Google Paid $10 Million for a Bankrupt Airline's Data: Data Is the Moat in the AI Era
In the AI era, a company's most valuable asset is its data—Google paid $10 million for a bankrupt airline's data; AI agents with legitimate access are the new biggest threat, and data infrastructure must be rebuilt.
The video won't play here. Listen to the audio instead:
The argument · tap a timestamp to hear it
Models and compute have zero switching costs; only data is uniquely yours
Eon's founders point out that models and compute have near-zero switching costs, leaving data as the only truly unique asset. Two days ago, Google bought data from bankrupt Spirit Airlines for $10 million, with another bidder reportedly Mercor. This is no longer an isolated case: CEOs are often asked 'would you sell your data,' and AI labs even approach hedge funds on Wall Street to buy data. Real enterprise data has become scarce fuel for training agents, and even a bankrupt company's data can become an auction item.
Real enterprise data is scarcer than imagined; much public data is synthetic
The host questions that 'data is the new oil' didn't hold in the slogan era, but post-training and reinforcement learning have turned it into real demand. Companies like Applied Comp offer fine-tuning services on datasets, but Eon believes truly usable real enterprise datasets are extremely scarce, and much public data is synthetic. Spirit Airlines' data can train airline customer service agents, but also serves as a panoramic sample of how a large enterprise operates—hierarchy, middle management, and worker collaboration—exactly the material agent training lacks most.
Data silos aren't a tool problem; incentives are misaligned
Data tools are plentiful; the problem is data locked in business units. In the past, data teams chose their own projects and tools (Fivetran, DBT, Monte Carlo). Now CEOs, educated by ChatGPT, demand data activation, but data is scattered across 20-year-old systems, even on servers no one dares to shut down. Extracting data requires engineers, at the cost of security, compliance, and production disruptions; business owners and platform teams have fundamentally misaligned incentives. Eon's approach is to first auto-map, classify, and build a semantic layer, then continuously supply data without impacting production.
Security models were built for humans; now the intruders are agents
Existing security models target human threats. At AWS, Eon saw a customer whose environment was 60% exposed to ransomware because resources weren't properly mapped, classified, or tagged. Now the same threat comes from non-human actors—AI agents with legitimate access entering databases and dropping a table in an instant. Detection and recovery methodologies are similar, but the speed is extreme. A few months ago, nearly every enterprise customer either feared agent attacks or had already experienced one; organizations must design defenses assuming they are already compromised.
More agents mean dashboards become more essential, not less
Enterprise data stacks were built around humans asking questions via dashboards; agents will dynamically reason over larger datasets, access SaaS and historical data, and take direct action. The guests believe dashboards won't disappear but will become more common—humans need them to understand what agents are doing in the environment. Coding agents will write most code, agents will activate other agents, and tracking non-human identities (NHIs) becomes a core challenge, with many NHI security companies emerging. Non-technical employees building agents with Lovable may not know data compliance boundaries, creating new actors in the organization unbound by rules.
Legacy pipelines aren't connected because full context wasn't needed before
Take buying coffee with a card: the transaction is written to a database, extracted, and sent elsewhere for separate processing. Pipelines aren't connected because full context wasn't needed—you only handled predefined questions. Today, if you can intelligently collect, efficiently store, and activate data, you enable combinatorial insights: merging a list of 'people in New York who like burgers' with 'people in New York who like pizza' yields a new audience. Data ingestion is growing crazily; old ETL tools can't handle multi-party noise, and vendors like Databricks are reinventing themselves.
AI transformation is faster than cloud, yet customers have less control
Eon's two founders did large-scale cloud migrations at AWS, with customers running from thousands to hundreds of thousands of servers; transformation required both technical and human investment. AI-era transformation is an intensified version of the cloud era, but customers are losing control—fearing system crashes, data leaks, and IP exfiltration—and control has become a inhibiting factor. Cloud is abstract, but AI is understandable to everyone (the ChatGPT moment), so boards push companies to adopt AI from both 'opportunity' and 'fear' angles. Traditional enterprises needing rapid deployment can only shorten sales cycles by bringing in Silicon Valley engineers with mature solutions and playbooks.
The fastest AI path for big companies is buying a startup outright
How enterprises consume software is changing. Large traditional companies want AI but internal processes take one to two years, so they bring in top engineers and ready-made solutions to accelerate; product-led growth (PLG) is particularly effective in AI infrastructure, with Cognition as a case of PLG first, then moving to large customers. Another aggressive approach is directly acquiring a startup and transforming it into an AI company to improve margins and efficiency faster. The guests believe 'the revolution is just beginning'—most companies are still at the start of their AI journey, but eventually all will be pushed forward.
In their own words · checked verbatim
The most valuable thing that you have Is actually your data.
They bought data for $10 million, because they think it's very important in a dark perspective.
We have a lot of data in the organization. We now realized data's new oil.
I actually think we'll see more dashboards because this will be the only way to kind of let us figure out what they have going on in the wall.
Figures
| Ransomware-exposed customer environment percentage | 60% | 15:00 |
| Server scale involved in cloud migration | thousands to hundreds of thousands | 28:13 |
Glossary
- NHI / Non-Human Identity
- Identity credentials used by machine accounts, service accounts, and AI agents in systems; a key concept in agent security.
- PLG / Product-Led Growth
- A business strategy where the product itself attracts and converts customers, with enterprise sales following up for expansion.
- Data Foundation
- Eon's proposed unified data layer integrating storage, protection, classification, semantic layer, and AI access.
How to listen
CTOs, data platform leads, and AI engineers focused on data assets and security in the AI era, plus investors evaluating data company valuations.
The cold open preview can be skipped; start at 1:27 for the main discussion.