There is a particular kind of person (you have met him, possibly you have hired him) who believes that the way to make a place valuable is to make it nice. Knock down the ugly bits. Pour concrete. Build something rectangular and well-ventilated with a name like "InnovateHub" and an atrium you could land a helicopter in. This person is an engineer of the soul, and he is almost always wrong.
I have been thinking about King's Cross, because King's Cross is the most instructive thing to happen to British real estate in fifty years, and almost nobody has drawn the correct lessons from it. The correct lessons are all counter-intuitive, which is precisely why they are correct. If the lessons were intuitive, the spreadsheet people would already have acted on them, and there would be no edge left to exploit.
Lesson one: the dereliction was an asset, not a liability
For most of living memory, the land behind King's Cross and St Pancras was a wasteland of gas-holders, coal sheds and railway sidings, with a reputation for vice that made estate agents whisper. By every metric a property developer holds dear, it was a problem to be solved.
But consider what cheapness and a bad reputation actually do. They keep out the timid. They keep out the corporate. They lower the cost of being there, which means they lower the cost of experimentation, and the single most valuable input to any creative ecosystem is the freedom to do something that might not work. When Central Saint Martins moved its art school into an old granary in 2011, it was not despite the grime. The grime was the point. You cannot put four and a half thousand art students into a glass-and-steel business park, because a glass-and-steel business park communicates, in every fibre, that mistakes are not welcome here.
This is the first piece of psycho-logic the engineers never grasp: a place that looks finished is a place where nothing new can happen. Slightly broken is a feature.
Lesson two: they kept the useless things, and the useless things became the prize
Here is what a rational developer would have done with a set of Victorian gas-holders: scrapped them. They hold no gas. They are giant iron skeletons of an obsolete technology, the literal definition of a stranded asset.
Instead they kept the gas-holder frames and put flats and a park inside them. They kept the coal drops, the dirtiest, most industrial structure imaginable, a place built solely to dump coal off trains, and turned them into Coal Drops Yard, where you may now buy a flat white for the price of a small Victorian funeral.
The reductive view says: this is inefficient nostalgia. The correct view is that the old ironwork is doing an enormous amount of psychological work that no new building could do for any money. It signals permanence, story, costliness-that-cannot-be-faked. You cannot build a hundred and fifty years of patina; you can only inherit it. Economists have a word for goods whose value rises because they are hard to produce, Veblen goods, and a regenerated coal depot is a Veblen good the size of a postcode. The "useless" heritage is the most valuable thing on the site, for the simple reason that it is the one thing a competitor with a bigger budget cannot replicate.
Lesson three: the most digital industry on earth has chosen to behave like a medieval guild
Now we arrive at the genuinely strange part. The companies pouring into King's Cross are AI labs, Google DeepMind, and lately Meta, and now Anthropic taking room for hundreds, and OpenAI signing for a permanent London home, and a procession of others leasing more than a million square feet since 2025.
These are the most location-independent businesses that have ever existed. Their product is literally the abolition of distance. They could, by their own logic, headquarter on a fishing trawler. And what do they do? They cram themselves into a single one-mile radius around a railway station, paying handsomely to sit within walking distance of the very rivals trying to poach their staff.
The death-of-distance prophets of the 1990s told us the internet would dissolve geography. They had it exactly backwards. The more virtual your work becomes, the more ferociously you crave physical proximity, because when your only scarce resource is a few thousand brilliant minds, you need them bumping into each other in a coffee queue, defecting between labs, gossiping at the same lunch spot. Agglomeration isn't a bug the AI industry has failed to optimise away. It is the optimisation. The cleverest people on earth have independently rediscovered the logic of the medieval guild quarter, where all the goldsmiths lived on Goldsmith Street, and for the same reason: talent is contagious, and contagion needs contact.
And note the lovely detail the brokers report, these firms are not, in fact, paying a premium over the wider market. The eye-watering rents reflect a shortage of good space, not AI money behaving stupidly. The premium isn't in the rent. The premium is in the postcode on the email signature, which is a much cheaper thing to buy and a much dearer thing to possess.
Lesson four: they renamed it, and the name did the work
My favourite intervention in the whole story is the one that cost almost nothing. Somebody decided to call the area around King's Cross, Euston and Bloomsbury "the Knowledge Quarter."
This is alchemy in the purest sense, a change in perceived reality achieved entirely through language. The British Library did not move. UCL did not move. The Francis Crick Institute was always going to be a world-class research lab whatever the road signs said. But by drawing a circle on a map and giving it a confident, slightly grand name, you change what the place means to the person deciding where to put their lab. You convert a collection of separate institutions into a scene, and people will pay, relocate and queue to be part of a scene in a way they never will to be near a list of buildings.
We routinely spend billions on steel and glass and dig tunnels under the Chilterns to shave twenty minutes off a train, while sneering at the marketing department that solves a harder problem with two words and a logo. This is the great blind spot of the technical mind: it trusts only the interventions that are expensive and tangible, and dismisses the ones that are cheap and psychological, which are, with maddening regularity, the ones that actually work.
The thing nobody planned
If you want the perfect emblem of all this, walk into King's Cross station and find the queue of tourists photographing themselves at Platform 9¾, pushing a luggage trolley into a wall for a train that does not exist, to a school that does not exist, in a book. A fictional platform is among the station's most reliable generators of real footfall and real affection. No transport planner put that in the business case. You could not have.
Which is rather the moral of the whole place. King's Cross did not become London's answer to Silicon Valley by being engineered into it. It got there by keeping the things a rational analyst would have scrapped, leaving room for the things a rational analyst could not predict, and understanding, as the spreadsheet never will, that the value of a place lives almost entirely in the heads of the people looking at it.
The opposite of a good idea, as I am fond of saying, can be another good idea. The opposite of "demolish the derelict coal depot" turned out to be a very good idea indeed. We should try having more of them.
