Bernie Sanders and Donald Trump agree on something. That alone should make UK boards pay attention, but the substance matters more: both now argue the US government should take equity stakes in the big AI companies so the public shares in the rewards. Reuters columnist Mike Dolan’s analysis of this convergence lands on a conclusion with direct consequences for anyone building on OpenAI or Anthropic — turning AI firms into quasi-utilities “may no longer be a pipe dream”. If the companies at the top of your AI stack end up part-owned by Washington, the terms of your supply chain change, and the UK’s own sovereignty push starts to look less like ambition and more like necessity.
Nationalisation is back in fashion
The immediate trigger was a New York Times op-ed in which Sanders proposed the US government take 50% stakes in the big AI companies, held through a new sovereign wealth fund, with the returns directed to the workers most affected by automation. His argument is that AI “is built on our collective intelligence: our books, songs, artwork, journalism, computer code, scientific research, videos, conversations, images and ideas spanning generations” — so the public already paid for the raw material.
What stopped this being a firebrand’s trial balloon is that the White House appears to agree. Digital outlet NOTUS reported that senior US officials held preliminary talks with major AI companies, including OpenAI’s Sam Altman, about firms voluntarily ceding shares to the government, with returns possibly paid out as dividends to American households. By the Friday of that week, Trump told reporters the idea “almost becomes a partnership with the American public” and that his team would look into it.
Nor is this hypothetical as policy. The same administration took a 10% stake in Intel last year, followed by stakes in rare-earth firms and, only last month, quantum computing companies including IBM. Cornell professor Nicholas Mulder argues we are in the fourth wave of nationalisation in a century, with governments taking half a trillion dollars in assets worldwide since 2020 — the first surge of its kind since the 1970s.
| Number | What it is | Why it matters |
|---|---|---|
| 50% | Public stake Sanders proposes in big AI companies | Sets the ambitious end of the debate; legislation is planned |
| 10% | US government stake in Intel, taken August 2025 | The template already exists and has been repeated |
| $10bn → $50bn | Value of Washington’s Intel stake after the shares quintupled | Success will feed political appetite for more stakes |
| $500bn | Assets nationalised globally since 2020 (Mulder) | This is a wave, not a one-off — and it predates the AI debate |
| 2 | Frontier labs (OpenAI, Anthropic) heading to public markets | IPOs create the exact moment a government can buy in |
Strategic Reality: The two companies most likely to receive a US government shareholder are the two companies most UK AI strategies depend on. This debate is not American domestic politics; it is a possible change of ownership at the top of your supply chain.
From regulating AI to owning it
Dolan’s sharpest observation is about motive. Governments are not considering AI stakes because socialism is back; they are doing it because AI and quantum computing are increasingly treated as critical national resources, in a post-pandemic world of supply chain anxiety and security rivalry. The logic that nationalised steel and energy in earlier waves now points at compute and models.
The economics reinforce the politics. Former IMF chief economist Ken Rogoff warned that governments failing to secure a place in the AI supply chain may face mass job displacement without the tax revenues or state capacity to contain the fallout. “No one really knows what such a world would look like, let alone how to keep it from tearing itself apart,” he wrote. Fear of that world, Dolan argues, may be enough to push governments from regulating AI to owning a piece of it.
The open question is what kind of shareholder a government makes. Dolan notes the split view: a state stake can mark a company as too big to fail and underwrite its success — Intel’s shares have quintupled since Washington bought in — or it can deter private capital, politicise governance and expose taxpayers to losses. There is also the question of appetite. Trump has already said he “should have asked for more” from Intel. Minority stakes have a way of growing.
Critical Context: OpenAI and Anthropic are both preparing public share sales. An IPO is precisely the moment a government can acquire a stake without expropriation — which means the window in which this debate becomes reality is the next few quarters, not the next decade.
The UK angle: sovereignty rhetoric meets someone else’s utility
Here is where the op-ed collides with the week’s UK news. Rachel Reeves used her Mansion House speech to demand a “serious plan on AI sovereignty”, calling AI the defining technology of our generation and committing the UK to the first G7 Digital Sovereign Bond by early next year. The UK is spending real money on sovereign compute and digital infrastructure.
But there is an uncomfortable asymmetry in what that money buys. UK sovereignty spending purchases infrastructure: data centres, compute capacity, connectivity. The models that run on that infrastructure remain the property of two Californian companies that may soon count the US government among their shareholders. Owning the pipes is not the same as owning the water. If Washington takes stakes in the frontier labs, the UK’s position becomes that of a country whose critical AI capability is supplied by the quasi-utilities of another state — a familiar position for anyone who remembers European dependence on foreign-owned energy, and not a comfortable one.
| Stakeholder | What US stakes in AI firms would change | So what |
|---|---|---|
| UK firms buying AI tooling | A political shareholder enters the vendor relationship | Model access and pricing gain a diplomatic dimension; watch for public-interest conditions |
| Firms building on frontier APIs | Vendor governance answers partly to Washington | Portability across providers, including open-weight models, becomes a sovereignty measure, not just a cost one |
| UK public sector buyers | Procuring from part-state-owned foreign suppliers | Expect harder questions from ministers and auditors about dependency |
| Policymakers | Sovereign compute without sovereign models | Pressure grows for UK stakes, golden shares or capability guarantees of its own |
Hidden Cost: A government shareholder is a silent counterparty in every contract you sign with its company. Export controls on advanced chips already show how quickly commercial access can become a foreign-policy instrument. Equity stakes deepen that lever — and no UK contract clause can negotiate it away.
What UK leaders should do about it
None of this requires panic, but it does reward positioning. The right response scales with how load-bearing AI is for your organisation.
If you are early in adoption, treat this as one more reason to avoid deep exclusivity. Both frontier labs are competing hard for enterprise business ahead of their listings, so the leverage sits with you; use it to keep your prompts, evaluation sets and integration code portable.
If you are established on one vendor, add political ownership to your vendor risk register alongside financial health and roadmap churn. Ask your supplier directly what a government shareholding would mean for non-US customers’ access and terms. The quality of the answer is itself information.
If AI is critical to your operations, plan for the scenario where frontier models become instruments of national strategy. That means a tested fallback on a second provider, a considered position on open-weight models for workloads that tolerate them, and board-level awareness that your core capability may soon sit inside companies with a state on the share register.
Implementation Note: The cheapest hedge available today is architectural, not contractual. An abstraction layer that lets you route workloads across two providers costs weeks of engineering now and buys options against ownership outcomes nobody can predict.
The challenges nobody puts on the slide
Four second-order problems deserve attention before this debate resolves.
First, utility status cuts both ways for buyers. Utilities are stable, regulated and slow. If frontier labs become quasi-utilities, the relentless feature velocity UK firms currently enjoy may ease. That is good for governance teams drowning in model updates, and bad for anyone whose competitive plan assumes the current pace continues. Build plans that survive both tempos.
Second, partial nationalisation can crowd out the challengers. A state-backed incumbent raises the barrier for new entrants, and today’s healthy duopoly competition could calcify into an administered market. The mitigation is to keep genuine optionality alive in your own stack, including smaller and open providers, rather than assuming the market will do it for you.
Third, the UK could be pushed into reactive spending. Dolan warns that beyond America, the push for state stakes in tech firms “may turn panicky”, forcing catch-up investment in regional ecosystems. A UK response driven by fear of exclusion rather than a clear capability plan would spend sovereignty money badly. Firms should judge UK sovereign-compute announcements by what they secure access to, not by their size.
Fourth, dividends politicise dependency. If American households receive AI dividends, every dollar of international revenue those companies earn becomes a matter of US domestic politics. UK customers would be contributing to a foreign public’s payout, a framing that will eventually occur to a UK politician too, with unpredictable consequences for how AI procurement is debated at home.
The takeaway
The serious point beneath the striking headline is that ownership of the AI stack is now a live political question in the country that builds most of it. Sanders wants half; Trump wants a partnership; the precedents — Intel, rare earths, quantum — already exist. Whether or not Washington ends up on the frontier labs’ share registers, the direction of travel is clear: AI is being reclassified from private product to strategic resource.
UK organisations should do three things this quarter. Add political ownership of your AI vendors to the risk register, in writing, so the scenario has an owner. Make portability real enough that a change in vendor governance is an inconvenience rather than a crisis. And read the UK’s sovereignty announcements — including whatever follows Reeves’ Mansion House commitments — with one question in mind: does this buy Britain control of any layer of the stack, or just better plumbing beneath someone else’s utility?
Take Action: Write down which layer of your AI stack you actually control — models, compute, integration, data. For most UK firms the honest answer is the bottom two. That gap between what you depend on and what you control is the strategic exposure this debate is really about, and mapping it costs nothing this week.
Source and attribution
This analysis draws on “Will AI become a public utility?” by Mike Dolan, Reuters Editor-at-Large for Finance and Markets, published by Reuters Open Interest on 9 June 2026. Original commentary available at reuters.com. The opinions in the source are the author’s own.
Editorial analysis and UK business framing by Resultsense. We make sense of AI in the UK — turning research, policy and announcements into what they mean for the people building and buying these systems. For more analysis, explore our insights or get in touch.