If Reeves wants Britain to lead in AI, she must offer tax incentives
The government is aware of the levers it’s able to pull to compete on AI: planning, electricity connections, visas, higher education policy, direct subsidies. But where’s tax? Asks Tim Sarson
A couple of weeks ago one of those mini news stories came along, the sort that is so neatly zeitgeisty that it demands further consideration. Mayor of London Sadiq Khan wrote to the CEO of Anthropic, Dario Amodei, to invite the company to move to London.
I’ve heard no reports of Mr Amodei’s response. But it’s still a fascinating story. Why? Because it brings together many strands of the current news agenda.
The global economic discourse this year, at least before the recent energy shocks, has been dominated by the rise of AI and the small collection of mainly US companies that are driving it. AI and quantum computing were pivotal features of Chancellor Rachel Reeves’ Mais lecture last Tuesday.
Then there’s the way US domestic political disputes become global stories and influence our own debates here, in this era of politics by social media.
And this is a further example of the inexorable rise of the metro mayors; another topic heavily featured in the Mais lecture. Note that it was Sadiq Khan making the approach, not Keir Starmer or Rachel Reeves. And the invitation was specifically to London, not the UK.
I’m not here to talk about the politics of all this, so let’s instead step back and think about the real future for the AI industry in the UK. When our chancellor talks about wanting to secure our place as a world leader in innovative technologies, is this wishful thinking or does our government hold the levers to get us there?
First, we need to be clear what it means to be a “world leader”. There are three big AI buckets policymakers talk about: physical infrastructure such as datacentres and the kit that goes with them; innovation and commercialisation by the technology companies that develop and productise the AI models; and take-up by industry and households to drive productivity growth and efficiency.
I’m going to focus today on that middle one: attracting the businesses that develop and make money out of the technology and ensuring that their employment opportunities and investment dollars come our way, not elsewhere.
Strong domestic talent pool
I’ve been pleasantly surprised recently by just how strong our domestic AI talent pool is. It’s a “problem” for some multinationals who would like to keep their valuable functions in places with low tax rates but are finding all the good people here. Yes, Silicon Valley will always win on sheer numbers and the depth of investor pockets, but our universities and start-ups have been churning out an enviable number of luminaries, up to and including Nobel prize winners. And, with apologies to Messrs Burnham and Houchen, they’re concentrated in London and the Oxbridge corridor.
We’re a major European hub for datacentres and several of the US giants driving the AI revolution have large offices here.
But that’s not enough. Never mind the megawatts of datacentre capacity we’re building here, they’re counting it in multi-gigawatts over in the USA. And those technology giants own almost all the Intellectual Property (IP) that’s driving AI growth. In due course, when the technology matures, they will be raking in most of the taxable profits that come from it.
Britain’s job is to secure at least our fair share. We can’t compete in market scale or depth of talent pool with the USA, but we can try our best to get a piece of the global action while keeping as much as possible of our own. It’s always a little sad to see a promising UK startup snapped up by a US buyer whose first action after completion is to strip all the IP out to the mothership. How can we stem the flow?
The government is aware of the levers it’s able to pull: planning, electricity connections, visas, higher education policy, direct subsidies. But where’s tax? We sorely lack a tax incentive to bring AI profits here.
I asked AI app Claude what might induce its owners to set up shop in Britain. It seemed appropriate to get ideas from the horse’s mouth. Claude came back with five asks, and its number one is to extend our Patent Box regime to cover AI applications.
Extend the patent box regime to cover AI applications
I agree with Claude.
A successful software company here pays 25 per cent tax on its profits. It may claim a Research and Development (R&D) credit, but these are more like a subsidy of operating costs than a tax break, and they come regardless of IP ownership. They encourage businesses to spend money here, and the regime has been very successful, but it doesn’t induce them to bring in their profits. That’s what our patent box regime does: gets the tax rate down to 10 per cent on certain types of income. The Patent Box is great. I’ve no doubt it’s one of the main reasons we still have a successful life sciences industry in this country.
But most AI models and applications are unpatented. In fact, software as a class is rarely patented. So, if you’re a tech or AI company here you’re most likely not benefiting. In the US if you combine the right credits and deductions with their preferential rate on exports of goods and services, you can be paying a far lower rate. Not forgetting the 12.5 per cent on offer in our neighbour across the Irish Sea.
We need an IP box regime that extends to include innovative software applications including AI. I have enough experience of large multinationals buying UK businesses and immediately exiting their IP to the USA to be confident that a competitive tax rate would stem that flow.
Officials will worry that it would be hard to circumscribe and rather open-ended, and that it will cost a lot of money. But that circumscription is possible: just ask the Netherlands, whose innovation box is wide enough to include AI, or France which grants benefits to IP that is patentable even if not patented. And, as our pharma experience shows, it may well pay for itself. The more favourable the regime, the more profit it makes sense to allocate there.
As one of Claude’s arch-rivals, ChatGPT, might put it: that’s not hallucination; that’s gravity.
Tim Sarson is head of tax policy at KPMG