Sweden Outsources Its Digital Spine, Funds Data It Can't Ship

The Contradiction Nobody in the Boardroom Wants to Name

This week, HCLTech renewed and expanded its digital transformation deal with a Swedish commercial vehicle giant. The same week, Vinnova announced backing for MASSIV+, an industry-academia collaboration for "actionable climate data." And at Karolinska, a trustworthy AI event is framing deployment and adoption as still unsolved problems. In 2026.

Let me spell out what this means. Sweden's largest industrial companies are simultaneously outsourcing their core digital infrastructure to offshore IT service providers and writing checks for academic AI projects that produce papers, not products. BCG is now telling Nordic companies that transformation is a "must." But nobody is asking the obvious question: who actually owns the transformation stack when the capability sits in Noida, not Norrköping?

I run HEIMLANDR.IO from Jönköping. We build AI automation for business, custom AI solutions, and ship software. I am not a neutral observer. But I am a builder, and builders see things that consultants miss.

The Outsourcing Pattern That Eats Competitive Advantage

Let me describe what the HCLTech deal actually looks like from the inside, because I have seen this pattern in Swedish industry for 15 years.

A large Swedish manufacturer decides it needs to "digitally transform." The board approves a budget. The CIO, who has 40 people internally and needs 400, signs a multi-year managed services agreement with an Indian IT giant. HCL, TCS, Infosys, Wipro. Pick one. The deal covers everything from ERP integration to cloud migration to IoT platforms.

Year one: things move. Dashboards appear. Executives get slide decks showing KPIs. Year two: customizations pile up. The offshore team grows. Internal knowledge shrinks. Year three: the Swedish company cannot make a meaningful technical decision without consulting the service provider. The IP lives in someone else's codebase. The data flows through someone else's architecture. The people who understand the system sit 7,000 kilometers away.

This is not outsourcing. This is a slow transfer of your digital nervous system to a third party who bills by the hour.

I am not blaming HCLTech. They are doing exactly what they are paid to do, and they do it well. I am blaming Swedish boards who treat software as a cost center instead of a strategic asset. When your core differentiation is engineering, and you hand the digital layer of that engineering to someone else, you are hollowing yourself out.

The Academic AI Trap

On the other side of the same boardroom table sits the innovation budget. Swedish companies love to fund research. Vinnova facilitates it. Universities participate. The MASSIV+ project for climate data is a perfect example. It aims to deliver "actionable" climate data through industry-academia collaboration.

I have no doubt the researchers involved are talented. I have no doubt the data will be interesting. What I doubt is that any of it will be deployed at scale in a production environment within three years.

Here is why. Academic AI projects optimize for novelty and publication. Commercial deployment requires optimization for reliability, integration, maintenance, and user adoption. These are fundamentally different objectives. Karolinska's own event this week is essentially admitting that trustworthy AI deployment in healthcare is still an unsolved problem. This is Sweden's premier medical institution. If they have not cracked deployment, what chance does a Vinnova-funded climate data consortium have?

The pattern is: fund research that produces a proof of concept, publish results, present at a conference, and then watch the prototype rot on a server somewhere because nobody budgeted for the unglamorous work of turning it into a product.

Meanwhile, American and Chinese companies are shipping AI into production every single day. Not better AI. Shipped AI. There is a massive difference.

Sweden vs. The World: Where We Stand

Let me put this in global context.

In the US, companies build internally. Big tech hoards engineering talent. Even mid-market companies have 50-person engineering teams building proprietary systems. The cultural default is: if it matters, we build it ourselves. Yes, they use contractors and offshore teams too. But the architecture, the core logic, the data, those stay in-house.

In China, the state pushes domestic capability. AI automation for business is a national priority with commercial deployment targets, not paper targets.

In India, the IT services giants are the capability. They are not just body shops anymore. HCL, TCS, and Infosys are building their own AI platforms, their own IP, using the knowledge they extract from client engagements. Every Swedish manufacturer that outsources its transformation is training its future competitor's AI systems.

In Sweden, we have the worst of both worlds. We outsource the building to people who do not share our strategic context, and we fund research that does not reach the market. We have phenomenal engineering culture, incredible domain expertise in manufacturing, energy, transport, and medtech. And we are systematically failing to convert that into digital capability we own.

The EU AI Act adds a layer of regulatory complexity that, frankly, makes it even harder for small and mid-sized Swedish companies to deploy AI. The compliance burden falls disproportionately on European builders while American and Chinese companies move fast and sort out regulations later. I am not saying we should abandon responsible AI. I am saying the regulatory framework as it exists today protects incumbents and punishes builders. That is the opposite of what Sweden needs.

What Actually Needs to Happen

I am going to be direct.

Swedish industrial companies need to stop treating software development as something you buy by the kilogram from the cheapest provider. Custom AI solutions are not a commodity. AI agent development is not something you outsource to a team that does not understand your factory floor, your logistics chain, your regulatory environment.

This does not mean you need 500 engineers in-house. It means you need a core team that owns the architecture, owns the data, owns the decision-making about what gets built and why. You can work with partners. You should work with partners. But the partners should be people who build with you, not for you. People close enough to understand your context. People who pick up the phone in your timezone.

At HEIMLANDR, we work exactly this way. Our AI agent development is built around understanding the client's actual operational reality, not delivering a generic platform with a Swedish-language skin on top. Our SaaS development starts with the business problem, not the technology stack.

The Vinnova model needs fixing too. Fund deployment, not just research. Require commercial milestones. Put builders in the room from day one, not as an afterthought in year three when the grant period is ending and someone says "we should think about commercialization." The gap between Swedish academic AI and deployed AI is not a technology gap. It is an incentive gap.

Where This Goes: 2027-2030

Here is what I see coming, and it is not comfortable.

AI-driven development tools are accelerating at a pace that will reshape who can build software and how fast. Projects like OpenHands and Agno are making it possible for smaller teams to build and manage agentic systems at scale. The everything-claude-code movement, now one of the hottest repos on GitHub with 165,000+ stars, is optimizing how developers work with AI coding agents. The cost of building software is dropping. Fast.

This means the argument for massive offshore teams gets weaker every quarter. If a five-person team with AI agents can do what a 50-person offshore team does today, the economics of outsourcing flip entirely. The companies that invested in understanding their own digital architecture will be able to move at unprecedented speed. The companies that outsourced everything will be stuck renegotiating contracts with service providers who have no incentive to make themselves obsolete.

On the AGI trajectory: we are not there yet, but the path is visible. When general-purpose AI systems can handle complex multi-step engineering tasks, the competitive advantage shifts entirely to domain knowledge, proprietary data, and the ability to direct AI systems toward the right problems. If your domain knowledge lives in an offshore provider's documentation system and your data sits in their managed cloud, you have nothing left when the AI capabilities become commoditized.

Swedish regulators are not ready for this. The Swedish government still talks about AI in terms of "national strategy" documents that read like they were written in 2021. EU policy is focused on risk classification and compliance frameworks. Nobody is talking about industrial sovereignty in practical terms. Who owns the digital transformation stack of Sweden's largest exporters? That should be a question for the Riksdag, not just a procurement decision in a Gothenburg office park.

Blockchain development adds another dimension here. Supply chain transparency, carbon credit verification, provenance tracking. These are real problems for Swedish manufacturers, and blockchain solutions can solve them. But only if the companies building them understand the industrial context. Tokenizing climate data from MASSIV+ on a chain that actually integrates with SAP and production systems is hard, specific work. It is not a hackathon project.

What to Look At

If you are a CTO or founder processing this, here are specific things worth your attention right now:

Agno (39,600+ stars). An open-source framework for building, running, and managing agentic software at scale. If you are evaluating whether a small internal team can replace a large outsourced one for AI workloads, start here. This is what the future of lean engineering teams looks like.

Daytona (72,000+ stars). Secure, elastic infrastructure for running AI-generated code. As AI writes more of your codebase, you need a safe environment to execute and test it. Daytona solves that problem. Relevant for anyone running AI agents in production.

Hoppscotch (79,000+ stars). Open-source API development ecosystem. If you are pulling your API development workflow away from proprietary tools and into something your team controls, this is the best alternative to Postman right now. On-prem option matters for companies with data sovereignty requirements.

The BCG Nordic transformation report. Read it critically. It correctly identifies that transformation is mandatory. It less clearly identifies who should own the resulting systems. Fill in the blanks yourself.

The Real Question

I sit in Jönköping and I watch Swedish industry make the same mistake over and over. We have world-class manufacturing. World-class engineering. Reasonably good education systems. A culture that values precision and quality. And we are handing the digital expression of all of that to companies whose incentive is to keep us dependent.

The contradiction is not complicated. You cannot fund "actionable climate data" and simultaneously outsource the entire infrastructure that would make climate data actionable. You cannot talk about AI sovereignty and then sign a five-year managed services deal that puts your core systems in someone else's hands.

Someone in the boardroom needs to say this out loud. If your board will not, forward them this post. The future belongs to companies that own their digital backbone. Not the ones that rent it.

Build things. Own things. That is the whole strategy.

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.