Three Big Ideas #10
Special Compute Zones, an (in)tangible policy tweak, and a dig at invisible graveyards
Welcome to our weekly Three Big Ideas roundup, in which we serve up a curated selection of ideas (and our takes on them) in entrepreneurship, innovation, science and technology, handpicked by the team – and today, with a guest contribution!
⚡ Jack Wiseman, Inference Magazine
Since 2012, the computational power used to train the largest AI models has grown 100 million-fold. This has become enormously energy intensive: the most recent model from Meta used an estimated 27 megawatts of power capacity, which is approximately the same power required for 88,000 UK households. Were trends to continue, by 2030, training the largest model would use more than twice the UK’s entire energy generation capacity.
But as things stand, no developer would choose to build an AI datacentre in the UK: it can take up to 15 years to get a grid connection, and, once connected, the UK’s industrial electricity prices are four times higher than in the US and 45% higher than in France.
To rectify this, Britain needs to increase its energy supply. We investigated two options for doing so: one using a combination of wind, solar, grid-scale batteries, and natural gas backup; the other solely with nuclear power. Our modelling suggests that nuclear route has significant advantages for environmental impact and safety. Not only would the cost-minimising allocation of renewables plus batteries and backup require over 200 square kilometres of contiguous land area per gigawatt, and colocation with an LNG import terminal, it would also have 40% higher carbon emissions, and lead to 27 times more ‘expected deaths’ than using nuclear power, based on historical patterns.
However, at present, building nuclear power in the UK is hard. Compared to South Korea, our reactors are 4.5 times more expensive and take at least twice as long to construct. In our diagnosis, this is a choice. We recommend the creation of Special Compute Zones, with an alternative planning and regulatory approval process for nuclear power, AI datacentres, and the transmission and networking infrastructure. If these reforms allowed UK projects to close the cost gap with South Korean reactors by two thirds, nuclear power would become 37% cheaper than an equivalent blend of renewable power for datacentres. With these changes, the UK can become the world’s best place to build the infrastructure for the Fourth Industrial Revolution.
☁️ Philip Salter, Founder
As Stian Westlake and Jonathan Haskel argued in the groundbreaking work Capitalism without Capital, modern economies are increasingly driven by intangible assets – such as research and development (R&D), software, branding, and organisational processes – rather than traditional tangible assets like machinery and buildings. A recent paper by Timothy DeStefano et al. hammers home the importance of policymakers understanding this point.
Investigating the UK’s Annual Investment Allowance (AIA), which allows firms to deduct the cost of capital investment against profits, the authors find that the tax break boosted tangible IT investments by, on average, 41.3% for hardware and 27.8% for software, but also slowed the adoption of cloud services by 17 percentage points.
As a blog on the paper states: “This is a significant distortion, given that 56% of firms used the cloud by the end of our study period. Crucially, the capital incentive also slowed the diffusion of other data-intensive technologies, reducing AI and big data adoption by 3 and 18 percentage points, respectively. These effects were particularly pronounced for small and medium-sized enterprises, precisely the firms most likely to benefit from cloud computing’s flexibility and variable cost structure.”
The capital incentive policy also dampened demand for data-analytics workers. Their back-of-the-envelope estimates suggest that in the absence of the AIA’s distortive effects big data adoption in the UK could have been 14% higher, and AI adoption 30% higher.
There are plenty of bad policies the Government could adopt in their efforts to try to get Britain’s businesses to adopt AI – and I’ve heard a lot of them from other business groups over recent months in meetings with Ministers – but here is a straightforward one. We’ve done this before. Alongside Startup Coalition we successfully made the case for including intangibles in R&D Tax credits. It’s time the AIA followed suit.
🪦 Eamonn Ives, Research Director
One of my favourite turns of phrase in public policy debates is ‘the invisible graveyard’. Though rather morbid, it tells an instructive tale. Originally coined by economist Alex Tabbarok, the invisible graveyard refers to all of the deaths that occur while a drug which could have prevented them is stuck waiting to be approved by regulators. The notion forces us to remember that a strategy of not taking risks is far from being risk-free in itself.
Invisible graveyards don’t just exist in pharmaceuticals either. We can apply the concept across a swath of sectors that are tightly regulated. Red tape that stifles nuclear energy means we burn much more fossil fuel than we might otherwise, heating up the planet and polluting the atmosphere. Our inability to swiftly legislate to permit autonomous vehicles on our roads means, as I wrote in an earlier Three Big Ideas issue, that more crashes happen. Rules that make it harder than necessary to put lab-grown meat on supermarket shelves are bad news for consumers, and even worse for chickens.
Note that none of this is to say that rules are bad in and of themselves. Rather, it’s an argument for regulators to always be mindful that, where a problem exists, insisting on making a possible solution as safe as is conceivably possible isn’t always a recipe for safety.
Currently, applications are still open to Chair the newly established Regulatory Innovation Office – which could be a gamechanger for pioneering tech in Britain. Whoever gets the gig, let’s hope they’re aware of the invisible graveyard, and do what they can to keep it as empty as possible.
 | A guest post by
|