Zero is not a valuation

Somewhere in most bankruptcy files is an appraisal that values the debtor's data at zero, or does not mention it at all. That is rarely negligence. It is what the standard methods produce when pointed at this asset, because all three of them fail at once.

Cost fails because the data was never bought. It accumulated as a byproduct of operations, and there is nothing to depreciate. Comparables fail because dataset sales are private, bespoke, and almost never disclosed, so there is nothing to compare. Income fails by construction: the debtor never licensed its data, the cash flow attributable to it is zero, and a discounted cash flow on zero is zero. The methods are not wrong so much as out of jurisdiction. They return a number anyway, and the number is precise, defensible, and meaningless.

The bankruptcy process happens to contain the honest alternative, because a section 363 auction is a market test, and market tests keep embarrassing the models. When 23andMe moved through Chapter 11 in 2025, the stalking horse agreement valued the assets, principally the genetic database, at $146 million. The auction that followed closed at $305 million. The market disagreed with itself by more than two to one inside a few weeks, and no appraisal had produced either number. A decade earlier, creditors' experts in the Caesars bankruptcy concluded that the Total Rewards customer database was the most valuable asset in the case, with figures near a billion dollars, in an estate everyone else had been valuing by counting hotel rooms.

Models cannot reach these numbers because data is worth what a specific buyer will pay for a specific use. That depends on what the buyer already holds, what they lack, and what acquiring it saves them in time. Those inputs sit inside the AI laboratories that have become the largest buyers of licensed data, and they are invisible from the estate's side of the table. No expert report can conjure them.

None of this means every estate is holding a hidden nine-figure asset. Most are not. It means a zero appraisal is evidence the method ran out, not evidence the asset is worthless, and the two conclusions call for different responses. Where the debtor's data falls in a category with live laboratory demand, the correct appraisal is a cheap, structured market test: describe the asset accurately, stop the deletion clock, and put it in front of the people who buy such things. The number comes from the auction. It was never going to come from the model.