Databricks priced its latest round at $188 billion this week, a private secondary valuation, not a public listing, which means the number comes from investors buying existing shares from employees and early backers rather than from a stock exchange setting the price. That distinction matters because it happens the same week New York became the first US state to impose a one-year moratorium on new data center construction, and the same week Apple's trade-secrets suit against OpenAI moved toward discovery. Three stories, one substrate: whoever wants to build the next training cluster now has to clear a state permitting freeze before it clears a compute order, and whoever wants to keep selling enterprise AI tools now has to survive litigation before it collects the contract. Databricks sells the picks and shovels, the data infrastructure that feeds models rather than a model itself, and picks-and-shovels valuations do not carry the same exposure to either fight.

That is the read the $188 billion number is actually pricing. A model vendor's value sits downstream of two things it does not fully control: where the next data center gets built, and whether the litigation around a specific product survives contact with a courtroom. Databricks' value sits upstream of both, in the layer that ingests, cleans, and pipes data into whichever model wins, on whichever cluster gets approved. New York's moratorium runs through July 2027, and it is the first state-level construction freeze aimed squarely at the industry's physical bottleneck: power draw and water for cooling. Every model company with a training run scheduled past that date now has to route around New York, the same way it already routes around BIS export controls for chip access. The data layer does not need a zoning variance to keep growing.