näringsliv, in Swedish, on February 23rd, 2026. This piece was translated from Swedish by Claude. Some phrasing may differ from a human translation.

Sweden’s new AI strategy is long on words and short on numbers. The government prioritises talking about how important AI is rather than showing it in action. To reach the top of the world, something absolutely crucial is missing.

So it finally arrived — Sweden’s AI strategy.

It was released on a Friday afternoon — hardly prime time for major announcements. That is rather when you tend to reveal things you would prefer people to forget, or at least miss. Once you read the strategy, the timing becomes easier to understand.

There are no bombshells here. It is 24 airily written pages, full of self-evident truths. For the concrete goals that are mentioned, a plan — and a budget — for how to achieve them is usually absent.

There are also 12 pages of action plan, in which “action” is a generously applied term.

The following quote is telling: “The Government has commissioned an inquiry into privacy-preserving methods for a more data-driven and collaborative public administration.” Sure, an inquiry is some kind of action. But it was probably not what many were waiting for when the strategy was finally presented — a year after it was supposed to have come into force.

Strategy is, as we know, a vague concept. One way to assess a strategy is therefore to consider whether there is anyone who could say the opposite of what is being claimed. Only then have you distinguished yourself.

We can test the exercise on our country’s new AI strategy. Here are some quotes from it. Would any country actually say the opposite — aiming for worse data management, for example?

Another remarkable thing is that there is barely a single figure mentioned in the material. In the strategy itself there are none at all, and in the action plan there are a few individual items but without an overall budget framework. Vinnova is investing 570 million kronor in “Advanced digitalisation” and 118 million kronor is to be invested over four years in expanding fibre-optic cables.

We have to go to an opinion piece from the Tidö coalition parties from September 2025 to see what they actually planned to allocate. There it states that 479 million kronor will be invested in 2026, and around 500 million kronor per year from 2027–2030. Rounded up, that comes to around 2.5 billion kronor.

That might sound like a lot of money, so let us put this figure in perspective. 2.5 billion kronor is approximately 0.9 percent of the amount invested in AI company Anthropic in its most recent funding round a few weeks ago. And that is talking about just one single American company.

We can compare with other countries too. The United Kingdom is investing roughly 10 times more than Sweden. Canada even more — around 30 billion kronor. Even tiny Singapore, with half Sweden’s population, is investing around 7 billion kronor.

There are other resources besides money, however, and here there are some bright spots.

Together with the Wallenberg Foundations, a Swedish language model is to be developed. An interesting initiative, particularly from the perspective of digital sovereignty.

The model is to safeguard the Swedish language, our culture, and our norms. The work is being done together with writers, media companies, and publishers to ensure copyright remains intact. That might sound like a matter of course, but it is substantially more than the American equivalents have managed — or even attempted to achieve.

The Swedish language model is an ambitious project that draws on WASP, Sweden’s largest individual research programme, which is largely financed by the Knut and Alice Wallenberg Foundation. Here the Swedish state will contribute data — a welcome and encouraging move. The Royal Library is also involved — an excellent partner for this.

Another area that sounds promising is the billions that could potentially be saved annually by making the public sector more efficient with AI. The ambition to become the best in the world at using AI in public administration is admirable. Here there are resources that could be freed up for more important things than administration.

Finally, let us return to the opening of the strategy, in its very first sentence: “Sweden shall be among the ten leading nations in artificial intelligence (AI) in the world.” Beyond the fact that it does not specify exactly how we will measure this, further questions are laid bare.

What happens to this ambition after this year’s general election? Who will pay for it all? Do we as a country have a clear timetable for this work?

There are no clear answers — either in the strategy or in the action plan. Instead we got yet another paper product full of self-evident truths.