Iqbal Latif
And physics does not grant him the city. A tower on the street reuses the same airwaves every few hundred metres; a satellite hundreds of kilometres up must share them across a patch of ground many kilometres wide. In a desert, at sea or in a village with no road, the net is a blessing, and I say so. In Karachi, Lagos or Manhattan, where millions of phones stand shoulder to shoulder, the sky cannot carry what the street carries. The big American carriers have refused to rent their customers to him, and T-Mobile’s chief financial officer said at an investor conference on 10 September that satellite signals do not go through walls and a million satellites cannot break physics. His own prospectus tells the story: the average Starlink customer paid $99 a month in 2023 and $66 in early 2026. That is a network reaching poorer customers, which is good. It is not the arithmetic of a network that will own the world’s conversation. Starlink is a real achievement, and its connectivity business reported about $11.4 billion of revenue last year at a healthy margin. But a good business at sea is not a monopoly of every telephone on earth. His future is not financed by escaping the Earth economy. It is financed by capturing more of it. And I have not even counted the rest: the tunnels, the chip factory his filing describes at length without a date or a cost, and every other venture that arrives with its own promise and its own bill.
Now the robots, and here I move from his contract to his word. In Riyadh in October 2024, he said that by 2040 there will be at least 10 billion humanoid robots, priced at $20,000 to $25,000 each, more robots than people. That is a forecast made from a stage, not a milestone in a filing, and I treat it as one. But it is his forecast, it is the basis on which he has told investors Tesla could be worth $25 trillion, and he has never withdrawn it. So take him at his word. Ten billion robots at his price cost $200 to $250 trillion, about two years of everything the human race produces. If a robot lasts about 10 years, which is my assumption, the first ones wear out before the fleet is even complete, and replacing them becomes a permanent bill of $15 to $18 trillion a year. Building them is its own problem. The world has about 1.5 billion vehicles on its roads, built over more than a century, and it makes about 90 million a year. Ten billion robots by 2040 means building about 700 million a year from now on, nearly eight times the entire world car industry. His own Fremont line is designed for 1 million a year, so the world would need around 700 such factories. Every robot needs copper for its motors, lithium for its battery and rare earth magnets for its joints, most of them processed in one country, China, and all of them already fought over by electric cars, wind turbines and power grids. If each robot draws an average of 300 watts, again my assumption, 10 billion of them need about 26,000 terawatt hours a year. That is close to all the electricity the world generates today, before a single data centre is switched on to think for them. And if he is only a tenth right, if there are 1 billion robots rather than 10, the bill is still $20 to $25 trillion, close to a whole year of the world’s investment, and every argument below still stands.
Then the checks. He says everyone will have a high income, and the Treasury should simply issue them. There are about 8.2 billion people on earth. Suppose a high income means even $30,000 a year, my figure and not his; then the bill is about $246 trillion a year, twice the entire world economy. Do it only in America, for 340 million people, and the bill is about $10 trillion a year, roughly twice everything the federal government collects in taxes. His answer is that prices will fall so far that a small income buys everything. But if prices fall that far, the revenues of the companies that own the robots and the satellites fall with them. And if their revenues fall, their market values cannot climb to the $7.5 trillion his share grant needs. If goods become nearly free, a company worth $7.5 trillion needs enormous income from somewhere else: from the machines, the networks, the rights, or the sheer volume it controls. The prices can fall for the buyer while the income gathers in the owner. His pay package bets on that. His speeches promise the falling prices and leave out who owns the machines.
Now let me put his future on one page, as a banker would. The world economy today is about $124 trillion a year. Of that, the world invests roughly a quarter, about $30 trillion, in everything it builds: homes, factories, roads, power, hospitals and the replacement of what wears out. Take his timetable. Ten billion robots by 2040 costs about $15 to $18 trillion a year for 14 years, and replacing them costs about the same every year after that. If the orbital data centres are built over 20 years, they add $5 to $10 trillion a year. Mars and the Moon, even on his cheapest numbers, add a few hundred billion more. Together that is roughly $20 to $28 trillion a year, every year, for a generation.
Set that beside the $30 trillion the whole world invests, and his future needs between two thirds and nearly all of it. So there are only two roads. Either the world stops building almost everything else, the houses, the grids, the schools and the water, for 20 years. Or the world must first become 16% to 22% richer than it is today, and it must do so before the robots have produced a single thing. In a good year the world economy grows by about 3% in real terms; even in its best recent year, the rebound after the pandemic, it grew by about 6%. At a good pace, finding that extra output takes five to seven years of the whole world’s growth, and every dollar of that growth already has a family, a firm or a government waiting to spend it.
Where does the money come from? There are only four places. The first is the world’s savings, and they are already spoken for. The second is government borrowing, which the rich countries already carry at record levels and which must be repaid from taxes that shrink as wages shrink. The third is the stock market, which is where he has gone: the listing, the $25 billion of bonds, and the index rule that made every fund tracking the Nasdaq 100 a buyer of his shares. The fourth is the pension funds, which hold nearly $60 trillion across the world’s largest markets and which now own his shares whether their members chose them or not.
What does the rest of the world suffer while all this is paid for? First, its ordinary investment starves. Every tonne of steel, every chip, every megawatt and every engineer’s year sent to Mars, the Moon or orbit is one not sent to Manchester, Lagos or Sukkur. The grid is not upgraded. The hospital is not built. The water plant never comes. Second, materials become dearer. Ten billion robots, millions of launches, a million satellites and a fleet of data centres all want the copper, lithium, magnets and electricity that the rest of us already need. When that much new demand arrives for the same things, their prices rise, and a family that only wants to run a fridge and a fan pays more for its power because the robots and the orbital computers bid for it first. (To be continued)
