Wrong Companies Specific
What If the AI Supercycle Is Neither Bubble Nor Boom — But a Railroad Story?
The entertainment stories of the week share an underlying theme: institutions discovering how little control they retain over information in a media environment where suppression amplifies. ABC, owned by Disney, blocked a Rosie O'Donnell Trump parody for Jimmy Kimmel's show — and the decision itself became the story, broadcasting the parody's existence to an audience far larger than the original broadcast would have reached. Wisconsin Democratic gubernatorial frontrunner Kayla Hong faced criticism from within her own party — including from Obama's former chief strategist David Axelrod — over a 'cancel Thanksgiving' remark made in a CNN interview, a week before her primary. A congressional candidate who shot a Flock automated license-plate-reader camera on camera and subsequently raised thousands in small-dollar donations demonstrated how a very specific surveillance concern can mobilize constituencies across ideological lines that agree on almost nothing else.
The policy story with the most direct effect on people's lives in this segment is the Trump administration's decision to end Medicare Part D's premium stabilization program a year ahead of its scheduled sunset, potentially raising prescription drug costs for 25 million enrollees on a compressed timeline. For a population largely on fixed incomes, drug cost increases are not abstract.
A writer identified as the author of 'Call Me' disclosed that she lost a $2.4 million book deal after AI-generated claims about her circulated — the author declaring the claims false. The incident illustrates two distinct failures: if publishers are making seven-figure decisions based on AI-generated research without verification, that is an editorial process failure; if AI systems are generating false biographical claims, existing legal remedies for defamation were not designed for the speed and scale at which such material spreads.
The 'What If We're Wrong' question this week targets the most confident claim in financial media: that the AI investment cycle is either a bubble approaching collapse or a transformation that justifies current valuations. Both Burry and Dalio's warnings and the Situational Awareness fund collapse have made the bubble thesis increasingly the modal view rather than the contrarian one. The more useful challenge is to ask what the bull case looks like if the skeptics are wrong on timing. The railroad buildout of the 1870s and 1880s appeared to be speculative excess — many individual railroad companies went bankrupt — but the physical railroads, once built, generated economic value for a century. Investors who lost money were not wrong that railroads were important; they were wrong about the timeline and about which specific companies would capture the value. The specific indicator worth watching: corporate AI revenue as an auditable line item in earnings reports. Most companies today report AI-driven cost savings or efficiency gains, not AI-native revenue. If specific, verifiable revenue figures attributed to AI products do not appear in meaningful volume by mid-2027, the bubble argument strengthens substantially. If they do, the Burry-Dalio timing analogy will have been wrong even if the structural observation was astute. A second scenario that current analysis underprices: a significant deflationary shock to compute costs. If the next generation of chips drops inference costs by 80 to 90 percent between 2025 and 2027, it would validate the technology while potentially undermining the companies whose valuations rest on selling expensive compute — because the companies that win in a cheap-compute world are not the same as those that win in an expensive one.