In this episode of The Deep Dig, we take apart one of the most repeated slogans of the modern tech era—"data is the new oil"—and expose it as a 20-year misdirection. Tracing the phrase from Clive Humby's 2006 talk through The Economist's 2017 cover story, we show how the metaphor was stripped of its original meaning and weaponized to naturalize mass surveillance. We run "data as oil" through four axes of basic economics and watch it collapse, then reveal the resource that actually behaves like oil: compute. Drawing on Pulitzer-adjacent George Polk Award investigative reporting into hidden data centers, a March 2025 superintelligence strategy paper, and a string of dueling peer-reviewed studies on algorithmic influence, we argue that AI systems don't watch you—they compress you, discarding the unique, irreducible parts of who you are as statistical "error."
Category / Topics / Subjects
The "Data Is the New Oil" Myth
Economics of Data vs. Compute
Data Center Secrecy and Local Governance
AI Compute as Geopolitical Resource
Algorithmic Compression of Human Identity
Latent Persuasion and Algorithmic Influence
Algorithmic Monoculture and Systemic Risk
Surveillance, Power, and Accountability
Best Quotes
"Data is the new oil... It's completely economically illiterate. It makes zero sense when you actually look at the math."
"A warehouse full of oil doesn't get you slapped with a $2 billion lawsuit by the European Union under the GDPR. Oil doesn't get you sued."
"They turned global surveillance into geology to avoid accountability."
"The residual is where you live."
"You aren't being watched. You are being rounded off."
"Stay sharp, stay irreducible, and whatever you do, never let them file you under noise."
Three Major Areas of Critical Thinking
The Anatomy of a Load-Bearing Lie: Examine why "data is the new oil" survived for two decades despite failing on all four economic axes—rivalry, fungibility, asset-versus-liability, and returns to scale. Analyze who benefits from the metaphor's persistence: how framing surveillance as "resource extraction" launders creepy behavior into something noble, smuggles in an implicit property claim, and manufactures a false sense of inevitability. Consider the broader lesson that a bad metaphor refusing to die in public consciousness is often keeping someone's profitable business model alive—and what other "common sense" tech narratives might function the same way.
Misdirection and the Architecture of Secrecy: Discuss the gap between how legitimate commodities behave (transparent markets, public ownership records, spot prices) and how the data economy actually operates (shell LLCs like Sidecat, Mellin Enterprises, and Montauk Innovations; NDAs gagging public officials; data centers traceable only through diesel-generator air permits and industrial water filings). Evaluate the claim that opacity isn't merely hiding the truth of the metaphor—it is the refutation of it. Then weigh the central reframe: that compute, not data, is the scarce, rival, geopolitically contested "fissile material" of the AI era, and why aiming public anxiety at data privacy may be diverting attention from where real power is being consolidated.
Compression, the Residual, and the Erasure of the Self: Consider the "three cardboard boxes" model of lossy compression—where an algorithm keeps your median, generic traits and discards the jagged, unique edges that make you you. Reflect on the three escalating claims of harm: latent persuasion (an autocomplete-style assistant measurably shifting users' actual opinions), algorithmic monoculture (the loss of human variance that once functioned as a societal safety net, so that rejection by one model becomes rejection everywhere at once), and population-scale conformity (the unresolved scientific brawl across the 2023 Facebook study, its 2024 rebuttal over 63 "break-glass" changes, and the 2026 X experiment showing asymmetric, persistent effects). Debate what it means—practically and ethically—to be treated as an "error term," and confront the closing provocation: are we already smoothing our own edges to avoid being flagged as statistical noise?
For A Closer Look, click the link for our weekly collection.
