Read the headline again: a company lost $7 billion in a quarter and its stock jumped 13%. That is not a market pricing risk. That is a market that has stopped asking what it is buying.
The cheer has a source. SoftBank says its $54.6 billion in OpenAI is now worth $99.6 billion, and that gains from OpenAI's latest funding round aren't even in the numbers yet. Notice what kind of money that is. It is a mark, an estimate, a number set in a round in which SoftBank is itself a leading buyer. SoftBank is grading its own homework and the stock market is giving it an A. Meanwhile the losses are concrete: Intel's share price fell, and SoftBank has sunk $9.48 billion into power-generation equipment for U.S. data centers, then had to change its procurement plans. Paper is up. Steel and turbines are the problem.
This is Masayoshi Son's signature move, and we have seen it before. The Vision Fund treated WeWork's valuation as reality until reality arrived. Son is a man who has famously ridden a fortune up and nearly all the way back down. Now he is at the center of the largest private build-out of physical infrastructure in a generation, and his own filing admits it is bumping into shortages and long delivery times for power equipment.
Why should anyone but SoftBank's shareholders care? Because that power equipment is not an abstraction. It is meant to plug into American grids, and American grids are shared. When a hyperscale data center arrives, someone pays for the transmission lines, the substations, the new generation, and the higher capacity prices that follow. In the country's largest grid region, capacity auction prices have exploded, with data center demand cited as a main driver, and utility bills are already a political problem in states that have never voted on whether they want to underwrite the AI race. The upside of this boom is being booked privately, at valuations like $99.6 billion. The cost, spread across ordinary customers, is not.
And when the bets sour, watch who is asked to help. An OpenAI executive has already mused publicly about a federal backstop for the industry's financing, then retreated when the reaction was ugly. Treat that as a trial balloon. The sequence is familiar: privatize the gains, call the risk visionary, then discover the enterprise is too important to fail. We bailed out banks that told us the same story about their own assets.
The Left needs to be honest about its own temptation here. Democratic governors and mayors compete for data centers like they are Amazon's second headquarters, handing out tax abatements for facilities that employ a few dozen people once the concrete dries. Some on our side are so eager to be seen as pro-growth, pro-innovation and pro-clean-energy that they have waved through deals without asking who covers the grid upgrades. That is not progressive industrial policy. It is subsidy without strings.
The demand is simple. Large energy users should pay the full cost of the capacity they force the public to build, under tariffs that protect residential customers. Public subsidies should come with binding commitments on jobs, power sourcing and water. Any talk of federal guarantees for AI financing should be met with a flat no, or with public equity and public control in return. If this technology is as transformative as its financiers claim, they can afford to carry their own risk.
SoftBank's investors may be right. OpenAI may be worth every dollar of that mark. But a number that rises because the people holding it say so is not evidence of anything. Before the public is asked to power it, insure it or rescue it, the people selling the dream should have to show that it is real.
How it may affect me
If you pay an electric bill, the data center boom may reach you through higher capacity and transmission charges. Check whether your state utility commission is considering "large load" tariffs that make big data centers cover their own grid costs, and comment or contact your representatives. If your retirement fund holds broad tech or AI-linked stocks, you also have indirect exposure to these valuations.
