Alibaba’s managers burned seventy-six percent of their net profit last quarter to build AI infrastructure, then turned around and sold three-point-seven percent of the company — 710 million new shares — to outside investors to keep building it. Fabiana Negrin Ochoa reports for The Wall Street Journal the placement will raise $10.2 billion at HK$112.70 per share, a discount to the HK$123.00 Hong Kong close, with 100 percent of net proceeds committed to “full-stack AI capabilities, including infrastructure.” The deal is managed by CICC, HSBC, Morgan Stanley, and UBS.
The honest case for the raise is not weak. The managers who sign these things believe AI infrastructure is the next foundational layer — comparable, in their minds, to the railroads that built this country a century and a half ago. The discount to the share price is the cost of that conviction. The foregone dividend is the price of being early. The capital expenditure is what investment looks like before it becomes earnings, and the company that arrives first owns the ground rent of the new economy. There is a respectable conservative case that says: let capital find its work, let the patient shareholder wait, and the returns will follow.
The trouble with that case is who pays for the patience. The seventy-six percent drop in net profit is not a natural disaster; it is a decision. Someone at Alibaba decided that this quarter’s earnings — and the dividends that would have flowed from them, and the reinvestment in the working business that would have produced the next quarter’s earnings — were worth less than the speculative future embedded in a data center. That decision was made by managers whose compensation rises with the share price, not with the dividends. The small shareholder who bought the stock for the dividend is being asked, in Chesterton’s old phrase, to lend the company his money and pretend he hasn’t.
This is the rentier move dressed as the builder’s move. The patient shareholder is told to wait because the patient shareholder has no seat at the table where the decision was made. The investment banks underwriting the placement — the same firms that brought you the mortgage-backed security, the credit-default swap, and every other financial instrument that ended in a taxpayer bailout — collect their fees whether the data center returns a dime or not. The community that will house the server farm, if there is one, is not consulted on siting, tax abatement, water use, or grid load. The state that will subsidize the build — whether in Beijing or in some American county that has been promised a thousand construction jobs — does not ask whether the subsidy would have built more distributed wealth in a different shape. The returns, if they come, accrue to capital that did not do the work. The risks, if they don’t, accrue to labor and place that did.
I’ve seen this movie before. In 1909 the Chicago & North Western Railway surveyed a line through central Wisconsin and put the depot a mile south of Friendship — on the land of men who had optioned it for less than the asking price of the merchants who thought they could hold up the railroad. Adams was born; Friendship did not get the dividend. The rail jobs that would have raised our wages and populated our Main Street went elsewhere, and the merchant class that survived was the merchant class that owned the depot land. The same pattern held, a hundred years later, when Heartland Farms — five generations old, twenty-seven thousand acres of irrigated potatoes under contract to Frito-Lay — got large enough to consolidate the family farms that used to make a living on the same ground. The value leaves the place. The capital stays with the capital. The Alibaba placement is the same arithmetic, scaled to planetary dimensions: three-point-seven percent of a hundred-billion-dollar company, sold at a discount, to fund an infrastructure whose returns will accrue to the same capital class that already owns three-point-seven percent of everything else. The dilution of the small shareholder is the price of admission to the next platform monopoly. The foregone profit is the rent of being early. The community that hosts the server farm will get the property tax abatement and the water bill.
The cooperative tradition was invented for exactly this moment. The Rochdale Pioneers did not dilute their members to build a steam engine; they pooled their capital, owned the engine together, and distributed its returns to the people who had earned them. Mondragon — the world’s largest worker cooperative federation, seventy thousand employees in the Basque Country — runs on the same logic: ownership where the work is done, returns where the capital was raised. Adams-Columbia Electric Cooperative, headquartered two miles from my desk in Friendship, electrified this county on that principle in 1936, when the for-profit utility said we were too sparse to serve. The infrastructure was real. The returns stayed local. The dilution was zero.
The harder truth is that no such counter-model exists at the scale of a national AI buildout. The capital required is so vast, and the supply chain so consolidated, that even a federated system of regional cooperatives would face a procurement wall built by the same handful of firms whose shareholders we are now being asked to dilute. The remedy is not to refuse the build — AI infrastructure will be built, by someone, somewhere. The remedy is to refuse its ownership structure. Distributed ownership of the means of computation is the next frontier of the same fight we have been fighting since the railroad bypassed Friendship. The capital must be raised where the work is done. The returns must stay where the capital was raised. The dilution must be zero.