Apple loses chip priority to AI partner
Also a spreadsheet error weighed 1.5 Empire State Buildings.

The Custodian's Closet Showdown: Apple Learns About First-Come, First-Served

Apple, the former undisputed anchor tenant of the world’s most exclusive chip foundry, has reportedly been asked to start using the general capacity sign-up sheet like everyone else. For years, Apple enjoyed special treatment at TSMC's manufacturing facilities, having what amounted to reserved parking for the bleeding-edge production processes. Now, all that personalized goodwill is being quietly moved into a shared Slack channel.

The sudden chill comes courtesy of Nvidia, whose voracious appetite for AI chips has allowed it to grab the top customer spot in at least one or two quarters of 2025. TSMC CEO, Mr. CC Wei, made the trek to Cupertino last August to personally deliver the unwelcome news: not only are prices going up, but Apple no longer has guaranteed access to production capacity. This is not a malicious act; this is simply the market rewarding the loudest, fastest growing line item on the invoice. Apple is now in the same waiting room as everyone else, frantically trying to secure 2nm capacity so their next iPhone will still feel sufficiently "Pro", while Nvidia just keeps hitting the refresh button on their purchase order form.

The New HR Dashboard Accidentally Tags People for Reorganization

Palantir has delivered a new piece of organizational excellence software called ELITE, or Enhanced Leads Identification & Targeting for Enforcement, to the folks at Immigration and Customs Enforcement, ICE. This is an app that uses advanced analytics to identify and prioritize "high-value targets" by populating a map with potential people and giving each a dossier, complete with a "confidence score" for their current address.

The company is trying very hard to optimize a workflow, and the app is simply optimizing for geographical density, which is a key performance indicator in any field operation. An ICE agent mentioned during testimony that they use the tool to find "target-rich areas," preferring dense populations over less populated ones, which is just standard logistics for maximizing throughput. It is a textbook example of a data analytics firm building an exceptionally efficient hammer and then being mildly surprised by the exact type of nail its client decided to smash.

The Decimal Point that Ate Half a Million Tons of Copper

A technical paper released by Nvidia quietly set the commodities market on fire by suggesting that a single 1 Gigawatt data center would require up to 500,000 tons of copper for its internal rack busbars. For a brief, terrifying moment, analysts had to confront the possibility that the global AI build-out would consume almost half of the world's annual copper supply, all thanks to one document.

The true figure is closer to 200 tons, meaning the error in the paper was a factor of approximately 2,500x, most likely a simple unit conversion mistake between pounds and tons. This innocent typo gave us a moment to reflect on the sheer lack of physical intuition in financial modeling, since 500,000 tons of copper is roughly the weight of 1.5 Empire State Buildings. One must wonder if anyone paused to consider whether the foundation of their future multi-billion-dollar data center would survive the installation of the busbars. Nvidia has since corrected the figure to something far less terrifying, and the copper market is adjusting its expectations accordingly.

Briefs

• Creepy Links: The new URL shortener is explicitly designed to make your links look as suspicious as possible. Finally, a tool for the honest, self-aware malware distributor.
• Tech Writers: An open letter was written to management everywhere who believed that hiring an AI meant they could fire all their technical writers. This is like firing your janitorial staff because you bought a Roomba.
• Wikipedia Reaches 25: The massive, collective source of slightly-out-of-date trivia celebrates 25 years of existence. It is the internet's oldest, most dependable employee who everyone trusts but nobody actually pays.