The AGI Waiting Room: Why Swedish Companies Burn the Wrong Clock

Everyone is waiting. Almost nobody should be.

Demis Hassabis just pushed AGI timelines back again. Sequoia's David Cahn says 2026 will split companies into those that built real AI infrastructure and those that didn't. The Guardian ran a piece about trillions in AI investment with no guaranteed return. And in every single one of these global conversations, zero Swedish companies appeared.

Zero.

That last fact should bother you more than any AGI timeline debate. Because it tells you something specific about where we are in Sweden right now. We are not in the game. We are in the waiting room. And we chose to sit down.

The two failure modes

There are exactly two ways to get AI wrong in 2025. Silicon Valley is demonstrating the first one: pouring hundreds of billions into infrastructure bets predicated on AGI arriving on a schedule that keeps sliding. Nvidia's market cap assumes a future that may take fifteen years, not five. That is a real risk and people like Cahn are right to flag it.

But here in Sweden, we are demonstrating the second failure mode. And honestly, it is worse.

The second failure mode is watching the AGI debate from a distance, seeing smart people disagree about timelines, and concluding that the rational move is to wait. To form a committee. To commission a report. To do a pilot with a consultancy that charges 2 million SEK to tell you what GPT-4 can do.

This is not caution. This is paralysis dressed up as strategy.

The thing Swedish decision-makers keep missing is simple: you do not need AGI to transform your operations. You do not need it to cut your customer service costs by 40%. You do not need it to automate your document processing pipeline or build internal tools that actually work. The AI that exists right now, today, the boring kind, is already enough to create massive competitive separation. And every month you wait, the gap widens.

What I see from Jönköping

I run HEIMLANDR.IO from Jönköping. Not Stockholm. Not San Francisco. Jönköping. When I talk to Swedish CEOs and CTOs, I hear the same things over and over:

"We're watching the space closely."

"We want to wait until the technology matures."

"We did a ChatGPT workshop in Q3."

A ChatGPT workshop. In Q3. Of 2024. And they think they are being responsible.

Meanwhile, a mid-size logistics company in Rotterdam has already deployed AI agents that handle 60% of their procurement communications autonomously. A 200-person SaaS company in Austin rebuilt their entire onboarding flow with AI and cut time-to-value from 14 days to 3. These are not trillion-dollar moonshots. These are practical, boring, profitable applications of current technology.

Swedish companies are world-class at process optimization. We built Lean manufacturing cultures before it was a buzzword. We have incredibly high digital literacy. The raw ingredients are all here. But we are sitting on them because the AI discourse in Sweden has become contaminated by two things: AGI hype (which makes everything feel speculative) and EU regulatory anxiety (which makes everything feel risky).

Both are real concerns. Neither is a reason to do nothing.

The Nordic gap is real and growing

Let me put this in context. The US is spending like a country at war. China is building AI infrastructure at state scale. The UK, despite its size, has positioned itself as an AI regulatory sandbox and attracted real talent. Even smaller economies like the UAE and Singapore are making outsized moves.

Sweden? We have good universities. We have Ericsson doing interesting things in telecom AI. We have a handful of startups. But at the company level, at the level of the 10,000 mid-size Swedish businesses that form the backbone of this economy, AI adoption is shockingly low.

Breakit covers Swedish tech closely, and even there the AI coverage tends to focus on what is happening in the US, not what Swedish companies are actually building. Because there is not enough to cover.

Norway has its sovereign wealth fund thinking about AI allocation. Denmark has strong public-sector AI experiments. Finland has been running AI education programs at population scale since 2018. Sweden has... Vinnova grants and a general sense that someone should probably do something.

The political class is not helping. The EU AI Act is well-intentioned but it is creating a compliance fog that gives risk-averse Swedish boards one more reason to delay. "Let's wait until the regulatory framework is clear." This is a trap. The companies that will thrive under EU AI regulation are the ones building now, because they will understand the technology well enough to comply efficiently. The ones waiting will face both a capability gap AND a compliance scramble when they finally move.

What the AGI timeline debate actually means for you

Here is my honest read on AGI timelines, for what it is worth.

I think Hassabis is right to push back on aggressive timelines. The problems remaining in reasoning, planning, and genuine understanding are not just engineering problems. They are research problems. Hard ones. Scaling alone will not solve them, and the people who say "just add more compute" have been wrong before.

But this is completely irrelevant to what you should be doing right now.

Think about it this way. If AGI arrives in 2028, you need to have your entire organization AI-literate and AI-integrated by then, or you will be obliterated by competitors who did. If AGI does not arrive until 2040, you still need current AI capabilities deployed across your operations because your competitors in Germany, the Netherlands, and the US are deploying them right now.

Both timelines demand the same action: start building today.

The only scenario where waiting makes sense is the one where AI turns out to be a complete dead end. Where LLMs plateau entirely, where no further progress is made, where the whole thing was a bubble. If you genuinely believe that, fine. But you should probably also short Nvidia and go on record, because that is an extremely contrarian position in mid-2025.

Where this goes: 2025 to 2030

Here is what I think the next five years look like, and what it means for Swedish builders.

2025-2026: The sorting. This is where Sequoia's split happens. Companies that built real AI workflows and AI agents into their operations will start showing measurable advantages in cost structure, speed, and customer experience. Companies that ran workshops will start feeling the pain. M&A activity will pick up as AI-mature companies acquire AI-lagging competitors at discount.

2027-2028: The integration wave. AI stops being a separate initiative and becomes infrastructure. Like the cloud transition, except faster. Companies that are still "evaluating AI strategy" in 2027 will be in the same position as companies that were still "evaluating cloud strategy" in 2018. Alive, but increasingly irrelevant.

2029-2030: The capability cliff. Whether or not we reach AGI, AI capabilities will be dramatically beyond what we have today. Autonomous agents handling complex multi-step business processes. AI systems that genuinely reason about novel situations. The companies that have been building and iterating for five years will be able to adopt these capabilities in weeks. The companies starting from zero will need years. The gap becomes permanent.

For Sweden specifically, I worry about the mid-market. Our big companies (Ericsson, Volvo, Spotify) will figure it out. They have the resources and the global exposure. But the 50 to 500 person companies, the manufacturers in Småland, the SaaS companies in Gothenburg, the logistics firms in Malmö. These are the ones at risk. And they are the ones I spend most of my time thinking about at HEIMLANDR.

The regulatory gap nobody talks about

The EU AI Act focuses heavily on risk classification and transparency. Fine. Important, even. But it almost entirely ignores the competitive dynamics of AI adoption speed.

There is no EU policy for what happens when 30% of mid-size European companies lose their cost advantage to AI-native American competitors over a three-year period. There is no framework for the workforce disruption that comes not from AI replacing jobs, but from AI-integrated foreign competitors taking market share from European companies that moved too slowly.

Swedish policymakers specifically need to wake up to this. We regulate the risks of AI adoption but we do not even measure the risks of AI non-adoption. That is a blind spot big enough to drive a recession through.

What to actually do (the practical part)

If you are a CEO or CTO at a Swedish company reading this, here is what I would do this quarter:

First, pick one painful process and automate it with AI. Not a pilot. Not an experiment. A real deployment that handles real work. Customer support triage, document classification, internal knowledge retrieval. Something where you can measure before and after. We build these kinds of AI solutions at HEIMLANDR and the pattern is consistent: the first real deployment changes the conversation inside a company more than any strategy deck ever will.

Second, build internal AI literacy from the top down. Your board needs to understand what current AI can and cannot do. Not from a McKinsey slide. From using it. Every executive should be using AI tools daily for their actual work. If your CFO has not used AI to analyze a financial model, you are behind.

Third, budget for AI infrastructure in 2026 now. Not "we'll see." Now. The companies that win the sorting will have dedicated AI budgets, dedicated AI roles, and dedicated AI infrastructure. This does not mean billions. For a 200-person company, it might mean 2-5 million SEK annually. That is a rounding error compared to the cost of falling behind.

What to look at

Some specific resources worth your time this week:

LangGraph on GitHub. If you are building AI agents (and you should be), this is the most practical framework for creating agents that can handle multi-step workflows with human-in-the-loop checkpoints. Your engineering team should be prototyping with this.

Model Context Protocol (MCP). Anthropic's open standard for connecting AI models to external tools and data sources. This is becoming foundational infrastructure for agentic AI. If your team is not familiar with it yet, that is a gap.

Sequoia's AI writing, particularly David Cahn's pieces on AI infrastructure economics. He is one of the clearest thinkers on where AI investment makes sense and where it does not. Read him even if you disagree.

Rapid MVP development. If you want to test an AI-driven product idea or internal tool without a six-month procurement cycle, this is what we do at HEIMLANDR. Get something real in front of users in weeks, not quarters. The speed of learning matters more than the perfection of the plan.

Stop waiting for permission

The AGI debate is fascinating. I follow it closely. I have opinions about scaling laws, reasoning capabilities, and whether transformers are sufficient architecture for general intelligence. These are important questions for researchers and long-term investors.

They are not important questions for you right now.

What is important is that your competitors, somewhere in the world, are not waiting. They are building. They are deploying. They are learning by doing. And every quarter you spend in the AGI waiting room, they are compounding their advantage.

Sweden has every ingredient needed to be an AI leader. Talent, infrastructure, digital maturity, a culture of pragmatic engineering. What we lack is urgency. And urgency is the only ingredient that matters when the clock is already running.

From Jönköping, I can tell you: the view from here is clear. The technology works. The opportunity is real. The risk of waiting far exceeds the risk of starting. Stop reading the AGI tea leaves and go build something.

Fredrik Brunnberg is the CEO of HEIMLANDR.IO, building AI and software solutions from Jönköping, Sweden. This is the daily HEIMLANDR briefing. If you found this valuable, share it with someone who builds things.