What makes a manufacturing superpower?

Some reflections on Breakneck: China’s quest to engineer the future by Dan Wang.

Dan Wang’s new book on China is rightly getting great reviews. It’s a compelling read, engagingly written, reflecting both the author’s deep understanding of China’s developing economy, and his personal sympathy with the Chinese nation. It is admiring of Chinese achievements over the last couple of decades , while being entirely clear-eyed about the deficiencies of the political system and its human costs.

The big idea behind the book is to compare and contrast the two great powers of the world today – China and the USA, summarising that comparison in a neat formula. For Wang, China is the Engineering State, while the USA is the Lawyerly Society – and from that contrast, the complementary strengths and weaknesses of the two nations can be derived.

What kind of state is China? According to Wang, it is a “Leninist Technocracy with Grand Opera tendencies”.

It’s Leninist – because it concentrates all power in the Chinese Communist Party, which will ruthlessly suppress any potential or emerging rival power structure, whether that’s an independent trade union, a civil society group, or an emerging class of plutocrats.

It’s a technocracy, because it is dominated by engineers, and brings an engineering mindset of finding technological solutions to all kinds of problems – whether material or social. In many ways, it’s a realisation of the high modernist vision of rule by experts, and Wang traces one of the big sources of difference between the China and the USA to the Western backlash against high modernism in the 60s and 70s.

Its “Grand Opera tendencies” refer the state’s tendency to intersperse periods of stability with occasional episodes of government triggered chaos – the Great Leap Forward, the Cultural Revolution – and recently, the excesses of the Net Zero policy. If this sounds like a flippant way to describe episodes containing great human tragedy, I should stress that Wang’s descriptions of, for example, some of the consequences of the one-child policy are truly harrowing.

What are the positives of the engineering state? China has built a huge amount of infrastructure – interstate highways, high speed railways, dams and power stations – at a pace – and low cost – that seems incomprehensible to Americans (or, indeed, we British). Wang acknowledges the downsides – human costs of displaced people, environmental damage, unsustainable debt overhangs. But his overall sympathies here are with the engineering state – the nowheres of “bridges to nowhere” frequently become somewheres as a result of those bridges.

Chinese achievements in infrastructure are impressive, but perhaps the most significant recent development in the Chinese economy has been the achievement of mastery in high technology manufacturing. This now amounts to technological near-supremacy in consumer electronics, telecommunications equipment and battery electric vehicles, as well as complete domination of solar cells.

This supremacy hasn’t come from frontier science, but from exalting process knowledge, and building and sustaining the engineering communities of practise that holds that process knowledge. “It’s another of the ways that the United States and China are inversions of each other. Americans expect innovations from scientists working at NASA, in universities, or in research labs. They celebrate the moment of invention: the first solar cell, the first personal computer, first in flight. In China, on the other hand, tech innovation emerges from the factory floor, when a new product is scaled up into mass production. At the heart of China’s ascendancy in advanced technology is its spectacular capacity for learning by doing and consistently improving things.”

Ironically, it has been US firms – particularly Apple and Tesla – outsourcing their own manufacturing that have driven the development of these Chinese engineering communities of practise. Apple taught Chinese engineers how to build iPhones, and in the process built sophisticated supply chains for electronic components which can then feed new industries, such as drones. Now, in a colder geopolitical climate, the Chinese Communist Party seeks to achieve similar dominance in areas like semiconductor manufacturing.

Silicon Valley used to be such a community of engineering practise, but in Wang’s view it has lost its way. Wang writes of Andy Grove, former CEO of Intel, that he “saw Silicon Valley transition from doing both invention and production to specializing only in the former. And he understood quite well that technology ecosystems would rust if the research and development no longer had a learning loop from the production process.”

The recent troubles of Intel, slipping behind TSMC at the frontier of semiconductor manufacturing, perhaps vindicates that view.

China has certainly made a firm decision to focus on manufacturing, developing science based industries like semiconductors & clean technologies, while actively suppressing the tech industry as Silicon Valley currently understands it. Xi has been discomfited that “the cutting edge of the economy seemed to have been driven by the vagaries of investors rather than the interests of the state”, while believing that social media promises “little gain in economic productivity while creating huge potential for social unrest”.

One aspect that Wang does not treat in great detail is the relationship between scientific research in China and its high technology manufacturing base. Anyone who reads the scientific literature will appreciate, not just the hugely increased volume of contributions from Chinese scientists, but also their quality and impact. What I don’t have a good sense of is how closely coupled this impressive academic effort is to the growing industry base.

But in my understanding, there are now areas where Chinese science is in the ascendancy, with brilliant individual scientists given huge resources to accelerate new applications. Quantum technology is surely one such field, under the leadership of scientists like Pan Jianwei.

What Wang does make clear is that he doesn’t go along with the comforting notion that the creativity of science can’t exist with the suppression of free speech in a Leninist state: “How is it that science can coexist with autocracy? Mostly, I believe, because the precondition for science is that abundant funds are far more critical to science than free speech, and that is something that dictators can deliver”.

But Wang does highlight the way in which the increasingly authoritarian tendencies of the Chinese state under Xi make the country increasingly uncomfortable for young creative types. Tech entrepreneurs have found that their wealth and business success is no protection against a change of mood in the Party-state, of the kind we saw in the 2021 crackdown on the tech industry.

We’re now entering a much darker world; Xi is preparing for the worst, asking China to “be ready to withstand the major test of high winds, choppy waters and even dangerous storms”. One such test, no longer inconceivable, would be outright war with the West. This fortress mentality has led to a focus on resilience, energy and food security rather than economic efficiency. In this environment, China’s manufacturing base is a huge strategic advantage.

Wang’s book is built around the contrast between China and the USA – but it’s still highly relevant for those of us in smaller countries like the UK. Wang himself was brought up in Canada, so perhaps this perspective isn’t entirely foreign to him [2].

While the UK is barely mentioned in the book, we know from his other writing some of Wang’s views about it. In his 2020 letter, he stressed that China couldn’t follow the example of the UK, “which is so successful in the sounding-clever industries—television, journalism, finance, and universities—while seeing a falling share of R&D intensity and a global loss of standing among its largest firms.” He followed this up with this comment on Twitter [3].

I don’t think the UK is a lawyerly society in the same way as the USA. It certainly suffers some parallel problems in its inability to build infrastructure, but its productivity slowdown and loss of high tech manufacturing has been even more marked. The lesson of the UK is that development isn’t always one way – countries can de-develop, losing technological capacity through a loss of process knowledge and the engineering communities of practise that sustain it.

Wang’s book ends with the hope that America can declare itself to be a developing country – “able to reform itself, get unstuck from the status quo, and ultimately unlock as much as possible of human potential.” I hope that every country can think this way.

Notes:
[1] A couple of other reviews of Breakneck that I found interesting:
Jonathon Sine

Henry Farrell

A nice interview with Dan Wang himself

[2] I’m reminded of a very apposite saying by the historian David Edgerton: “The UK is not a small USA, it is a big Canada”.

[3] So who are these British nay-sayers that so disloyally cheer Wang on in his denigration of the UK’s economic choices? Oh.

Another Modern Industrial Strategy

This is a slightly expanded version of an article published last week in Research Professional – The latest industrial strategy has made choices

Last week’s Industrial Strategy Policy Paper is the latest chapter in the chequered history of UK Industrial Strategies. For nearly three decades after Thatcher’s ascent to power, the UK’s strategy was not to have an industrial strategy, which was a concept associated with money-losing supersonic airliners and cars with square steering wheels. But that conventional wisdom has been challenged by a global financial crisis and nearly two decades of economic stagnation, so after a number of stops and starts over the last decade, a fully developed Industrial Strategy has now arrived.

This strategy is not entirely new – there are strong echoes of the 2017 strategy from Theresa May’s government, driven by the then Business Secretary Greg Clark. The five “critical technologies” of the 2023 Science and Technology Framework are preserved, and other interventions of the previous government, like Investment Zones and Freeports, survive with only minor rebranding. This continuity is to be welcomed.

The first test of any strategy is whether it makes choices. Here I think the signals are positive. The Green Paper identified 8 key sectors, but now 30 detailed sub-sectors, “frontier industries”, are identified. This welcome degree of granularity has been underpinned by a detailed methodology, evaluating both the economic potential of the frontier industries, and the degree to which policy can have an impact.

So, under the banner of Advanced Manufacturing, aerospace and automobiles appear as expected, but perhaps the inclusion of agri-tech is less anticipated. Clean energy industries include wind (onshore, offshore and floating), fission and fusion, but not solar. There’s more focus on non-STEM-based sectors than in previous strategies, including creative, financial services and professional and business services. Unsurprisingly, given the current geopolitical situation, defence is prominent, but here the final list of frontier industries awaits the publication of a Defence Industrial Strategy.

The geographical aspect of industrial strategy is much more prominent in this strategy than in predecessors. An analysis of industrial clusters leans heavily on the previous government’s cluster map, as well as accepting the importance of increasing the productivity of the UK’s underperforming second tier cities. But in identifying those places with existing strengths in the strategy’s priority frontier industries, or with the potential for better exploiting agglomeration economies, some places won’t make the cut. There will be gaps in the Industrial Strategy map.

Discussions of industrial strategy distinguish between “horizontal” interventions, that aim to improve the business environment without discriminating between sectors, and “vertical” policies focusing on particular types of industry. Here, the signals are that even the more “horizontal” measures will be tilted in favour of the (roughly) 30 frontier industries.

Addressing the cost of energy is prominent here; the effect of the very cost of energy relative to overseas competitors on the viability of many UK industries has been much discussed recently. Perhaps just as important as the cost of energy are the delays in making new grid connections for industrial electricity. Here there will be difficult choices too; if industrial energy prices are lowered, someone else will have to pay more, whether that is the domestic consumer, or the tax-payer. We’re paying for policy choices made a decade (and longer) ago, which resulted in an over-reliance on gas & an underinvested network.

Other “horizontal” interventions include trade policy, access to finance, addressing regulation and removing planning barriers. Skills and innovation are two areas where the choice needs to be made about how much to direct policy towards one’s selected sectors.

The skills section seems slightly underpowered; it highlights skills shortages in the priority sectors, recognises that we have too many underqualified young people, and acknowledges that adult and continuing education has declined. But solving these problems is laid at the door of Skills England (and counterparts in devolved nations). An infographic makes a welcome assertion of the central role of universities in delivering an industrial strategy, but without recognising the pressure the sector is under.

UK Government research and development spending since 2012. Outturn: ONS Research and development (R&D) expenditure by the UK government, 2023 Datasets. Allocations from CSR 2021 and CSR 2025 corrected for inflation using GDP deflators, OBR March 2025 predictions for future years.

For innovation, it’s appropriate to recognise that government spending on R&D is now at a high level – the last government did make significant increases in R&D spending, and the recent spending review locked those increases in place in real terms. The question will be, to what extent, that R&D spending will be directed at the industrial strategy priority areas.

Here the document is unequivocal on the principle – research funding, including that delivered through UKRI, will be pivoted towards the industrial strategy priorities. Again, the need to make choices is accepted. As widely trailed, R&D funding will be more rigorously sorted into three categories; basic curiosity-driven research, delivery of government priorities, and supporting private sector innovation-led growth. What remains to be decided are the relative size of these three buckets.

Will this industrial strategy deliver the economic growth that has been so lacking for nearly two decades? I think the two key requirements for success are genuine commitment across the whole of government, and consistency of purpose over time, with a mechanism for tracking progress and holding the government to account.

On the former, there are encouraging signs that the strategy has the support of Treasury; in addition to its sponsor department, the Department of Business and Trade, the priorities of DSIT and DESNZ are clearly reflected. But can the Home Office can be persuaded that some skilled immigration might be needed? Does DfE recognise that universities are actually quite an important part of the skills system?

Most importantly, this strategy probably needs a decade to make a significant impact. It’s good that it will be supported and monitored by an Industrial Strategy Council with statutory powers and supporting infrastructure. It would be even better if it could generate the kind of consensus that would allow it to survive the inevitable political changes to come.

The civic university in hard times

Universities in the UK at the moment are broke and unloved. In these circumstances, the temptation is going to be to withdraw to “core business” – teaching students, and for research intensives, doing the kind of research that pushes the institution up the international league tables, to attract the overseas students whose fees prop the whole system up. In a period of retrenchment, it might be tempting for managements to see supporting the role of universities in their communities as a dispensable luxury. I think this would be a profound mistake.

This isn’t to understate the difficulty UK universities find themselves in. Around three quarters of them are expected to be in deficit next year, and about a hundred are now actively restructuring or making staff redundant. This follows a 40% real terms erosion in fees for home students, and a business model, reliant on growing overseas student numbers, which has become both politically unpopular and exposed to geopolitical risk. The latest proposal – of a levy on international student fees – is both another financial blow, and a symbol of the way universities find themselves on the wrong side of culture war discourse. What’s quite clear is that, whatever recognition there might be in government of the university sector’s troubles, the sector is simply not a high priority for a government facing difficult issues on all sides.

Meanwhile the UK’s slow-burning economic crisis continues. At the root of this is the UK’s weak productivity growth since the GFC or before; productivity is now about 25% down on what it would have been if the pre-GFC trend had continued. This is the direct cause, not just of the very weak wage growth seen in the last 15 years, but also of the fiscal difficulties recent governments have faced, and the resulting pressure on public services. Much of what’s gone wrong with politics and economics in the UK can be put down to this productivity problem.

There’s a big regional dimension to the productivity problem. In contrast to most developed countries, the UK’s big second tier cities aren’t drivers of national productivity; instead their productivity is lower than the UK average. And beyond the big cities in the North and Midlands, there are very weak peripheral economies in in deindustrialised areas and coastal towns. In these places, local government services are under huge pressure, having received the brunt of the 2010’s austerity, while remaining legally obliged to maintain increasingly costly services like social care and supporting children with special educational needs and disabilities.

Thinking about these two factors together, it should be clear that no good will come of universities just asking the government to hand over more money, to carry on doing the things they’ve always been doing. However bad the crisis in universities is, the larger economic problems, and the constellation of social problems that follow from those, mean that universities are just too far down the list of priorities for the government.

Instead, universities need to lean in to a national effort to deal with the country’s economic and social problems, with a particular emphasis on the problems of the places where they are located and the communities that live in those places. Many universities are already doing good work in this area, but upping the intensity of these activities probably does mean that universities will have to do some things differently. Already, a letter from the Secretary of State to University leaders, sent last November, has identified 5 priorities, two of which are highly relevant here: “Make a stronger contribution to economic growth” and “Play a greater civic role in their communities”.

The two are closely linked. UK universities are institutions with international reach, but they’re still located in specific places and that’s where their influence is greatest. What can they do for the towns and cities where they are located? To our citizens, perhaps the most tangible impact of universities on their towns and cities is in the professionals they train – the teachers who teach their children, the nurses, doctors and health workers who treat them and their families. University research can provide an evidence base to inform policy for local and regional governments. Student and staff volunteering can support charities and community groups. The cultural institutions associated with universities, and outreach activities in schools and communities can support the cultural vitality of places.

We do, though, need to face the fact that, for many of the communities which neighbour our universities, the fundamental problem is that they are in materially poor places. If we are to be serious about asserting the economic value of our universities and the research they do, we need to demonstrate that, by supporting efforts to make communities more prosperous. This means supporting places in attracting and growing the kinds of businesses that produce tradable goods and services; this in turn will bring money into those places, and support higher wages for people with all levels of skills. In short, universities need to support inclusive economic growth in their towns and cities.

This will require universities to think differently. We’ve focused a lot on a relatively narrow, linear model of knowledge exchange, where university academics do research leading to protectable intellectual property, which then is used to produce spin-out companies. This remains important – though we do need to think about how such spin-out companies can be encouraged to scale-up in the UK to the point at which they produce material economic impacts on communities, rather than just making a few founders and venture capitalists wealthy (important though that is for creating the right incentives). But we need to take a much broader view of the innovation economies of our places.

Taking this wider view of our innovation ecosystems, increasing the number of spin-outs – both from students and staff – is important, as is the need for them to scale. But universities need to think hard about how they support the existing business base – not just big, technologically sophisticated firms that are accustomed to working with universities, but also the small and medium size enterprises who often find the labyrinthine structures of universities impossible to penetrate, and find it difficult to access financial support from government agencies. SMEs are important, but we do need to recognise the power of multinationals operating at the technology frontier in building manufacturing and innovation capacity in those parts of the country where productivity is lagging. Cities and regions are always anxious to attract inward investment, but this needs to be targeted on those firms whose presence can most effectively drive up a region’s wider innovation capabilities.

A key problem for many places with low productivity is a lack of skilled people. This is often framed as a supply problem – so, it is argued, simply increasing the supply of skilled people will help solve the productivity issues in weak regional economies. But the problem could equally be one of demand – in a town with few opportunities for high-skilled work, for people who don’t want to leave, there isn’t the incentive to put time and money into increasing their level of skills. Towns with weak economies are often locked into a low skills, low productivity, low innovation bad equilibrium – and to break out of that equilibrium, both halves of the problem need to be addressed. Existing businesses need to be supported in their innovation ambitions and new and innovative businesses need to be attracted, while those businesses need to be assured that the higher skills they need will be available to them. We need to link up the skills and innovation agendas.

This leads us to think about geographies. The excellent research facilities that our big civic universities have may seem a long way away from businesses in a town or suburb on the other side of the city centre – particularly given the poor state of transport in our cities outside London. So universities need to think hard about how to maximise their impact, not just on their campus and its immediate environs, but across whole city regions, and beyond.

This means working in partnership. The different HE institutions in a city region should work together, bound by civic university agreements, that commit them to concrete actions in support of their place. For example, the expertise of technology transfer offices can be shared across a city, the business schools could collaborate to provide management training to local SMEs, there can be a single approach to widening participation, and lifelong learning. Institutions could collaborate to develop initiatives for translational R&D and innovation diffusion focused on regional business.

These collaborations should go beyond HE institutions. Further education has a huge role to play in a joined up approach to innovation and skills; in many left-behind towns FE colleges are the anchor institutions, with strong employer networks. Partnership with businesses is crucial, and here business sector groups can be invaluable in articulating the collective needs of businesses in places. Strong relationships between universities and their local and city-regional governments are crucial. Universities should be working together with local and city-region authorities in developing and implementing local industrial strategies and growth plans.

My own work in the University of Manchester, as VP for Regional Innovation and Civic Engagement, has been guided by these principles. We’ve established a Civic University Agreement, supported by GM’s five universities and endorsed by the Mayor, and developed a collaborative body, Innovation GM, which brings together the HE sector with business sector groups and the GM Mayoral Combined Authority. We were awarded one of the pilots in the Innovation Accelerator programme, which we’ve used to support innovation diffusion in key sectors like digital and AI, and specifically to support the materials and manufacturing sector in Northeast Greater Manchester (Bury, Rochdale and Oldham), which remains one of the most economically and socially disadvantaged part of the conurbation, and forms a priority growth location through the “Atom Valley” Mayoral Development Zone. Collaborations with the FE sector have been growing, helped by the increasing coordination of GM’s nine FE colleges through the GM Colleges Group; Innovate UK’s very welcome Further Education Innovation Fund have allowed us to directly connect the FE colleges with GM’s innovation ambitions. Within the University, we’ve established a new unit, Unit M, with a specific mandate to own and drive our regional innovation agenda. I’m delighted that the government announced, in the June Comprehensive Spending Review, a Local Innovation Partnership Fund, which will build on the Innovation Accelerator pilots, extending them to 10 city-regions in the UK.

Finally, it’s worth coming back to money – and the current financial difficulties of the HE sector. It would, I think, be wrong to think of these civic activities as big sources of new funding, establishing new profit centres. Universities do need to be resourced to do new things, but this isn’t going to be an immediate boost to their bottom lines. What these civic activities will do, though, is give us a social licence to operate – and, in many cases, reconnect with the original purposes of our institutions.

Moore’s Law, past and future

Moore’s Law – and the technology it describes, the integrated circuit – has been one of the defining features of the past half century. The idea of Moore’s law has been invoked in three related senses. In its original form, it was rather a precise prediction about the rate of increase of the number of transistors to be fitted on a single integrated circuit. It’s never been a law – it’s been more of an organising principle for an industry and its supply chain – and thus a self-fulfilling prophecy. In this sense, it’s been roughly true for fifty years – but is now bumping up against physical limits.

In the second sense, Moore’s law is used more loosely as a statement about the increase in computing power, and the reduction of its cost, over time. The assertion is that computing power grows exponentially. This also was true, for a while. From the mid 1980’s to the mid 2000’s, computer power grew at a rate of 50% a year compounded, doubling every two years. In this extraordinary period, there was more than a thousandfold cumulative increase over a couple of decades.

The rate of increase in raw computer power has slowed substantially over the last two decades, following the end of Dennard scaling and the limitations of heat dissipation, but this has been counteracted to some extent by software improvements and the development of architectures specialised for particular applications. For example, the Graphics Processing Units – GPUs – that have emerged as being so important for AI are highly optimised for multiplying large matrices.

In the third sense, Moore’s Law is used as a synecdoche for the more general idea of accelerating change, that the pace of change in technology in general is exponential – or even super-exponential – in character. This of course is a commonplace in airport business books. It underpins the idea of a forthcoming singularity, as a received wisdom in Silicon Valley. The idea of the singularity has been given more salience by the recent rapid progress in artificial intelligence, and the widespread view that superhuman artificial general intelligence will soon be upon us.

In this post, I want to go back to the fundamentals – how much the basic components of computing can be shrunk in size, and what the prospects for future miniaturisation are. But this does directly bear on the question of the prospects increase in computer power, which has taken on new importance, as the material basis of the AI boom. AI has brought us to a new situation; in the classical period of fastest growth of computer power (the 80s and 90s) the supply of computing power was growing exponentially, and the opportunity was to find ways of using that power. Now, with AI, it’s the demand for computing power that is growing exponentially, and the issue is whether supply can match that demand.

Moore’s Law. From Max Roser, Hannah Ritchie, and Edouard Mathieu (2023) – “What is Moore’s Law?” Published online at OurWorldinData.org. Retrieved from: ‘https://ourworldindata.org/moores-law‘ [Online Resource]. Licensed under CC-BY.

A classical depiction of Moore’s law is shown in this plot from Our World in Data – with a logarithmic y-axis, a straight line indicates an exponential growth in the number of transistors in successive generations of microprocessor. The seemingly inexorable upward progress of the line conceals a huge amount of innovation; each upward step was facilitated by research and development of new materials and new processes. It also conceals some significant discontinuities.

For example, the earlier relationship between computer power and number of transistors was broken in the mid-2000s. Before then miniaturisation brought a double benefit – it gave you more transistors on each chip, and in addition each transistor worked faster, because it was smaller. The latter relation – Dennard scaling – broke down, because heat dissipation became a limiting factor

Another fundamental change happened in 2012. The fundamental unit of the modern integrated circuit is the metal oxide silicon field effect transistor – the mosFET. This consists channel of doped silicon, with contacts at either end. The channel is coated with a thin, insulating layer of oxide, on top of which is a metal electrode – the gate. It is the gate which controls the flow of electrical current through the channel. When physical limits meant that the planar mosFET couldn’t be shrunk any more, a new design flipped the channel into the vertical plane, so the transistors took the form of fins standing up from the plane of the silicon chip. Each side of the doped silicon fin is coated by insulating oxide and a metal gate, to form the finFET.

The patterns that make the circuits in integrated circuits are made by lithography – light is shone through a patterned mask onto a photoresist, which is subsequently developed to make the pattern physical. The lower limit on the size of the features that can be patterned in this way is ultimately set by the wavelength of light used. Through the 2010’s, lithography was based on using deep ultraviolet light created by excimer lasers – with a 193 nm wavelength. By 2020, this technique had been squeezed as far as it would go, and the 5 nm process node uses extreme UV, with a wavelength of 13.5 nm. The Dutch company ASML has a monopoly on the tools to produce EUV for lithography, each of which costs more than $100 million; the radiation is created in a metal plasma, and has to be focused entirely by mirrors.

I’ve referred to the 2020 iteration of fabrication technology as the “5 nm process”, following a long-standing industry convention of characterising successive technology generations through a single length. In the days of the planar mosFET, a single parameter characterised the size of each transistor – the gate length. There was a stable relationship between the gate length and the length characterising the node number, and there was a roughly biennial decrease in the node number, from the 1982 1.5µm process that drove the explosion of personal computers, to the 2002 90 nm process of the Pentium 4. But with the replacement of the mosFET by the finFET, circuit geometry changed and the relationship between the node size and actual dimensions of the circuit broke down. In fact, the node size now is best thought of as entirely a marketing device, on the principle that the smaller the number the better.

A better way to describe progress in the scaling down of the size uses an estimate of the minimum possible area for a transistor as the product of the metal pitch, the minimum distance between horizontal interconnects, and the contacted gate pitch, the distance from one transistor’s gate to another’s.

Minimum transistor footprint (product of metal pitch and contacted gate pitch) for successive semiconductor process nodes. Data: (1994 – 2014 inclusive) – Stanford Nanoelectronics Lab, post 2017 and projections, successive editions of the IEEE International Roadmap for Devices and Systems

My plot shows the minimum transistor footprint, calculated in this way, for each process node since 1994 (the 350 nm node). The first five nodes – until 2002 – track the exponential increase in density expected from Moore’s law – the fit represents a transistor density that doubles every 2.2 years. The last three generations of planar mosFET technology – until 2009 – show a slight easing of the pace. The switch to the finFET prolonged the trend for another decade or so. But it’s clear now that the “2 nm” node, being introduced by TSMC this year, confirms a marked levelling off of the pace of miniaturisation. For this node, there has been another change of geometry – finFETs have been replaced by vertical rows of nanowires, each completely surrounded by the metal of the gate electrode – GAA, for “gate all around”.

It has to be stressed that miniaturisation of transistors is far from the only way in which computer power can be increased. A good illustration of this comes from progress in making the ultra-powerful chips that have driven the current AI boom, such as Nvidia’s H100. The H100 itself was actually fabricated by TSMC on the “5 nm” node, the first to use AMSL’s EUV light source for lithography. But, as this article explains, only a fraction of the performance improvements of the H100 over previous generations are attributed to Moore’s law. Much of the improvement comes from more efficient ways of representing numbers and carrying out the arithmetic operations that underlie artificial intelligence.

Another factor of growing importance is in the way individual silicon chips are packaged. Many modern integrated circuits, including the H100, are not a single chip. Instead several individual chips, including both logic and memory, are mounted together on a silicon substrate, with fast interconnects to join them all up. The H100 relies on an TSMC advanced packaging technology known as “Chip on Wafer on Substrate” (CoWoS), and is an example of a “System in Package”.

What does the future hold? The latest (2023) iteration of the IEEE’s International Roadmap for Devices and Systems foresees one more iteration of the Gate All Around architecture. The 2031 node is a refinement of that which stacks two mosFETs on top of each other, one with a p-doped channel, one with an n-doped channel (this combination of p- and n- doped FETs is the fundamental unit of logic gates in CMOS technology – “complementary metal oxide silicon”, hence this is referred to as CFET). This essentially doubles the transistor density. After this, no further shrinking in dimensions is envisaged, so further increases in transistor density are to be obtained by stacking multiple tiers of circuits vertically on the wafer.

So what’s the status of Moore’s law now? I return to the 3 senses in which people talk about Moore’s law – as a technical prediction about the growth in the number of transistors on an integrated circuit, as a more general statement about increasing computer power, and as a shorthand for talking about accelerating technical change in general.

In the first, and strictest, sense, we can be definitive – Moore’s law has run its course. The rate of increase in transistor density has significantly slowed since 2020, and exponential growth with an increasing time constant isn’t exponential any more. The technology in its current form has now begun to hit limits, both physical and economic.

For the second, looser, sense, things are more arguable. Available computing power is still increasing, and we see the outcomes of that in advances such as the development of large language models. But this increased power is coming, less from miniaturisation, more from software, specialised architectures optimised for particular tasks, and advanced packaging of chips in “Systems in Package”. It’s this transition that underlies the fact that Nvidia is worth more as a company than TSMC, even though it’s TSMC that actually manufactures (and packages) the chips.

But I wonder whether these approaches will be subject to diminishing returns, in contrast with the classical period of Moore’s law, when constant, large, fractional returns were repeated year after year for decades, producing orders of magnitude cumulative improvements. We are also seeing as a major source of increasing computer power the brute-force approach of just buying more and more chips, in huge, energy consuming data centres. These kind of increases in computer power are fundamentally linear, rather than exponential, in character, and yet they are trying to meet a demand – largely from AI – which is growing exponentially.

It’s very tempting to take Moore’s law as an emblem of the idea that technological change in general is accelerating exponentially, but I think this is unhelpful. Technology isn’t a single thing that improves at a given rate; there are many technologies, and at a given time some will be accelerating, some will be stagnating, some may even be regressing. As we have seen before, the exponential improvement of a single technology never continues forever; physical or economic limits show up, and growth saturates. Continuous progress needs the continuous introduction of new technologies which can take up the baton of growth from those older technologies, whose growth is stalling.

It should be stressed here that when we talk about the end of Moore’s law, the technology that we are talking about isn’t computing in general – it is this particular way of implementing machine logic, CMOS (complementary metal oxide semiconductor). There are many ways in which we can imagine doing computing – the paradox here is that CMOS has been so successful that it has crowded out alternative approaches, some of which might have significant advantages. For example, we know that CMOS logic uses several orders of magnitude more energy per operation than the theoretical minimum (the Landauer limit).

Finally, it does bear repeating what an extraordinary period the heyday of Moore’s law and Dennard scaling was, with computer power doubling every two years, sustained over a couple of decades to produce a cumulative thousand-fold increase. For those who have lived through that period, it will be difficult to resist the belief that this rate of technological progress is part of the natural order of things.

The economic impact of AI: three scenarios

Eighteen months ago I wrote about the potential economic effects of artificial intelligence – drawing attention to a “new Solow paradox”, in which rapid technological progress in machine learning and artificial intelligence has coexisted with continuing stagnation in productivity. What has happened since then?

Technical progress has continued, driven by massive investments, both in the companies developing the technologies themselves and in the huge server farms that are needed to provide the computing power to train and implement the models. Interestingly, there has been a significant upturn in productivity in the USA since 2023 (analysed in this recent Resolution Foundation report, PDF). Some of this is due to the USA’s increasing production of oil and gas, and some from the tech sector itself. But there has been a significant growth in productivity in those service sectors that use, rather than develop, technology. It’s at least plausible that some of this is being driven by the adoption of AI. Perhaps we’re seeing the beginnings of a resolution of the new Solow paradox.

What’s going to happen next? To try and clarify some of the widely divergent assumptions, it’s perhaps helpful to sketch out some scenarios. My primary purpose here isn’t to try and guess the most likely future outcome – personally, I’m deeply uncertain as to what’s going to happen. Instead, I think it’s interesting enough to ask what assumptions various actors are operating under, and how those assumptions themselves constrain and influence the future.

1. Intelligence explosion

In this scenario, two dynamics lead to the transformation of economy and society through AI. The first is a process of recursive self-improvement, by which the application of AI technologies to develop the AI methods themselves leads to a runaway process in the growth of the power and effectiveness of those methods. The second is an increasing application of AI to the physical world, leading to rapid technological progress in all fields. The outcome is a winner takes all economy, in which the controllers of the new technologies enjoy unprecedented political and economic power.

One of the most compelling early use cases of LLMs has been to write code, so it’s a natural extension to think that AI systems can be used to do systematic computer experiments in order to find an optimise algorithms for machine learning – it seems a fair assumption that this is already happening, contributing to the progress we’re seeing in the development of LLMs and reasoning systems.

But to make the hoped for transformational impact on the economy and society, artificial intelligence needs to have a more direct interaction with the physical world than simply through existing corpuses of text. This needs the incorporation of real time data of all kinds, together with improvements in robotics to intervene in the world. Self-driving cars are the prototype system here, but if AI is going to accelerate technological progress itself we need to move to self-driving laboratories.

Fully automated scientific discovery leads to rapid progress in medicine, but the biggest impact comes in developing the hardware infrastructure of computing itself. The planar CMOS integrated circuits that computing depends on now are replaced by new 3d assemblies of nanoscale electronic components, brought together in systems of ungraspable complexity. And so increasing computer power in turn feeds the acceleration of this intelligence explosion.

What does the economy look like in this scenario? It’s winner-take-all – the firm, organisation or nation which achieves this goal first accumulates an unprecedented degree of economic and political power. On the other hand, technological progress becomes so rapid that there’s a hope that everyone benefits

Who believes in this scenario? My impression is that this is a relatively conservative version of the Silicon Valley consensus, summarised in the near-universal SV opinion that artificial general intelligence (a term usually left poorly defined) is imminent. If this is what one believes, then the logical course of action is devote all possible resources to achieving this goal as soon as possible. It’s obviously a matter of self-interest to be one of the controllers of such an all-powerful technology.

But altruists can also reassure themselves that entirely focusing on this goal is the most effective way of solving any global problem. The corollary is that normal approaches to scientific and technological progress will soon become obsolete, so the existing scientific enterprise becomes less and less relevant and doesn’t need to be sustained.

2. Excel in prose

In this scenario, the development of large language models has essentially solved the problem of automating verbal reasoning, in the same way that spreadsheets automated arithmetic and bookkeeping. As happened with spreadsheets before them, this leads to the quiet transformation of most business processes. Productivity growth recovers, perhaps even doubling, to return to levels seen in the 1990’s. Beyond LLMs, the application of machine learning and artificial intelligence to the physical world continues to make incremental progress, though technological progress in the physical world remains markedly slower than in the digital world.

The first killer application for LLMs was machine translation, now a solved problem. Many of the problems of information flow in big organisations are susceptible to automation by LLMs – producing meeting notes, summarising documents, generating routine communications. As in previous technologies, the key factor limiting the speed of uptake is the need to adapt existing processes to create places where LLMs can contribute, but the number of compelling use cases steadily expands.

In software engineering, LLM assistants dramatically speed up the process of writing code, leaving more time for higher order tasks such as designing system architectures. The effectiveness and efficiency of LLMs themselves is significantly improved; a focus emerges on fine-tuning LLMs on custom datasets to improve their reliability and accuracy. More generally, LLMs enable natural language interfaces to computer systems of all kinds, potentially reducing the barriers to their widespread adoption.

On the other hand, early enthusiasm for artificial intelligence as a transformational technology in areas such as biotechnology and healthcare leads to some disappointment, despite early successes like AlphaFold. It turns out that the limiting factors here are fundamental shortcomings in our understanding of how biology works, and while machine learning and laboratory automation provides useful new tools, LLMs, as a technology fundamentally based on manipulating language, provide no dramatic shortcuts to developing scientific understanding of the natural world.

What are the economic implications of this scenario? One should expect significant productivity gains, but with a delay as business processes have to be adapted to make the most of the new technology. As with any new technologies, many of the early movers go bankrupt, but the firms that survive and make the most money from the technology are those that successfully integrate the technology into wider suites of business oriented software.

Who believes in this scenario? I think this is close to a consensus view of those economists and policy makers not directly involved in the AI business. For example, a recent US National Academies Consensus Report Artificial Intelligence and the Future of Work, from a blue-ribbon committee of economists and computer scientists, co-chaired by Erik Brynjolfsson and Tom Mitchell, identifies AI as a general purpose technology with significant potential to drive productivity improvements. However, it observes that “achieving the full benefits of AI will likely require complementary investments in new skills and new organizational processes and structures.”

In this scenario, the most effective focus will probably be on technology diffusion and skills development. Technological advances in other areas will depend on continued research and development spending, with the existing scientific enterprise benefitting incrementally from machine learning techniques.

3. Crash and burn

History has featured a number of financial bubbles, in which asset prices rise beyond any seeming connection to their underlying value. Many of these are related to technological advances, in which the advent of new technologies encourages an irrational exuberance based on overoptimism, both on the speed with which the new technologies will have an impact, and on the ultimate scale of that impact. The classic recent example is the dot-com bubble of the late nineties, while one can go back in history to episodes such as the Railway Mania of the 1840s in the UK.

In this scenario the current enthusiasm for AI is revealed as one such bubble – in scale, one of the biggest in history. The bursting of that bubble exposes a scale of over-investment so large as to risk the stability of the whole financial system, while the technology itself disappoints. Ultimately, rationality returns, and the technology finds useful applications. As in the case of the dot-com bubble, the capital infrastructure installed may ultimately yield useful returns, though not to the original investors.

The initial danger signals are continuing technical difficulties limiting the reliability of large language models, and fundamental issues in the business models underpinning the very large investments being made in the computing infrastructure to support LLMs. Larger models, trained on more data, still suffer from “hallucinations” – factual statements stated with great confidence and plausibility that turn out to be incorrect. The perception that LLMs are demonstrating any kind of real intelligence turns out to be largely a case of anthropomorphism. In fact, what LLMs tell us is not how intelligent and original computers have become, but how unoriginal and derivative most human interactions are. Rather than being “stochastic parrots”, LLMs have turned out to be automated “catechisms of cliché”.

Meanwhile, for those use cases that do turn out to have some value, a panoply of rival models – many open source – destroys the pricing power of the tech giants. It turns out that there is no “moat”, no way of achieving and protecting the monopoly position that Silicon Valley tech firms aspire to. Lacking the financial returns that would justify the huge cost of building out the infrastructure for artificial intelligence, those investments are written off and valuations of the tech companies – especially the “Magnificent Seven” tech giants, which saw such a huge increase while the enthusiasm for AI persisted, collapse.

To get a sense of the scale of the bubble, the market capitalisation of the Magnificent Seven increased by about $7 trillion between January 2023 and May 2025. Microsoft, Alphabet, Amazon and Meta have reported combined capital expenditure on AI of $246bn in 2024, up from $151bn in 2023.. OpenAI’s “Stargate” project to develop AI computing infrastructure, aims to raise $100bn now, with a total target investment of $500 bn. Substantial additional investments will have come from private equity and venture capital.

The overvaluation of the Magnificent Seven, together with the excess capital investments in the private market, can be thought of as a “bezzle”, in the sense discussed by Michael Pettis here. While the bubble persists, the owners of those overpriced assets are in possession of apparent wealth that doesn’t reflect the real productive capacity of the economy, but does lead to increases in GDP through a number of channels. But the bezzle always reverses, in turn depressing GDP, as the loss of apparent wealth is distributed across the economy. The outcomes include the failure of financial institutions (sometimes subsequently bailed out at the expense of the taxpayer), people across the world losing their savings, and a freezing of investments in other areas of technology as the venture capital industry retrenches.

As the situation stabilises, useful, but not transformative, applications are found for large language models, and the massive installed infrastructure of high performance computing finds new applications in science and engineering.

This is definitely a contrarian scenario – but contrarians are not always wrong. They will be bracing themselves for financial turbulence.

Last words

These are scenarios, not predictions, and it is possible to imagine many other different possibilities. But given the constant tendency to treat technological progress as unfolding along a single fixed track, it’s important to hold on to the fact that the future is open, especially when the range of plausible outcomes seems so large.

Why productivity growth is important – Spring Statement 2025 edition

Last Wednesday saw the UK Government’s Spring Statement on the economy: essentially a response to a new forecast from the Office of Budgetary Responsibility. This was politically painful – a deteriorating forecast means that the government needs to either reduce public spending or increase taxes in order for the forecasts to meet its self-imposed constraints on public borrowing and the deficit. Those forecasts are always uncertain, and highly dependent on what assumptions are made about the future. This set of forecasts underline how important productivity growth – or the lack of it – is for the ability of the government to meet its commitments.

First, where are we now? Here is the latest productivity data. To recap, as I have been discussing for some time now, for much of the postwar period, productivity in the UK grew at a steady rate of 2.1% a year. That changed in the mid-2000s, in the run-up to the global financial crisis; the growth rate rather abruptly dropped to about 0.5% a year. The cumulative effect is that the productivity of the UK economy is now some 28% below what it would have been if earlier trend had continued.

UK Labour Productivity index, quarterly data. The solid blue line is a best fit to a function which assumes two periods of exponential growth continuous at a break point. See my post When did the UK’s productivity slowdown begin? for details of the fitting methodology. The best fit parameters are: pre-break growth rate: 2.1%, post-break growth rate: 0.48%, break at year 2005.2. Data: ONS, Feb 18th 2025 release.

Taking a more close-up look at the last twenty years, one sees that there is no sign of any recovery of productivity growth. On the contrary, the last few quarters seem discouraging. I think it is probably too early to conclude that we are moving into a period of even lower productivity growth – one needs to wait a while to see the long-term trends, and data is often revised. In addition at the moment there is a worry about quality of the survey data ONS uses to estimate total hours worked, which adds further uncertainty.

UK Labour Productivity index, quarterly data. Data: ONS, Feb 18th 2025 release, fit as in the plot above.

What is the OBR predicting for the future? It’s central forecast is based on the assumption that productivity growth increases by 2029 to a value which is more or less the average of the pre-break and post-break values – 1.25%, with an increase of 0.2% to the growth rate attributed to planning reforms leading to more house-building. It has also modelled the effect of an upside scenario, where productivity growth of 1.2% is achieved in 2025, and a downside scenario continuing recent lower trend growth of 0.3% a year.

UK Labour Productivity index, annual data, together with the OBR’s March 2025 scenarios for productivity growth up to 2029. Grey is the main March forecast, yellow & green are high and low productivity growth scenarios.

The next plot, from the OBR report, shows how much difference assumptions about productivity growth make to the fiscal projections. Without the assumption that productivity growth is about to increase significantly over the trend it’s followed for the last 15 years, the government’s fiscal targets for the debt and the deficit can’t be met. In this scenario, the government would need either to cut public spending further or increase taxes to meet those targets.

Effect of the different productivity growth scenarios on the OBRs fiscal forecasts. From the March 2025 Economic and Fiscal Outlook.

So, big decisions about the UK’s fiscal policy are being made on the basis of predictions of future productivity growth, and the question of whether the government will meet its fiscal targets are very sensitive to the assumptions being made. How much credence can we place on these forecasts? It does have to be stressed that the OBRs record on predicting productivity growth has been consistently overoptimistic since its establishment, as the plot below shows. I think this reflects a wider failure by the UK’s economic and policy-making community to appreciate the scale of the change to the UK economy in the 1990s and 2000s and its long term effects. There is no more important economic policy question than to understand the causes of the productivity slowdown – and to find policies to reverse this.

Successive OBR forecasts of productivity and outturn. Source: OBR

The end of wage growth in the UK

I’ve been writing about the UK’s slowdown in productivity growth for about a decade, as I discussed here. I think it’s fair to say that this issue is well-understood amongst economists and some policy people, but productivity is an abstract concept. So, it’s perhaps unsurprising that, even now, the seriousness of our economic situation isn’t fully understood by commentators and journalists, let alone the wider public.

But there’s one way in which our productivity slowdown has very visible everyday consequences – and that’s in the end of wage growth. As my plot shows, wages have flatlined in the UK over last 15 years. This long period of stagnation is unprecedented in living memory, & marks a decisive & unwelcome break from the UK’s postwar economic trajectory.

Average real weekly UK wages. Green: Composite Average Weekly Earnings series, corrected for inflation using consumer prices index. Thomas, R and Dimsdale, N (2017) “A Millennium of UK Data”, Bank of England OBRA dataset. Brown: ONS, Real Average Weekly Earnings, total pay, using CPI (seasonally adjusted). 18/2/2025 release.

The period from the end of the Second World War right up to the mid 2000s shows a remarkably consistent record of wage growth. There are moments of economic turbulence that are reflected in deviations from the trend of continuous 2.8% pa growth; a short-lived period of more rapid growth in the late 60s and early 70s – the Barber boom – with the excess growth unwinding in the mid-1970s crisis. And again, more rapid growth in the late 1980s Lawson boom, with the excess gains lost in weaker wage growth in the subsequent recession.

But nothing compares to the stagnation that we’ve seen since the global financial crisis. By the economic measure that arguably matters most to people at large – how their wages grow – the last decade and a half is by far the worst period since the war. In comparison, the economic turbulence of the 1970’s looks like a golden age.

UK labour productivity, index 2022=100. Data: ONS, 15/11/2024 release. Line: non-linear least squares fit to two exponential functions, continuous at the break point, which occurs at 2005 for the best fit. See When did the UK’s productivity slowdown begin? for more details of the fitting approach.

The end of wage growth in the UK is a direct consequence of the end of productivity growth. It’s worth making a couple of points about the link between productivity growth and wage growth. In the USA, that link is weaker than it was. But the UK is not the USA; while in the USA the labour share of GDP – the share of overall economic activity that goes to wages, rather than rewarding the owners of capital – has significantly fallen, this is not so in the UK. For whatever reason, in the UK, over the last decade, the labour share of GDP has actually increased.

Of course, my plot of wage growth presents a single average, and it’s a fair question to ask how the distribution of wages has changed with time – has this become more unequal, with more of the benefits of productivity growth going to higher earners? It turns out that, while there was a substantial increase in inequality in the 1980s, overall measures of income inequality have been relatively steady since then.

The wage growth plot explains so much about state of UK politics today. Few people have an intuitive feel in the abstract for what productivity growth – or its absence – means, but the sense of stalling living standards, and worse prospects for young people, is all too palpable.

The world of business R&D (and the UK’s place in that world)

Most research and development (R&D) in the world is done not in universities or research institutes, but by businesses – big businesses can do more R&D than medium size countries. A useful snapshot of this world is provided by the 2024 EU Industrial R&D Investment Scoreboard, which came out in December. The scoreboard lists the top 2000 companies in the world by their annual R&D expenditure, classifying them by sector and nationality of headquarters. In total, this amounts to total R&D spend of €1257.7 billion (converted at market rates), which the authors believe accounts for 85% to 90% of worldwide R&D funded by the business enterprise sector.

Unsurprisingly, the top companies are US tech firms – Alphabet, Meta, Apple and Microsoft – which between them spend €127 billion. Number 5 is the German auto firm Volkswagen, Asia provides numbers 6 and 7 in the shape of China’s Huawei and Korea’s Samsung.

Taking the world as a whole, the top sectors are Software, accounting for 19% of the total, Pharma at 18%, Automobiles at 15%. Tech hardware accounts for 16% and Electronic & Electrical hardware another 7%. These last two categories do have some overlap – the former includes Apple, Huawei, Intel, Qualcomm, Nvidia, Cisco and TSMC, while the latter includes Samsung, Siemens and Hon Hai (aka Foxconn).

How does the UK do? The share of world business R&D done by UK domiciled firms is 2.8%, and there are just two UK companies in the top 100 – the pharmaceutical companies AstraZeneca and GSK.

Of course, where a company is domiciled and where it does its R&D aren’t necessarily the same. Roughly half of UK business R&D is done by overseas owned companies – for example, the significant R&D carried out in the UK by the auto company Jaguar LandRover is ascribed in these statistics to its Indian parent, Tata Motors. This is a very high fraction of R&D done by overseas firms, by comparison with other countries of a similar size. The positive interpretation of this is that it is a testament to the attractiveness of the UK as a place to do R&D. But control matters, and this exposes the UK to the risk that this R&D may be more footloose than R&D done my domestically owned firms.

We can get a sense of the sectors that the UK focuses on by comparing the UK shares with the global fraction.

Pharmaceuticals is a clear leader for the UK – it accounts for 49% of the UK owned business R&D, which amounts to 7.5% of the world total. There is an interesting aspect to this, however – it is completely dominated by the two giants, AstraZeneca and GSK. This is in contrast to the USA, where there is a significant tier of relatively recently founded companies that have emerged from the biotech revolution – such as Gilead, Amgen, Moderna, Regeneron and Vertex, all with € multibillion R&D spend. UK pharma scale-ups – like Bicycle Therapeutics and Immunocore – are still an order of magnitude smaller.

The other area of specialism for the UK is Banking – this accounts for 17% of the UK’s R&D; this represents 41% of the world R&D in this sector. Of course, there may be issues of what is classified as R&D in different companies.

Where UK firms are largely absent is in Software, Tech hardware and Electronic & Electrical hardware. Between them, these sectors dominate global business R&D, accounting for 42% of all business R&D. But the UK accounts for just 0.6% of world Software R&D, 0.45% in Electronic & Electrical hardware, and a tiny 0.046% of world R&D in Tech hardware. Once again, this doesn’t take into account of R&D carried out in the UK by overseas firms – for example, DeepMind’s work will be ascribed to its US owner, Alphabet. But it does suggest that the UK has largely missed out on innovation in the fastest moving areas of new technology in its domestically owned firms.

Finally, one might ask how effective markets are at allocating resources to the areas where the need for innovation is greatest. Given the urgency of climate change, and the need for innovation to drive down the costs of low carbon energy, it’s depressing to see that business R&D in the Alternative Energy sector accounts for just 0.23% of the world total, with Oil and Gas still accounting for 1.05%.

Ten years of banging on about productivity

An advantage of having a long-running blog like this one, which is now twenty years old, is that it gives me an easy way of checking out some of the topics that were most exercising me at various points in the past. But it’s still a bit of a shock to realise that it’s been 10 years since I started banging on about the dramatic fall in the UK’s productivity growth rate. My plot emphasises that nothing has happened in the last ten years to change that dismal trend.

UK labour productivity, index 2022=100. Data: ONS, 15/11/2024 release. Line: non-linear least squares fit to two exponential functions, continuous at the break point, which occurs at 2005 for the best fit. See When did the UK’s productivity slowdown begin? for more details of the fitting approach.

The consensus at the time was that the problem was likely to be self-correcting, and that one could expect a resumption of the earlier trend of productivity growth (if not a recovery of lost ground). Nothing illustrates this better than the successive predictions of the Office of Budgetary Responsibility, which took until 2017 to realise that a new dawn for productivity growth might not be immediately around the corner.

Successive estimates, made between 2010 and 2017 by the Office of Budgetary Responsibility, of future productivity growth, as reported in the November 2017 Economic and Fiscal Outlook

In discussing productivity, my initial focus was on research and development. I’d already drawn attention to the long fall in the UK’s business R&D intensity, and in my June 2014 piece Business R&D is the weak link in the UK’s innovation system I connected this with two of the UK’s problems – its slump in productivity growth, and its persistent current account deficit.

In a follow-up piece, Rebuilding the UK’s innovation economy, I linked low business R&D to the wider problem of short-termism, as identified in the 2012 Kay Review of UK Equity Markets and Long-Term Decision Making, writing that “our shrinking R&D base is part of a bigger problem of short-termism, in which the structures of our capital markets and the reward structures for company managers excessively reward good financial performance in the present at the expense of longer term prospects for growth”.

What did I think ought to be done about it? What I was clear wouldn’t work by itself is what I call “supply side innovation policy” – the idea that if one supports basic science in universities and provides a supply of skilled people, that will automatically lead to economically significant private sector innovation: “But there’s no point driving [university based] scientists to be more collaborative with applied researchers in the private sector, if the private sector doesn’t have the R&D capacity to collaborate with”.

Instead, I argued that we needed to build that capacity, in key areas like health related research and energy: To be clear, here I’m not primarily talking about academic science; what is needed is directed R&D focused on delivering products. For some of these products – for pharmaceutical and medical innovations – the government will be the main customer, as well as being the direct financial beneficiary of savings in areas like the social care budget.

For energy, I argued that the emphasis should be on driving costs down for low carbon energy, including nuclear: “In the case of energy, the costs will be imposed on future customers through long-term guaranteed prices. For example, the Hinckley Point deal, for just one nuclear power station, will result in the transfer of several tens of billions of pounds from domestic and business electricity users to the overseas providers of the technology and finance. Instead of simply standing back and paying these bills (or imposing them on future taxpayers and customers),the government should use its power as the purchaser or guarantor to make sure new technologies are developed for the which the UK can capture significant value.”

To understand the causes of the productivity slowdown more fully, one needs to dive into a more detailed analysis of productivity performance in different sectors. My January 2015 piece,
Growth, technological innovation, and the British productivity crisis, took a first look at this, using early estimates from the ONS for total factor productivity by sector. Here there was a contrast between steady growth in ICT and manufacturing, and peaks – and subsequent falls – in two broad sectors: Agriculture, forestry & fishing, mining & quarrying, utilities, and Financial and Insurance. For the sectors including mining and quarrying, the peak was at 2003, and I believe that this largely reflected the output of the oil and gas sector – production of North Sea oil peaked around 2000. For the sectors including financial services, the peak was at 2007, so unsurprisingly correlated with the onset of the global financial crisis. My conclusion was essentially that the UK had been a victim of what economists call “Dutch disease”, following the peaking of North Sea oil and the bursting of a financial services bubble: “Manufacturing and ICT have been squeezed out by the apparently greater, but ultimately unsustainable, returns from oil and finance, and when those wells dried up the economy was left stranded.”

I still believe that Dutch disease is a helpful, though partial, lens to look at our productivity slowdown through. But since then we’ve learnt a lot more about the proximate causes of the UK’s sluggish productivity growth, and the sectors that have most contributed to the slowdown. A good overview can be found in The Productivity Agenda, from the University of Manchester based Productivity Institute (including a section by me on R&D).

For example, careful econometric analysis has identified that the key contributors to the slowdown, in sectoral terms, have been transport equipment, pharmaceuticals, computer software and telecommunications. This is counterintuitive, in that these sectors are generally thought of as strengths of the UK economy.

More generally, there seems to be a consensus that low levels of investment – in both the public and private sectors – is a major proximate cause of low productivity growth. I would include in this low overall investment record, the low levels of business R&D that first attracted my attention, more than a decade ago. Since it is widely accepted that technological innovation provides the basis for productivity growth, and R&D provides a formal structure for developing and implementing technological innovations, I would still insist on its importance.

But this doesn’t not yet answer the question of what it is about the UK’s political and economic system that has led to this state of chronic underinvestment, or what should be done to address that. The consequences, though, are clear – slow productivity growth has led to stagnating living standards and difficulties in funding public services at the level people expect. For my last words here, I’ll quote what I wrote in July 2014:

“The erosion of the UK’s capacity to technologically innovate was not inevitable – it was the unintended consequence of a series of political and policy choices over decades. We need to reverse this loss of capacity. This needs to be done as part of a broader rethinking of the variety of capitalism the UK economy is currently based on. Without this rethinking, we will be condemned to continue on our current trajectory of low growth, poor trade performance and ultimately, loss of national sovereignty.”

Taking Anglofuturism seriously

Regular readers of this blog won’t need reminding that the UK is in a stagnant bind, with economic measures like productivity and GDP per person flatlining since the global financial crisis (or earlier). The consequences are felt well beyond these arid economic aggregates; wage growth has slowed down, successive governments find it hard to fund acceptable public services, and there’s a palpable sour sense of malaise in our politics.

One interesting response to this has been the emergence of a loose constellation of commentators, activists and pressure groups, a techno-optimist movement calling for more houses to be built, for the barriers apparently stopping the country building infrastructure to be swept away, for cheaper and more abundant energy.

Britain Remade wants to “reform the planning process to deliver more clean energy projects, transport infrastructure, and new good quality housing at speed”, while the yimby Alliance , as keen subscribers to the “Housing theory of everything”, focus on the need to build more houses. A very widely talked about paper, Foundations: Why Britain has stagnated focuses on housing, infrastructure, and the cost of energy. Rian Chad Whitton likewise focuses on high energy prices, connecting this with the decline of the UK’s manufacturing base.UKDayOne focus on science, innovation and technology as the motor for UK growth and prosperity, particularly emphasising AI and nuclear power.

I’m going to follow Tom Ough and Calum Drysdale in gathering these strands together under the banner “Anglofuturism”. Their eponymous, and interesting, podcast embraces a cheerful and optimistic version of this vision, with its whimsical AI generated illustrations of flying pubs and thatched space stations.

But I believe the term (in its current manifestation, at least) was coined by the journalist Aris Roussinos, in rather darker hues. This was a call for rebuilt state capacity in a definitively post-liberal world, a vision that owed less to Adam Smith, and more to Thomas Hobbes, which some readers might think more appropriate to deteriorating geopolitical situation we face.

I don’t think there is an entirely consistent underlying political ideology here, but I think it’s fair to say that there’s a common centre of gravity on the centre right. This isn’t the place to analyse political antecedents or implications, and I’m not the right person to do that, but I do want to make some remarks about this emerging movement.

There is much in this agenda that I applaud and agree with. The UK needs to get back to productivity growth, and there is no fundamental reason why that shouldn’t happen. We haven’t reached some final technological barrier – far from it. And I think there’s a profoundly humanistic perspective at work here – people should be able to enjoy the fruits of prosperity.

Of course, there is an opposing argument that believes that continued economic growth is inconsistent with planetary limits. It’s clear that we need to move to a new model of economic growth that doesn’t impose externalities on the global environment, and in particular we need to shift our energy economy to one that doesn’t depend on fossil fuels. But to embrace “degrowth” is in my view both politically infeasible and, if sufficient will and resources are applied, technologically unnecessary. To put it another way, the last 15 years in the UK have been an experiment in degrowth, and the results have been ugly.

There’s an undercurrent of generational justice here too. The perception that young people in the UK can’t look forward to the same lifestyle as their parents is profoundly depressing. Nowhere is this more obvious than in the unaffordability of housing.

Where I think these analyses are less convincing is in identifying the origins of our current problems. In particular, I think an explanation of our current productivity stagnation needs to account for its timing. It’s certainly convincing to argue, as these authors do, that we would be better off if the UK had built more infrastructure over the last few decades, but I don’t think they really convince in talking about what conditions would have produced that outcome. Anglofuturism, in all its varieties, could be accused of willing worthy ends, without really specifying the means.


Labour productivity in the UK since the Industrial Revolution. Data from the Bank of England A millennium of macroeconomic data dataset, plot & fits by the author.

The Foundations paper puts a lot of blame on the 1947 Town and Country Planning Act – and the wider Attlee settlement. But I don’t think this makes sense in terms of the timing. As my figure shows, the period of fastest productivity growth in the entire history of the UK took place between 1948 and 1972. In fact, Roussinos harks back to this period, referring to “the optimism and high modernism of the post-war era, a vanished world of frenetic housebuilding and technological innovation where British scientific research could lead the world, and produce higher living standards through its fusion with well-paid, high-skilled labour.”


Labour productivity in the UK since 1970. ONS data, fit by the author. For the rationale for putting the break around 2005, see When did the UK’s productivity slowdown begin?

What needs to be explained is that the current slowdown began in the mid-2000s. There is some overlap with a developing consensus view from mainstream economics that the immediate problem has been a lack of investment in the UK economy (see e.g. The Productivity Agenda). This includes public investment in hard infrastructure, private investment in capital goods, and investment in intangibles like R&D. In my own work I’ve emphasised the significant reduction in the R&D intensity of the UK economy between 1980 and 2005, and given the generally technocentric flavour of the Anglofuturists, I’m surprised that this aspect isn’t more prominent in their arguments.


From Research, innovation and the R&D landscape, by R.A.L. Jones, in The Productivity Agenda.

Even if one agrees that investment levels have been too low, there isn’t really a consensus about the ultimate cause of the lack of investment. One common thread is a sense that building infrastructure in the UK has become too expensive because of excessive regulation. In one sense, this is a reflection of the fact that the comparative advantage of the UK is to be found in professional services. One can celebrate that fact that the UK has become a “services superpower”, but the downside was caustically expressed in this comment from Dan Davies

Giles Wilkes has discussed what he terms the “crud economy” at a bit more length. Economic actors respond to incentives, and this doesn’t always direct activity towards where we need it. As Giles puts it: “We need vastly more clean energy, actual hard defence equipment for handling nasty rogue nations, the soldiers to use it, and much more numerous and productive care and health workers for the ageing population. Mitigating the dangerous effects of climate change is going to take real physical capital and effort. These are actual hard problems – and being able to produce more streaming videos, intelligent AI-related chat, or brilliant legal ‘solutions’ to financial market problems is not exchangeable for the assets we need for the real problems. Just because the lawyer’s fee is expressed in dollars, and so is the cost of transforming the US electricity system, doesn’t mean the two can get traded together.”

One thing all branches of Anglofuturism agree on is the need for abundant, cheap energy, and on the bad economic effects that current high industrial energy prices are causing. This clearly causes strong feelings, to judge by the violent on-line reaction to Tom Forth’s entirely reasonable, from a classical market liberal perspective, comments about this, arguing that, while this situation was not good, it was “a smaller problem and of a lower priority than many other restrictions on growth in Britain.”

I agree that it would be better if energy prices in the UK were lower, but I think it is important to understand how this situation has arisen. High industrial energy prices now are causing serious problems for what industry remains in the UK, but I don’t think they can be blamed for the UK’s greater degree of deindustrialisation that its neighbours. This took place at a time when energy prices were low and falling.

The decision the UK government made in the 1980s was that energy was just another commodity whose supply could be left to the market. As it happened, this coincided with a moment in time when the UK switched from being a net importer of energy, to being a net exporter, having found abundant supplies of natural gas and oil in the North Sea. North Sea oil and gas production peaked around 2000, and the country switched to being an energy importer again in 2004. The UK’s relative success in decarbonising its electricity supply initially relied on an early switch from coal to gas; even after the more recent expansion of offshore wind the price of electricity is set by the internationally traded price of gas. This was fine until it wasn’t – in the 2022 gas price spike.

If our problem is that we rely on imported gas, whose fluctuating price is beyond our control, together with offshore wind, which is necessarily intermittent (as well as being generated a long way from where it is needed, connected by an inadequate grid), would it not be better if a much higher proportion of our energy was generated by nuclear fission?

An enthusiasm for nuclear power is a common thread running through all strands of Anglofuturism, and it’s one with which I have much sympathy. For all the progress there’s been in renewable energy, in 2022, 77.8% of our energy still came from oil, gas and coal, and I think it’s going to be difficult to have a fossil-fuel free energy economy which doesn’t depend on some nuclear power to provide firm energy . I deeply regret the failure of the nuclear new build programme of recent governments – of the 18 GW of new generating capacity planned in 2014, only 3.2 GW is even under construction.

But I think it is important, and salutary, to understand why this failure has occurred. My recent blog posts go into the story of the UK’s civil nuclear power programme in some detail . There are ways in which the regulatory and planning framework for civil nuclear could be streamlined, but the fundamental problem with Hinkley C wasn’t the fish disco. It was the fact that the UK government wanted the Chinese state to pay for it, and the French state to build it, as the UK state no longer had the will or capacity to do either.

The UK’s own civil nuclear industry was killed in the 1990s; in an environment of high interest rates and low natural gas prices, and an ideological commitment to leave energy supply to the market, there was no place for it. I do think the UK should recreate its capacity to build nuclear power stations, including the small modular reactors that are currently attracting much attention, but I don’t think this will happen without substantial state intervention.

I agree with the Anglofuturists that we shouldn’t resign ourselves to our current economic failures. I think we need to ask ourselves what has gone wrong with the variety of capitalism that we have, that has led us to this stagnation. It’s a problem that’s not unique to the UK, but which seems to have affected the UK more seriously than most other developed countries. The slowdown seems to have begun in the 2000s, crystallising in full at the Global Financial Crisis.

This timing points to changes in the nature of capitalism and political economy that took hold in the decades after 1980, with the ascendancy of
market liberalism, the doctrine of shareholder value in corporate management, and an enthusiasm for outsourcing government functions to private contractors, no matter how central to the core purposes of the state they might appear to be. In the UK, even the Atom Weapons Establishment has been run by private contractors since 1989, with the government only taking ownership and control back from SERCO in 2020.

We have a new form of globalisation that followed from abolishing capital controls, together with a conviction that one doesn’t need to worry about the balance of payments, even though the persistent trade deficits the UK has run since then has meant ownership and control of national assets has moved overseas. We have a financial system that seems unable to direct resources to those activities that lead to long-term growth. We have a hollowed out state, that now lacks the capacity even to be an informed and effective contractor for services.

I agree with the Anglofuturists that our current stagnation isn’t inevitable, and I applaud their lack of defeatism. It doesn’t have to be this way – but to get beyond our current malaise, I think we need to ask some deeper questions about how our economy is run.