Boden Isn't a Data Center Town Anymore. It's a Sovereignty Test

Three companies with three completely different reasons to exist just landed in the same 40 kilometer stretch of Norrland, fighting for the same grid connections. That is not a data center boom. That is a sovereignty test, and most of the people writing headlines about "gigawatts landing in the north" have no idea they are watching one.

Let me lay out what actually happened this week, because the details matter more than the press releases.

Three Buyers, One Grid, Very Different Incentives

HIVE Digital Technologies, a Canadian bitcoin miner that has been quietly reinventing itself as an AI infrastructure company, just signed a 10-year lease with a Swedish sovereign client for the Boden AI data center. That deal sits on top of a $220M AI infrastructure raise and a Swedish data center acquisition HIVE made earlier this year. Read that again: a Canadian crypto miner is now the landlord for Swedish sovereign AI compute. That sentence should make every policymaker in Stockholm sit up straight.

At the same time, Google is breaking ground in Horndal, about 250 km south, and attaching a EUR 5M skills fund to sweeten the deal for local politicians and unions (SVT covered the groundbreaking, and it reads like every other hyperscaler PR playbook from the last decade). And CoreWeave, the American AI cloud that has become Nvidia's favorite GPU landlord, just partnered with Conapto to expand AI cloud capacity running on Swedish renewable power.

Three deals. Three capital structures. One resource: cheap hydro, cold air, and grid capacity that Svenska kraftnät cannot expand fast enough to satisfy everyone asking.

Here is the part nobody wants to say out loud: only one of those three buyers is actually accountable to Swedish law in a way that matters for sovereignty. The sovereign client leasing from HIVE has a real obligation structure. Google and CoreWeave do not, not in the way that counts if a foreign government ever leans on them, or if US export control law decides Swedish workloads running on American-owned racks are suddenly a national security concern for Washington, not Stockholm.

What "EU Cloud Sovereignty" Actually Means, and Why Boden Breaks the Definition

Everyone in Brussels loves the phrase "EU cloud sovereignty." Almost nobody can tell you what it means operationally. Does it mean the servers are physically in the EU? Does it mean EU law governs the data? Does it mean an EU entity owns the infrastructure? Does it mean the parent company answers to an EU regulator, or does a US CLOUD Act subpoena still reach through a European subsidiary? Boden is where that fuzzy definition gets tested in public, because you now have three different answers sitting on the same power grid:

The Physical Sovereignty Illusion

Google's Horndal site and CoreWeave's Conapto expansion are physically in Sweden. Electrons flow from Swedish hydro. That satisfies the loosest definition of sovereignty, the one that shows up on marketing slides: "our EU data centers." It does nothing for legal sovereignty. A US parent company operating a Swedish subsidiary is still reachable by US law when Washington decides it wants to be reached. Ask any GDPR lawyer who has read the Schrems II ruling twice.

The Sovereign Client Model

The HIVE deal is more interesting precisely because HIVE is not a hyperscaler. It is a Canadian miner that pivoted into infrastructure and is now, functionally, a neutral landlord leasing capacity to a Swedish sovereign client on a 10-year term. That structure looks a lot more like sovereignty in practice: Swedish entity, Swedish jurisdiction, Swedish accountability, foreign capital just providing the box and the power contract. It is not perfect. HIVE is still a foreign, publicly traded company whose board answers to Toronto and Nasdaq, not Stockholm. But the client relationship is where the actual control lives, and that is the model regulators should be studying, not celebrating gigawatt announcements as if megawatts equal sovereignty.

Sweden vs. The World: We Are Winning the Wrong Race

From where I sit in Jönköping, the Nordic pitch to global capital has been genuinely excellent for a decade. Cheap renewable power, political stability, cold climate that cuts cooling costs, predictable grid regulation. Ireland burned its own goodwill chasing data center growth without enough grid investment and is now rationing new connections around Dublin (RTE has covered the moratorium extensively). Sweden looked smarter by comparison. But "we have cheap power and stable politics" is a location pitch, not a sovereignty policy. The US does not have this problem because American hyperscalers building American data centers under American law is definitionally sovereign for the US. China does not have this problem because nobody foreign is allowed to own critical compute infrastructure there in the first place. Europe, and Sweden specifically, is the only major bloc trying to attract foreign capital into critical AI infrastructure while also claiming the result will be sovereign. That is a much harder trick than Brussels admits. The EU Chips Act and the Digital Markets Act both gesture at this problem. Neither one actually answers the ownership question when the datacenter operator is Canadian, the GPUs are Nvidia's, the cloud software stack is American, and the only genuinely Swedish party in the chain is the client leasing capacity and the electrons underneath it.

Is Swedish and EU Policy Ready? No.

Svenska kraftnät is approving grid connections roughly in the order requests arrive, with some weight given to regional balance, but there is no mechanism I have seen that prioritizes sovereign or EU-controlled workloads over foreign hyperscaler expansion when both are competing for the same substation capacity. Breakit has reported on the queue length for grid connections in Norrland multiple times this year, and it keeps getting longer, not shorter. That means the substation fight is being decided by who submitted paperwork first and who has the deepest pockets to pay for grid upgrades, not by any strategic assessment of what Sweden actually wants running on its power. A Canadian miner turned AI landlord and an American hyperscaler are, right now, on equal legal footing to a Swedish sovereign AI project when it comes to accessing the grid. That is not a policy failure that will resolve itself. It requires someone in Stockholm or Brussels to actually decide that sovereignty means something before the queue clears, not after. The EU AI Act sets rules for how models behave. It says almost nothing about who is allowed to own the compute those models run on. That gap is where this entire fight lives, and right now the gap is empty.

Where This Goes: 2029 and the Compute Ownership Reckoning

Follow the trajectory forward. Within 2 to 3 years, AI compute demand does not plateau, it compounds, because every serious step toward more capable, more agentic AI systems needs more inference capacity sitting closer to users and data. The gigawatts landing in Norrland today are a rounding error compared to what gets requested by 2029. At that scale, "who owns the substation" stops being an infrastructure question and becomes identical to "who controls the AGI pipeline." If your national AI capacity sits on foreign-owned racks, running foreign cloud software stacks, subject to foreign export control law, your sovereignty claim is fiction the moment a geopolitical crisis makes a foreign government decide your workloads are a problem. This already happened in miniature with chip export controls. It will happen again with compute access. Sweden and the EU have maybe 3 to 5 years to build a real answer: a legal and grid-allocation framework that actually distinguishes sovereign-controlled compute from foreign-leased compute, and that gives the former real priority when substations run out. Right now that framework does not exist. HIVE's Boden deal is a preview of what a good answer looks like operationally, a foreign capital provider, a domestic accountable client. But it happened because of deal structure and negotiation, not because Swedish policy demanded it. That is luck, not strategy.

What to Look At

If you are building anything that touches this infrastructure question, whether you are running workloads in a Swedish data center or trying to understand what "sovereign" means for your own stack, here is where I would spend time this month:

• awesome-selfhosted — the single best argument for why sovereignty starts with what you can run yourself, not what you rent from someone else's cloud, foreign or domestic.
• Kubernetes — if your workloads are portable across clouds and physical locations, you have optionality when a substation fight or a jurisdiction fight goes the wrong way. Vendor lock-in is a sovereignty risk, full stop.
• Netdata — you cannot make sovereignty claims about infrastructure you cannot observe. Full stack visibility into where your workloads actually run and how they perform matters more now than it did a year ago.
• Ansible — automate your deployment so moving between providers, or between sovereign and non-sovereign infrastructure, is a decision you can act on fast, not a six-month migration project.

What You Should Actually Do

If you run a company that touches sensitive data, government contracts, or anything that could plausibly be called critical infrastructure, stop treating "we use a EU data center" as a sovereignty answer. Ask who owns the entity operating that data center, what jurisdiction the parent company answers to, and what happens to your data if a foreign court orders disclosure. Most companies I talk to in Sweden have not asked that question once. This is exactly the kind of architecture decision we work through with clients at HEIMLANDR. When we build [custom SaaS](https://heimlandr.io/services/saas-development) or [AI solutions](https://heimlandr.io/services/ai-solutions) for Nordic clients who care about data residency and legal exposure, the infrastructure layer is not an afterthought, it is the first conversation. And if your business logic genuinely needs to be enforceable and auditable regardless of who owns the underlying cloud, [smart contract development](https://heimlandr.io/services/smart-contracts) gives you a layer of accountability that does not depend on trusting a foreign parent company's compliance department. Software development in Sweden has an advantage right now that most founders are not using: we can build sovereignty into the architecture instead of hoping a data center lease gives it to us for free.

The Real Headline

Nobody is going to write "Substation capacity allocation policy needs urgent reform" as a headline, because it is boring and it does not photograph well next to a groundbreaking ceremony with a EUR 5M skills fund attached. But that is the actual story happening in Boden and Horndal right now. Three very different companies are fighting for the same finite grid connections, and the winner of that fight, not the press release, decides whether Sweden's AI infrastructure serves Swedish interests or just happens to be located here. Gigawatts are not sovereignty. Ownership structures are sovereignty. Sweden needs to start writing policy that understands the difference before the queue for Norrland's substations closes and the decision gets made by default.

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.