The usual version of the abundance story goes like this. AI gets good. Companies build robots. The robots are expensive and narrow at first, so only the big players can afford them, and those players pull further ahead while everybody else has a difficult decade. Then, at some point, the robots start building robots. Curve goes vertical. Everything gets cheap. Prices fall, not because money became scarce but because supply became absurd, and we all sit around trying to work out what to do with ourselves.
I have told a version of this myself. And I now think the interesting question inside it is the one people skip past fastest, which is: how long is the difficult decade, and what exactly is holding it up?
Because it is not the intelligence. Design and compute will be cheap and early. It is not even the self-replication.
It is copper.
The bottleneck moves from the mind to the ground
A self-replicating factory still needs copper, lithium, neodymium, gallium, and an amount of electricity that has to come from somewhere physical. So the moment robots can build robots, the constraint does not disappear. It simply relocates, from intelligence to geology, permitting and grid capacity. And those three things have one lovely property in common: none of them respond to exponentials. A copper mine takes fifteen to twenty years from discovery to first pour. You cannot prompt engineer a resource consent.
Do one thing, just run the arithmetic yourself and see how it feels. Start with a global fleet of a hundred thousand capable units. Double it every eighteen months, which is a heroic assumption for anything involving steel. You reach a billion units in about twenty years. Halve the doubling time to nine months, be as optimistic as you like, and you still need seven or eight.
So the suffering window is a decade or two. Not a season. Not “a rough couple of years while things adjust.”
And a decade is long enough for something specific and permanent to happen, which is the actual point of this post.
The real chaos mechanism is debt, not unemployment
The deflation part of the story is usually told too gently. Prices fall, things get cheap, lovely.
Except debt is nominal and does not deflate. Mortgages, sovereign bonds, corporate leverage, all of it gets heavier in real terms at exactly the moment wage income is disappearing. Irving Fisher wrote this up in 1933 and it has not become less true since. The chaos in the chaos decade is not primarily people having no job. It is people having no job while the number on the mortgage stays exactly where it was.
The same is true of the price falls also. They will not arrive evenly. Anything that can be manufactured or copied goes towards zero. Anything that cannot be manufactured stays expensive and gets worse: land, location, attention, originals, the good school zone. Vanilla life becomes extraordinarily cheap in calories and gadgets and stays brutally expensive in address.
Which means the fight quietly migrates from income to land, and mostly nobody announces it.
Nobody shared the surplus out of kindness
Here is the part that should worry you most, and it is historical rather than speculative.
Surplus got shared in the twentieth century because elites needed two things from ordinary people. Their labour, and their bodies in uniform. Mass production and mass conscription are what bought us the welfare state, the eight hour day and the pension. It was leverage, dressed up afterwards as decency.
Robots remove both levers at roughly the same time.
So the universal token that everyone assumes is coming has no structural reason to be generous. It will be exactly as generous as the moment it is negotiated in, and not one dollar more. And taxing the robot owners requires a state that can still enforce against them, at precisely the moment enforcement is also getting automated.
Metal dissolves, politics sets
So which one actually binds, the metal or the politics?
Metal genuinely dissolves, and this is the optimistic half. It is a stock problem with a known clock and several exits. Aluminium substitutes for copper in windings. Sodium substitutes for lithium in stationary storage. And most importantly, every robot ever built is ore permanently lifted above ground, so once the installed fleet is big enough, recycling beats mining and the constraint quietly stops mattering. Nobody has to agree to anything for that to happen. It is metallurgy and time.
Politics does not dissolve. It resolves, into a settlement, and then it sets like concrete. There is no mechanism by which an ownership arrangement gradually stops binding on its own. Somebody has to lose an argument.
So they are not two parallel constraints. They are sequential, and the first one causes the second. The mineral shortage is what makes early robots scarce and expensive, which is what concentrates ownership in whoever can afford the first units, which is what hands that group twenty years of compounding surplus with which to purchase the rules. By the time metal stops binding, the political question has already been answered, by people whom the shortage made very rich.
The scarcity phase is not a delay before the distribution fight. It is the distribution fight, conducted quietly, before anyone has announced that it started.
The motivation problem is actually a coverage problem
The other thing people get wrong is motivation. If nobody needs to work, why would anyone work?
Except we already have the answer and it is sitting in a shed in every town in New Zealand. Men’s Shed. Volunteer fire brigades. People who show up on a Saturday for no money at all and do beautiful work.
But look closely at what those people are doing. They chose the task. They set the hours. They can walk out. And the work is visibly useful to someone standing nearby. That is a narrow and very pleasant slice of the work that will still need humans.
The residual jobs in an abundant economy are the ones robots cannot take yet. Night shift in aged care. The awkward judgement call nobody wants to sign. The sewer that has to be crawled at two in the morning, in winter, in the rain. No shed has ever formed around that. I have not once seen a volunteer roster for it either.
So it is not a motivation problem. It is a coverage problem. Intrinsic drive produces wonderful work, distributed completely unevenly across the tasks that actually need doing. My guess is those roles get bought back at genuinely startling rates, because the token buys vanilla and only the leftover buys status, and status will be the last currency with any bite left in it. The sewer engineer of that era does not earn a wage. She earns standing.
And on the reproduction question, I think the fear is upside down. Every society that has become rich has undershot replacement without anybody needing a licence. The abundance problem is almost certainly too few children, not too many, and the binding constraint is energy and waste heat rather than headcount. I would bet on enormous payments to encourage births long before I would bet on permits to allow them.
What to watch
Stop watching the robot demos. Everybody is watching the robot demos, which is precisely why they contain no information.
Watch the boring instruments instead. Grid connection queues and interconnection priority. Mining and resource consent decisions, and how long they take. Copper and the aluminium substitution rate. Recycling economics, because the day urban mining beats digging is the day the metal constraint starts dying. Data centre power contracts. And the least glamorous indicator of all, which is who is quietly acquiring land next to substations.
Those decisions are being made right now, in public, by people who believe they are doing infrastructure policy. They are actually drafting the constitution of the next economy and they have no idea.
Where to invest, and I am nobody’s financial adviser
The general principle falls out of the argument rather than from any cleverness on my part. Manufactured things go to zero, so do not own the manufactured thing. Own the constraint.
That points at a handful of categories. Generation and grid, because electricity is the one input nothing substitutes for. Land and location, especially industrial land with power access already attached. Recycling and materials processing, which is the boring end of the boring end. Substitution technology, the aluminium and sodium side of the story. And ownership stakes in the automation itself rather than salary from it, because wage income is the thing being deprecated and dividend income is the thing doing the deprecating.
The category I would be most careful about is anything whose moat is human labour cost arbitrage. That moat evaporates from both directions at once.
And the personal version, which matters more than the portfolio version. In every automated system I have ever helped design, the human keeps the veto and the machine gets the clerical labour. So be the veto. Judgement, accountability, taste, the ability to say no and sign your name to it. Those hold value in a way that “I can execute the task” simply does not any more.
The good news, which is real
Here is what I keep coming back to. The chaos is a distribution problem, not a physics problem, and distribution problems are made of decisions. Somebody decides them. That is genuinely better news than a shortage, because shortages need twenty years and a mine, whereas decisions only need enough people paying attention at the right moment.
And the window is open right now. Not in the sense of a crisis, in the much better sense that the boring paperwork is still boring, the queues are still public, the consents are still argued in ordinary rooms by ordinary people, and almost nobody has connected any of it to the word abundance yet.
So the most useful thing you can do about the age of abundance is probably to read a grid connection queue on a Sunday morning, be the one person in the room who understands what it is really about, and then say so out loud. That is not a small act. That is how settlements get written by more than one side.
Also, and I say this with love, do get on top of your mortgage.
Written for KiwiGPT.co.nz · Generated, Published and Tinkered with AI by a Kiwi