Can Lego save the world? That’s one idea that stuck with me reading How Big Things Get Done, a new book by Bent Flyvbjerg and Dan Gardner. Flyvbjerg is perhaps the world’s leading authority on the failure of megaprojects — or how big things get done, but woefully late and at woeful cost — and so he makes an unlikely optimist.
Over the decades, Flyvbjerg, a management professor at Oxford university, has assembled a database of large projects from high-speed rail lines to hosting the Olympics. His findings are so dismal that he has proposed the “Iron Law of Megaprojects”: they are over time and over budget, over and over again. Even worse, there is a long tail to these disappointments. A significant minority of megaprojects are not just late and expensive, but catastrophically so.
Despite this gloomy evidence, he and Gardner make the case that we could work wonders if instead we used a principle most familiar from Lego sets. That principle is modularity: a complex Lego model is assembled from a limited range of bricks, each of which is precision-manufactured and interchangeable with other bricks.
Modularity has a number of advantages. The first is that the individual components can be manufactured at scale, which rapidly reduces costs. In the 1930s, an American aeronautical engineer named TP Wright made a careful study of aeroplane factories. He concluded that the more often a particular model of aeroplane was assembled, the quicker and cheaper the next plane became. The workers learnt the best ways of working, and special tools would be developed to assist with particular tasks. Wright found that the second plane would typically be 15 per cent cheaper than the first. The fourth would be 15 per cent cheaper than the second, and the eighth plane 15 per cent cheaper again. Every time accumulated production doubled, unit costs would fall by 15 per cent. Wright called this phenomenon “the learning curve”.
Later researchers have found learning curves in more than 50 products from transistors to beer. Sometimes the learning curve is shallow and sometimes it is steep, but it always seems to be there. And because modular projects repeatedly use the same plans and structures, they harness the efficiency of the learning curve.
There are other advantages to modular projects. They are more likely to be able to use factory-made components, and when you make complex things in factories, you are less at the whim of the unexpected than when you make them on a building site — especially if that building site is deep underground or offshore.
By their nature, modular construction projects are more likely to be able to keep going even when there is a problem with one element of the structure. This helps explain why, in Flyvbjerg’s database, modular projects are all but immune to the most dramatic “black swan” cost overruns, which are always a risk for other large projects.
Such are the joys of modularity. Now turn to the problem of climate change, and an intriguing pattern emerges. Low-carbon energy projects include some of the most modular and the least modular designs in Flyvbjerg’s database. Solar and wind are at the modular end, while nuclear and hydroelectric projects are at the opposite pole. Perhaps no surprise, then, that solar and wind projects are rapidly falling in price.
I have no objection in principle to nuclear power, but I wonder whether it will ever be possible to make clean, safe nuclear power at a reasonable cost, unless nuclear plants are able to switch to a much smaller, more modular design. Nuclear power stations have been supplying power to the grid since the mid-1950s, but they never seem to get much cheaper, perhaps because we have been unable to repeat the same designs often enough to climb the learning curve. I keep reading news stories about companies with big plans for small reactors, so perhaps it is not impossible.
Still, the contrast with solar energy is striking. Silicon photovoltaic cells started supplying practical power around the same time: the American satellite Vanguard 1 was the first to use them, carrying six solar panels into orbit in 1958. (The sun always shines in space, and what else are you going to use to power a multimillion-dollar satellite?) At the time, those solar panels produced half a watt at what was no doubt a painfully high cost.
By the mid-1970s, solar panels were down to $100 a watt, or $10,000 for enough panels to power a lightbulb. By 2021, the cost was less than 27 cents a watt. Why? This is the learning curve in action. The learning curve for photovoltaic cells has been estimated to be 20 per cent per doubling — steeper than for aeroplanes.
Chris Goodall, author of The Switch, notes that the world produced 100 times more solar cells between 2010 and 2016 than it had in all the decades before 2010. Batteries — an important modular complement to solar PV cells — are also racing down a steep learning curve. There is a similar story to be told about wind power. Wind turbines are made of standardised components, and a wind farm is made of standardised turbines. The price of wind power, too, has fallen faster than most proponents could have dreamt two or three decades ago.
I am no expert on nuclear energy, but I am assured that modular reactors should be possible. I hope so. We need big things to happen in our ability to generate clean energy. And the best way to go big is to start with small, repeatable blocks.
Written for and first published in the Financial Times on 27 January 2023.
My first children’s book, The Truth Detective is out on 15 March (not US or Canada yet – sorry).