Google wins $10M bankruptcy auction for Spirit Airlines' enterprise data and code to train AI models
Google won an Aug 14, 2026 bankruptcy auction for defunct Spirit Airlines' corporate data at $10M, beating Mercor's $7.5M backup bid. The scope covers ~100M emails, ~500M Teams chats, ~30M lines of code and ~3.4M payroll records across ~80,000 accounts, with passenger and loyalty PII carved out and a third party de-identifying before transfer.
Where the rake sits
Spirit's estate takes a one-shot $10M creditor recovery; Google captures all downstream value — model improvements and travel-product gains amortised across its stack — because Google (and its third-party PII scrubber) does every step of the enrichment. The rake is fully on the buyer side.
What happened
- Bankruptcy court filing dated Aug. 14, 2026; Axios reporting Aug. 17, 2026.
- Google's winning bid: $10M; backup bidder Mercor at $7.5M (a two-bidder auction)
- Scope: ~100M emails, ~500M Microsoft Teams chats, marketing/HR/project/financial docs, ~30M lines of code, ~3.4M payroll records across ~80,000 email accounts.
- Explicitly excluded from the sale: 97.5M passenger profiles and 50.2M Free Spirit loyalty records (PII carved out)
- A third party will de-identify the data before transfer; buyer agrees not to attempt re-identification.
- Spirit ceased operations May 2026 after failing to emerge from its second Chapter 11; estate carries ~$8.1B in debt.
Who is involved
Buyer of the Spirit data; will use the de-identified emails, chats, code and operations records to improve products and train AI models, per the Aug. 14 bankruptcy court filing.
Alphabet, NASDAQ: GOOGL; ~$350B FY2024 revenue, >$2T market cap
Defunct US ultra-low-cost carrier (Spirit Aviation Holdings); ceased flying in May 2026 after failing to emerge from its second Chapter 11, and is now selling its de-identified corporate data trove via bankruptcy auction.
~$8.1B in debt at bankruptcy; data cache includes ~100M emails, 500M Teams chats, ~30M lines of code, ~3.4M payroll records, 80,000 email accounts
The reading
The enrichment happens entirely on Google's side: Spirit's estate hands over roughly 100 million emails, 500 million Teams messages and about 30 million lines of code per Time, and Google alone trains models and improves products on top. Spirit, in wind-down, does none of the downstream work.
What Google is paying for is a corpus of real corporate workflow (finance, ops, revenue management, pricing models, booking curves, 3.4M payroll records per Skift) that lets its models generalise agent behaviour into enterprise work - Surge's Nick Heiner framed it to Time as a bet that the agent 'is going to generalize into the rest of the economy'. The buyer's decision it feeds is which next-token prediction and which agent action to take inside enterprise tasks.
Crossed: the de-identified enterprise dataset, software code, and operations/pricing data, with PII scrubbed by a third party per Google's statement. Did not cross: customer and loyalty data (97.5M passengers, 52.4M loyalty members, 740,000 co-branded cardholders per Skift), and the Spirit brand and aircraft per InfosTourisme.
Not determinable. Mercor's $7.5M underbid anchors a market price but the auction ran inside a Chapter 11 estate under time pressure, and union objections (AFA-CWA) delayed the hearing from Aug 19 to Sept 9 per NewsNation - whether $10M under-prices what Google extracts is a judgement the public facts do not settle.
Why it matters
The case is a clean sell where the asset boundary was set by a bankruptcy court and PII carve-out rather than by commercial design, and where a distressed seller on a court timetable transferred a decades-deep corporate record into a buyer whose enrichment stack turns it into training signal for a $2T model business. It is a useful reference point for what raw substrate clears at when the seller has no ability to run the enrichment themselves and no time to price-discover.
The argument this deal tests: Data monetisation: three routes to value creation
Related deals
- Chegg pivots to licensing decade of proprietary STEM academic content and expert network to AI labs2026-05-13
- Reddit's capital-light data model: $1M capex against $300M+ cash flow demonstrates non-consumption and non-rivalry2026-04-30
- Universal Music Group licenses catalogue to ElevenLabs for multi-year AI music platform collaboration2026-09-10
- Customers Bank signs OpenAI deal to automate lending, deposits and payments using its own operational data2026-04-27
Sources
- axios.com — primary
- ppc.land — corroborating
- news.bloomberglaw.com — party background
- skift.com — party background
Announced 2026-08-14 (filed)