Disruption Is a Rounding Error. Pipes Are Where the Money Moves.

I got a deck last week from a startup pitching "AI-powered transformation for the insurance industry." Forty slides. Zero customers. Meanwhile atNorth quietly opened a new metro data centre in Sweden this week, no keynote, no confetti, just capacity going live because someone needs to run compute somewhere. Guess which one is actually making money right now.

This is the thing nobody wants to say out loud at the conference stage: industry disruption is a rounding error compared to what's happening underneath it. The real 2026 money is in the boring layer. Data centers. Cross-border trade infrastructure. Government-to-government tech deals. Pipes, not promises.

The Transformation Pitch Is Dead, Long Live the Pipe

BCG published another piece this week arguing transformation is "a must" for Nordic companies. Sure. But read between the lines and what they're actually describing is anxiety, not strategy. Every Nordic company knows they need to "do something" with AI. Almost none of them know what that something produces in revenue this year. That gap between anxiety and revenue is exactly where the vertical AI pitch decks live, and it's exactly why so many of them stall at pilot stage.

Compare that to what's happening at the infrastructure layer. atNorth isn't selling anyone a transformation story. They're selling megawatts and rack space, and Swedish digital capacity needed it yesterday. Business Sweden just opened an office in Peru, not because Peru needs a fintech app, but because trade corridors need infrastructure, legal scaffolding, and someone to make the introductions between governments. Latvian delegations are in Sweden this week signing tech and defence cooperation agreements. None of this is disruption. All of it is plumbing. And plumbing is where the contracts get signed, because plumbing is what everyone needs regardless of which AI model wins the benchmark this month.

The pattern is consistent across every deal I'm watching close in 2026: the companies getting paid are the ones who stopped saying "we will transform your industry" and started saying "we will move your data, your goods, or your compute from point A to point B, reliably, at a price you can budget for." That's not exciting. It's also the only pitch that survives a CFO's second meeting.

Microsoft Just Told You Where the Squeeze Is

Computer Sweden reported this week that Microsoft is killing volume discounts on cloud licensing. Read that again. The largest infrastructure provider in the world just tightened the screws on the exact layer everyone assumed was a commodity with permanently falling prices. That is not a footnote. That is a signal.

Infrastructure pricing power is consolidating at the top, right when demand for compute is exploding because every company on earth is trying to run some flavor of AI workload. If you are a mid-size Swedish company that assumed cloud costs would keep drifting down while your AI ambitions scaled up, you now have a budget problem you didn't have six months ago. This is exactly why the smart move for a lot of founders right now isn't "build bigger models," it's "build smarter automation on infrastructure you actually control the economics of." That's the entire premise behind why we push AI agents that run lean, not bloated model calls stacked on model calls burning your margin on every API round trip.

Where Custom AI Solutions Actually Pay Off

This is also why "custom AI solutions" beats off-the-shelf SaaS wrappers for companies with real infrastructure exposure. A generic AI product built on someone else's pricing assumptions breaks the moment the underlying provider changes terms, which is exactly what just happened. When we build custom AI solutions for clients, step one is always mapping the actual compute and data cost structure before writing a single agent. Nobody wants to hear that in a sales pitch. Everyone wants to hear it after their Azure bill doubles.

Sweden vs the World: We're Good at Pipes, Bad at Selling Them

Here's where I get a little proud and a little annoyed at the same time. Sweden is genuinely excellent at infrastructure. We have cheap, stable, green power in the north. We have engineering culture that respects reliability over hype. atNorth, Ericsson's network infrastructure legacy, our grid, our logistics through Nordic ports, this is real, exportable, defensible capability.

What we're bad at is narrating it. Silicon Valley will take a data center opening and turn it into a "sovereign AI infrastructure moonshot" press tour. We take a data center opening and put out a press release with the megawatt count and move on. From Jönköping, watching both sides, I'd say the American approach overhypes the product but wins the capital allocation game. The Swedish approach underhypes the product and loses deals it should win purely on quality.

Compare this to the EU-Latvia and Business Sweden-Peru moves this week. These are government-to-government infrastructure and trade plays, the kind of deal that never makes a viral LinkedIn post but moves actual GDP. Dagens Industri covers these deals in three paragraphs buried on page nine. In the US, a comparable trade infrastructure deal with a Latin American government gets covered as geopolitical strategy on the front page. Same substance, wildly different narrative weight. That mismatch is costing Sweden capital, because investors follow narrative before they follow substance, unfortunately.

Asia, meanwhile, is not debating the narrative at all. China and the Gulf states are pouring sovereign wealth directly into compute and energy infrastructure with the explicit goal of owning the layer everyone else has to rent. That's the real competitive threat to European AI ambition, not some vertical AI startup in California. If Sweden and the EU don't treat data center capacity, energy supply, and cross-border compute agreements as strategic infrastructure on par with ports and rail, we will spend the next five years renting our AI future from someone else's grid.

Are Regulators Even Looking at the Right Layer?

Short answer: not really. The EU AI Act is entirely focused on model behavior, output risk, and use-case classification. All of that matters. None of it addresses the actual bottleneck forming right now, which is physical: energy availability, data center permitting timelines, and cross-border data movement rules that were written for a pre-AI-scale internet.

Sweden's own energy permitting process for new data center capacity is still slow relative to demand, and every Nordic country is now competing for the same limited pool of green power that AI compute needs. Nobody in Brussels is writing policy for "who gets priority grid access when a hyperscaler and a hospital both need more power in the same region." That question is coming within two years, and right now the answer is whoever has the better lobbyist, not whoever has the better policy framework. That's a gap someone needs to close before it becomes a crisis instead of a headline.

Where This Goes: AGI Doesn't Care About Your Vertical, It Cares About Your Grid

Here's my honest forward view. As models get closer to genuine general capability, the bottleneck stops being "which industry gets disrupted first" and becomes purely physical: compute availability, energy cost, and data movement latency across borders. AGI, whenever it actually arrives in a meaningful sense, is not going to care whether it's running healthcare diagnostics or supply chain optimization. It's going to care about where the cheapest, most reliable megawatt is sitting, and which government made it easiest to move data and capital across a border to get to it.

That means the countries and companies that win the next five years are the ones treating infrastructure as strategy today, not the ones with the flashiest vertical AI demo this quarter. Sweden's green energy advantage and engineering reliability culture are genuinely strong cards to play here. But we need to play them as infrastructure diplomacy, the way Business Sweden is doing with Peru, not as isolated corporate press releases.

For builders, this means the AI agent development and blockchain development work that actually survives the next few years is the work that assumes compute costs will keep shifting, energy politics will keep tightening, and cross-border rules will keep changing. Build for that volatility, don't build assuming today's pricing and today's regulation stays fixed. That's why our blockchain development work increasingly focuses on cross-border settlement and tokenized infrastructure ownership models, because government-to-government trade deals need exactly that kind of trust layer, and it's a lot less sexy than another consumer NFT play, and a lot more durable.

What to Look At This Week

If you're building anything that touches this infrastructure squeeze, a few things on my radar right now that are worth your actual attention, not just a bookmark:

• rtk: a Rust CLI proxy cutting LLM token consumption 60-90% on common dev commands. If Microsoft is killing volume discounts, this is exactly the kind of unsexy efficiency tool that protects your margin. Single binary, zero dependencies, no excuse not to test it.
• Graphify: turns your codebase, docs, and schemas into a queryable knowledge graph with local deterministic parsing, no vector store dependency. Relevant because vector store costs are exactly the kind of infrastructure line item that quietly explodes at scale.
• OpenHands: AI-driven development tooling worth watching if you're scaling agent development in-house instead of renting every capability from a closed API. Owning more of your stack is the infrastructure mindset applied to your own codebase.

What to Actually Do About This

Stop pitching transformation. If you're a founder or a CTO reading this in Stockholm, Gothenburg, or here in Jönköping, ask yourself honestly: are you selling a vision, or are you selling a pipe? If you're selling a vision, your sales cycle is going to keep getting longer and your pilots are going to keep stalling, because every CFO in Europe has now sat through forty AI transformation pitches and learned to smell them coming.

If you want to sell a pipe, that means getting concrete. What data moves faster because of you. What compute costs less because of you. What trade relationship exists because of you that didn't exist before. That's a harder sell to write a keynote about, and a much easier sell to close a contract on.

Practically, this is exactly the kind of work we do at rapid MVP speed at HEIMLANDR when a client comes to us with an infrastructure problem disguised as an AI problem. Build the pipe first, prove the flow, then decide if you even need the vertical story on top. Most of the time, the client discovers the pipe was the product all along.

The next five years belong to whoever controls compute, energy, and the legal corridors to move both across borders. Sweden has real cards to play here. The question is whether we play them as narrative and diplomacy, or keep quietly building good infrastructure nobody outside our own industry ever hears about.

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.