Friday, September 4
AI Power, Capital, and Guardrails
The AI capital machine is running at full speed. Crusoe (opens in a new tab) is reportedly raising $3B at a $30B valuation to feed the compute hunger underlying every major model push, while Accel is in talks to lead a $1B round for Thinking Machines (opens in a new tab) at a $40B valuation — a company most senior executives couldn't have named six months ago. These numbers arrive the same week OpenAI formally launched GPT-6 Astra (opens in a new tab), with Sam Altman publicly acknowledging the model's potential for serious harm, a rare admission that suggests even the builders are struggling to pace governance with capability.
Tesla's Cybercab launch (opens in a new tab) landed with a quietness that felt strategic rather than accidental — no grand Musk spectacle, just a subdued Austin rollout that analysts are already calling the company's fork-in-the-road moment between a car brand and a mobility platform. Tesla is now actively surveying potential buyers about purchasing and operating Cybercab fleets (opens in a new tab), a franchise-style model that would shift capital risk onto operators while Tesla collects software margin. Meanwhile, Oura files to go public (opens in a new tab), putting a number on the wearable health data market and signaling that investor appetite for autonomous, sensor-driven consumer products has not cooled despite a choppy IPO environment.
The trust infrastructure underpinning digital identity is cracking in plain sight. Hackers maintained a live feed (opens in a new tab) of every ID scanned by a verification company for over a year — an exposure so complete it calls into question the entire third-party KYC model that fintechs and platforms depend on. Separately, Abliteration.ai (opens in a new tab) is now openly commercializing the removal of AI safety guardrails, offering enterprises unconstrained model outputs as a paid service, and TikTok pulled out of a scheduled House committee meeting (opens in a new tab) specifically to avoid answering child safety questions, a maneuver that Congressman after congressman is now citing as evidence the platform cannot self-regulate.