Should Norway Just Buy OpenAI? The Absurd Idea That Won't Go Away

Every so often the internet floats an idea so obviously ridiculous that you dismiss it, and then it refuses to leave your head. This is one of them: Norway’s sovereign wealth fund should buy OpenAI. Outright. All of it.

The first reaction is a laugh. The second reaction, after you check the numbers, is less certain. Because the proposal isn’t really about Norway. It’s a Trojan horse for a much harder question — who should own frontier AI, and why have we all quietly agreed the answer is a handful of venture capitalists in Menlo Park?

The Money Is Not the Problem

Norway’s Government Pension Fund Global (GPFG) is the largest sovereign wealth fund on the planet, sitting on assets in the trillions of dollars. Norway has roughly 5.5 million people. Divide one by the other and every Norwegian citizen is notionally sitting on a few hundred thousand dollars of invested capital, funded by decades of North Sea oil revenue that the country decided — with a discipline most petrostates never manage — not to spend.

The fund already owns pieces of nearly every public company that matters. It routinely shows up on the shareholder registers of Apple, Microsoft, and Nvidia. So “Norway buys a stake in Big Tech” isn’t a provocation. That’s just Tuesday.

The provocation is scale. OpenAI’s valuation, after its most recent rounds, sits in the hundreds of billions. Against the fund’s total assets, that’s a single-digit percentage. A big bet, but not an impossible one. Pension funds make concentrated bets in that range and survive.

That’s the uncomfortable part. The reason Norway can’t buy OpenAI has nothing to do with whether Norway can afford OpenAI.

The Mandate Says No

Here’s where the idea actually dies, and it’s not where you’d expect.

GPFG cannot buy whatever it likes. It operates under a mandate set by Norway’s Ministry of Finance and ratified by parliament, and that mandate is deliberately restrictive.

First, the fund is built around listed equities. Private companies have long sat outside its permitted universe, with narrow carve-outs for real estate and renewable energy infrastructure. OpenAI is private. That alone ends the conversation under current rules.

Second, there are ownership caps. The fund cannot hold more than a set percentage of any single company. “Buying it outright” means 100%, which isn’t a rule violation so much as a category error. GPFG was designed from day one to be a passive minority shareholder spread across thousands of companies. It is not an operating company. It has no apparatus for running a frontier research lab, and building one would make it a fundamentally different institution.

Third — and this gets overlooked — OpenAI isn’t shaped like something you can buy. Its governance descends from a nonprofit controlling a capped-profit arm, and even after the restructuring fights, the question of who holds ultimate control has never had a clean answer. Microsoft’s stake and contractual rights are woven through the whole thing. There’s no single cap table you can write a check against.

So the blocker is rules and structure, not capital. And rules can be amended. Structures get restructured. That’s precisely why the argument keeps resurfacing instead of dying.

Norway Is Just a Stand-In

Nothing about this thought experiment requires Norway specifically. Norway is a convenient body double: a state actor with enormous capital, a functioning democracy, and no dog in the US–China fight. The real question being asked is what happens if frontier AI is owned by a sovereign entity that is neither Washington nor Beijing.

The case for it is straightforward. Frontier models are drifting toward public-utility status. Hundreds of millions of people use them daily for work, health questions, legal questions, homework, therapy-adjacent conversations. Yet the decisions about what these systems will and won’t do are made by a small group of executives and the investors who back them. If this were the power grid or the telecom backbone, we’d have regulated it two decades ago. A sovereign fund, the argument goes, operates on generational timelines and doesn’t answer to quarterly earnings — exactly the horizon you’d want for a technology this consequential.

The case against it is at least as strong. State ownership is not the same thing as public benefit. Handing a government direct control over the most persuasive text-generation system ever built is a surveillance and censorship risk, not a safeguard — and that’s true even for a well-governed democracy, because institutions outlast the people currently running them.

Then there’s the fiduciary problem, which is the one Norwegians would actually care about. GPFG is a pension. Concentrating a meaningful slice of national retirement savings into one unprofitable, unproven, litigation-exposed technology company is the exact inverse of the diversification principle the fund was built on.

And finally: talent. AI researchers currently choose between competing offers with eight-figure equity packages. Ask them to accept a civil-service compensation band and a decision process that runs through a finance ministry. Buy the company, watch the researchers leave, and you own a brand and some depreciating GPUs.

Why the Argument Keeps Coming Back

This proposal recurs because it sits on top of a real trend, not a fantasy one.

Governments are already spending seriously on sovereign AI — domestic data centers, national-language models, guaranteed GPU allocations. The EU is actively working to cut its dependence on American cloud providers. Gulf sovereign funds have moved aggressively into AI equity stakes and infrastructure buildouts, backing labs and compute at a scale that would have been unthinkable five years ago. Every one of them is answering the same question in their own dialect: is it acceptable to outsource your country’s cognitive infrastructure to a foreign private company?

The Norway proposal is just the maximalist version. Instead of buying 3% or building your own second-tier model, buy the leader outright. Its very impracticality is what makes the underlying choice legible: do we want ownership of AI companies, or do we want a say in how they behave?

The honest answer is the second one. And influence doesn’t require a cap table. It requires regulation, procurement leverage, standards-setting, and shareholder engagement — the last of which GPFG already does professionally. The fund has spent years voting its shares and publishing expectations on climate disclosure and executive pay, and companies have changed policy in response. You don’t need 100% to have a voice. Norway figured that out a long time ago.

What’s Left

Norway will not buy OpenAI. The mandate forbids it, the corporate structure won’t accommodate it, and no finance minister could defend it to a pensioner.

What survives is the question underneath. Who should decide the direction of a technology that hundreds of millions of people now touch every day? Right now the answer is: whoever led the last funding round. Maybe that’s fine. Maybe markets allocate this better than ministries would — they usually do.

But it’s worth noticing that we never actually chose that arrangement. It just happened, fast, while everyone was busy arguing about benchmarks. And the window for choosing something else is narrowing.