Google agreed in August to pay $10 million for Spirit Airlines' internal data, as much as SimpleClosure expects to return to all its startup clients this year. The sale is still contested, and the employees whose messages sit in the archive were never asked.
A market has formed for the email and chat archives of failed companies. The first big price tag came from a bankruptcy, and the court has not yet decided whether it can be paid.
The market is priced on an untested thesis. Buyers pay for archives, while employees and later users carry the legal risk.
Nobody has published a test showing that real archives beat the alternatives. Creditors who collect the money carry very little of that risk.
The benchmark is Spirit. iTnews reported on 19 August that Google would buy Spirit's business data for US$10 million.
The sale covers employee emails, Teams messages, spreadsheets and calendars. MIT Technology Review adds that Google outbid Mercor and puts the archive at about 100 million emails and 500 million Teams messages.
The word that matters is "agreed". The docket of Spirit Airlines in the Southern District of New York, Case 25-11897 before Judge Sean Lane, shows the sale of what it calls the Deidentified Data is still pending.
The Epiq docket lists two notices adjourning the hearing, one on 28 September and one on 9 October. The Epiq case page identifies the proceeding and its court record. An amended agenda for a 14 October hearing was filed on 12 October.
The price per item is strikingly low. Ten million dollars spread over 600 million emails and messages is about 1.7 cents each.
That figure is a ceiling for the messages because the same payment also covers spreadsheets, calendars and other files. Even so, the archive is priced as a bulk lot. At that rate a million messages would fetch about $17,000, and a founder with a small workspace would see far less.
The scale gap is more telling. SimpleClosure, a service that winds down startups, launched its Asset Hub on 16 April to sell source code and, in beta, workspace data.
MIT Technology Review reports that the company has handled more than 150 Asset Hub deals and expects to return over $10 million to startups by the end of 2026. One Spirit contract therefore equals the projected annual payout of the whole founder side of the business.
The figures measure different things: a price paid by a buyer and money returned to founders. The comparison still shows how thin the rest of the market is.
The one founder price on record is small. Keelan Miskell told the magazine he sold code for about $25,000 and expects about $15,000 more for workspace data, which is still pending.
That makes $40,000 in total. SimpleClosure says 47% of its deals include code or data. The company supplied figure suggests data is a routine line item rather than a rarity.
Labs say synthetic data has limits. Surge AI, a data supplier, argues that real workplace threads show judgment and disagreement that generated text lacks.
The claim is plausible but unproven. A buyer paying cents per item has little reason to test it before signing. MIT Technology Review reports that no company has published a comparison of real archives against alternatives.
Every quote therefore rests on a research thesis, and prices diverge by orders of magnitude. Price discovery has not happened yet.
The people missing from the transaction are employees. In a bankruptcy a debtor sells whatever has value to raise money for creditors. The proceeds go to senior creditors and investors rather than employees.
Estimates from Carta and AngelList put annual startup shutdowns up between 26% and 56%, so the supply of such archives is growing. No federal law bans these sales, according to Ifeoma Ajunwa, who wants employees to give express consent. Attorney Greg Ewing calls such data "legally radioactive".
Spirit's docket shows how that plays out. The Association of Flight Attendants filed a supplemental limited objection on 24 September, and the Air Line Pilots Association objected the next day. Springshot, Inc. filed its own objection on 9 October. A Consumer Privacy Ombudsman, the official a bankruptcy court appoints to review the sale of personal information, filed a supplemental report on 5 October.
The best defence of the sales is that the data is scrubbed and staff were monitored as a condition of employment. Deidentified is the debtor's label for the asset, and a label is a seller's statement about its own product.
Ajunwa's answer is that the monitoring rationale ends when the company does. Staff accepted monitoring by an employer. Nothing in that arrangement covers a sale to a third party after the employer is gone.
For a founder or board winding down a startup, the practical step comes before listing workspace data. Ask the broker for the buyer's identity, the scrubbing method and a written route for employees to consent.
At $15,000 to $25,000 the price is small against the exposure Ewing describes. The trade off is a slower closing and a lower bid.
For a buyer's data counsel, the same logic runs the other way. With no statute to rely on, contracts are the only protection, so provenance warranties are the item to demand.
A buyer who skips them inherits the open question of whether anyone asked the employees. That question will follow every later use of the data. The seller has been paid by then, and the buyer is left holding the archive.
The hearing on the deidentified data sale, adjourned twice, is the test. The sale order, or its denial, will fix whether a consent route for employees is required.
That ruling will matter more to this market than another price.