It's been a summer of discontent for America's hyperscalers. Despite marshalling over £1 trillion of new investment, opposition to firms building data centres has snowballed across the country and across the political spectrum, as politicians from Senator Bernie Sanders to Texas Republican Governor Greg Abbott have called for a halt to new builds.
It's a story all too familiar to students of Britain's Industrial Revolution. Before the famous loom breakers of the Luddite movement, there were the Southwark millers who, fearing the ruin of their trade, were thought by many to have set ablaze the first steam-powered mill in London. The burned-out hulk of the Albion Mill became a feature of the London skyline for years, and may have inspired a local artist, William Blake to pen a poem about its ‘dark satanic’ features.
Yet a U.S.-style tech-lash against AI is the last thing Britain can afford. The country is already struggling to close a huge productivity gap with the United States, which the IMF estimates has roughly doubled since the early 2000s, accelerating rapidly after 2010. After the financial crisis, US productivity growth remained around 1 per cent a year, while the UK’s fell to about 0.5 per cent.
This gap, rooted in a failure of UK firms to invest, matters for the simple reason that we cannot fix today’s cost of living crisis without improving wage growth, and we cannot sustainably increase wages faster unless our productivity growth improves.
AI has immense promise to help fix this problem. This summer, the IMF showcased research (Misch and others, 2025) that estimates that the UK’s Total Factor Productivity gains could range from 0.1 to 0.5 per cent per year in the medium term which is ‘above the average of European countries and on par with estimated US gains.’
Yet America’s battle of the data centres is a warning because voters are asking a simple question: What's in it for me? As US think tank Third Way has shown, voters fear the same old story of oligarch fortunes built on workers' losses. And that is simply no longer a bargain people are prepared to accept. As Third Way’s analysts summarised;
America’s opposition to data centers has less to do with their feelings about artificial intelligence and more to do with their anger and distrust of large corporations and government.
The British public fears something similar. Polling in May from the Policy Institute, King's College shows:
69 per cent of the public are worried about the economic consequences of AI-related job losses.
65 per cent of the public and 65 per cent of workers think the economic benefits will mainly go to wealthy investors and large companies.
Just 7 per cent of the public think the benefits will be shared fairly across society.
51 per cent expect AI to increase inequality and social polarisation.
This is a huge potential problem for Britain, because from an economic point of view we need AI to diffuse faster through today's economy. Evidence to our Business and Trade Committee inquiry showed that only a quarter of businesses were using AI in late 2025, rising to 44 per cent among firms with more than 250 employees.
Going forward the speed of future AI diffusion will to a large degree depend on workers embracing it, tinkering with it, playing with it, innovating with it. If we all collectively slow walk the adoption of new technology, then as a nation, we simply will not harvest the prizes that are potentially available.
How do we fix the trust gap? By creating a new ‘Hyper-scalers Social Contract’. At its heart must be a new way to share the wealth that is so widely anticipated. Markets anticipate this prize to be gigantic. As we can see from the valuation of tech stocks, investors are betting that the AI economy will generate fortunes worth hundreds of billions of pounds.
My suggestion is that we expand our UK Sovereign AI Fund by creating a National AI Dividend Fund: a permanent public wealth fund that captures a share of the upside when public capital, infrastructure, data and other scarce assets help create private value.
How would this work?
The UK is already the world's third largest AI market and there is a revolution in the build-out of data centres already underway. Britain already has at least 500 data centres but Ofgem said in July that another 315 data-centre projects had applied for grid connections. Some applications are likely to be speculative or duplicated, but Barbour ABI has identified 183 more projects either in the planning system or already under construction. And some of these facilities are huge: the proposed Elsham Tech Park near Scunthorpe might cover 1.76 million square metres, roughly 70 per cent of the area of Hyde Park.
Public policy can, however, help this AI revolution go faster in at least seven ways:
Power and faster grid connections. All of our evidence shows that the lack of electricity and grid connections are a fundamental constraint on building the nation’s AI infrastructure. AI Growth Zones, it is said, will offer ‘enhanced access’ to power and planning support to accelerate investment. But when firms like Microsoft are being offered grid dates of 2033, we know there’s a problem - and a value in solving it. Kao Data told us it was waiting up to 15 years for grid slots.
Faster, more predictable planning. Again, we’ve had lots of evidence about how planning problems create delays. Public policy can help with this. Firms face not only planning delays but overlapping jurisdictions of councils and mayoral authorities that confuse not simplify the process of getting stuff built.
Government procurement. This is big. I’ve heard time and time again how a government purchase order can provide revenue, validation and thereby help firms access finance. But right now, the British government is, as one witness told us, ‘addicted to American software.’ Boardwave told us the government spends upward of £20 billion a year on technology of which 90% is spent on American technology. The private sector tech spend - which is perhaps £110 billion - is only 60-70% American tech.
Public compute, research and innovation funding. Big investment in public compute, research and innovation funding helps create an economy in which AI flourishes faster - and helps build the labour pool of requisite skills. UKRI grants and R&D support - all part of the state-backed innovation system - help build the demand side for AI
Patient capital and state-backed investment from the British Business Bank helps crowd-in private capital into high-growth firms which take AI into the economy faster and on a larger scale
Public data as infrastructure. This is one of the biggest assets we have. It has to be used incredibly carefully, but we have incredible data assets in this country, from the NHS to Ordnance Survey, that, carefully used, could create tremendous value.
Standards and assurance. As recent stories this summer highlighted, this is going to become more and more important. But trust and assurance created from the AI Standards Institute creates private sector value in ways that is going to become more and more valuable. But right now, it is basically free to industry.
These are very significant assets but, deal by deal, the government risks leaving very significant value on the table for the future. So, how do we ensure that workers ‘own a piece’ of the machine’ which the state is helping build for the future.
Owning a Piece of the Machine
There’s an old, possibly apocryphal, story that when the famous British scientist Michael Faraday first demonstrated the importance of his electrical discoveries to William Gladstone, the great knight’s only comment was: ‘But, after all, what use is it?’ Quick as lightning, Faraday replied:
‘Why, Sir, there is every probability that you will soon be able to tax it!’
All new technological advances bring a debate about how to share the wealth and today is no different. Fiscally, the advance of AI will require us to fundamentally rethink the balance of tax between capital, labour and consumption (see this from Brookings). That is a big task. But in the meantime there are five core ideas on the table that, in their different ways, heed the warnings of Pope Leo XIV, who counselled that we must avoid:
‘the idolatry of profit that sacrifices the weak’
And remember that
‘among the goods that are universally intended for everyone (my italics), we must also include new forms of property, such as patents, algorithms, digital platforms, technological infrastructure and data.’
How should Britain capture its share?
There is no single obvious “AI tax” but the international debate has now generated an excellent range of options. I think they fall into roughly five broad strategies:
Tax the machine. First, we have taxes on AI use. New York State Representative Alex Bores has proposed a small tax on AI tokens, as has Bill Gates and the UK’s IPPR has floated something similar. The attraction is obvious: tax the activity that is growing. But a tax on use risks slowing good-for-productivity diffusion and as tokens are a proprietary thing, they might prove tricky to get at.
Tax displacement. A second approach would require firms to contribute when AI substitutes for labour. IPPR has proposed a worker-support levy to finance retraining and transition support. Dario Amodei argues that if AI produces large and permanent labour displacement, we might need ‘taxes on relevant companies or raising the capital gains tax.’ Defining exactly when technology has “replaced” a worker will of course not be easy, but these ideas are part of a broader debate for rebalancing the tax system between workers and machines, discussed by both Brookings and Bill Gates. Silicon Valley’s very own Congressman Ro Khanna has argued for ‘taxing agentic AI more than we tax human workers’ alongside measures to encourage worker equity participation.
Tax the rents, not the investment. A third option is to strike when exceptional profits arrive and tax the economic rents that emerge. IPPR calls this an “AI unearned rents tax.” It has a logic. Taxes on ordinary returns discourage investment. But windfall taxes are a tried and tested method for sharing windfall rents. If AI really does generate monopoly-like fortunes, rent taxation is bound to have a role.
Take equity, warrants or other claims on future profits. This is the most direct way of turning an AI boom into public wealth. The notion of sovereign wealth funds to help cushion the transition of automation has been around since at least 2017 when Sam Altman proposed American Equity. Today, Bernie Sanders is anchoring the radical end of this debate with draft legislation to require the largest US AI firms to pay a one-off 50 per cent tax in shares to a sovereign wealth fund. Senator Sanders’s American AI Sovereign Wealth Fund Act reflects ideas proposed by Jeremy Bearer-Friend and Sarah Polcz who independently proposed ‘a one-time tax payment by generative AI firms in the form of equity’. Representative Bores proposes a less intrusive alternative: ‘The federal government would take stakes in major AI companies structured as out-of-the-money warrants - the right to buy shares at a price that only pays off if these companies are wildly successful at replacing human labor’. OpenAI itself has also called for a Public Wealth Fund, but on the lines of the UK’s current Sovereign AI Fund, investing on behalf of the nation, ‘in both AI companies and the broader set of firms adopting and deploying AI.’
Charge properly for the public assets that make AI possible. For Britain, this may be the most immediately practical strategy. Canada’s emerging “common wealth” debate is super helpful because it asks us to think about the full range of assets that might need to be folded into building new sovereign wealth funds, including spectrum auctions, public-land rents, water-use charges, infrastructure fees, royalties, digital and data levies and public equity holdings. Australia as it happens is also moving in this direction. Ministers have been very clear that major data centres must pay for their own energy needs and grid connections, not shift costs onto existing consumers.
For my money, I think we need to combine these ideas and radically widen the scope and powers of our current AI Sovereign Wealth Fund. The principles are basically these:
If the public incurs a cost, recover it. If the public allocates something scarce, price it properly. If public knowledge or infrastructure creates commercial value, retain a royalty or share. If taxpayers take equity-like risk, take equity-like returns. And if extraordinary economic rents eventually emerge, tax the rents rather than the productive investment that created them.
What might that include? There’s at least seven different kinds of assets and revenue flows:
Infrastructure charges and scarcity rents from grid connections (which could be auctioned like spectrum licences), public land, water and other scarce public assets.
Royalties and revenue shares from public data, intellectual property, research and government-backed validation. Think of the value of carefully stewarded access to the vast public data sets we own.
Equity stakes and warrants where taxpayers invest capital, provide guarantees or materially de-risk projects. This is sort of already covered by the existing AI Sovereign Wealth Fund.
Success payments and clawbacks on grants, concessional finance and developmental procurement when supported firms succeed. Crucially, we need to be co-developing systems using public procurement contracts in which the state actually owns an interest. The co-development power of a £20 billion state tech spend is currently under-leveraged.
Land-value capture where AI Growth Zones and public infrastructure sharply increase site values.
Taxes on exceptional AI rents or windfall profits if very large excess returns emerge. I’m personally less optimistic about exactly how this is going to happen, but let’s see.
Selected existing public assets transferred in at the outset to help capitalise the fund.
In short: recover public costs, price public assets properly and take a share of the upside where taxpayers help create it.
The proceeds - cash, shares, warrants, royalties and other assets - should then flow into the National AI Dividend Fund, which becomes the pool for different returns generated by the public contribution to the AI revolution. In time, as stakes are sold and revenues accumulate, the fund then invests in a diversified range of assets exactly like Norway's sovereign wealth fund, which was used to pool the gains of the country's oil wealth.
The dividends from this fund should then capitalise new universal savings accounts for every young person, created through a universal roll-out of auto-enrolment savings accounts. These ‘sidecar’ accounts have been piloted by the National Endowment & Savings Trust, which helps run the nation’s auto-enrolment pension system, and found to be an inspiring success.
In time, a national fund worth around £100 billion would be enough to pay out to every young person at the age of 21, a one-off dividend of around £4,000 which could be used, in effect, as an individual learning account to help pay for retraining should and when the need arise. But we could widen the potential uses of this fund to include a deposit to buy a house.
As the fund builds and the dividends rise, it could well become the foundation for a system of Universal Basic Capital and a cornerstone for a new wealth-owning democracy for the 21st century.
When everyone from the Pope to Bernie Sanders to Bill Gates to Sam Altman is arguing for new ways to share the fortunes created by the AI transformation, we know it's time to act. We made a mistake back in the 1980s when we gave away the proceeds of North Sea oil in tax cuts, which were here today and gone tomorrow. We must not make that mistake again.
It was Harry Bridges, the legendary leader of America’s dockers, who understood that he could not turn back the tide of mechanisation transforming the shipping industry in the 1960s. So he set himself a different question: how could his members win “a piece of the machine”? The challenge, Bridges understood, was to make the machines work for the workers, rather than against them. That “Bridges test” is once again the challenge for progressives as we confront a very different future of work.
Good stuff here 👏
Great piece Liam. What does this mean for East Birmingham and the West Midlands? Are there immediate things we as a devolved area can do in this space eg invest local pension funds, local grid queue auction etc?