Tech Insider WARNS: "The Bubble Is About To Burst"

The AI investment buildout is not yet validated by consumer revenue; it relies heavily on subsidies, circular Big Tech spending, and speculative demand [13:26]. Rapid adoption is real, but it is not evidence that consumers will pay unsubsidised API prices .

Key points

• The interview cites 190 GW of data-center capacity in planning, which converts to $1.6–3T annual demand at $12M per megawatt, versus under $130B in current annual AI demand [10:43]. More than 70% of that demand comes from OpenAI and Anthropic, which carry $1.1T in cloud commitments and are fed by their Big Tech customers instead of independent consumers [08:01].
• ChatGPT reached 100M active users in 60 days, faster than TikTok, Instagram, and the web, with 60% of US adults using AI tools within three years . That adoption is real, but the video does not show users would pay the unsubsidised API price, which Zitron characterises as $2–4 per action [01:16].
• AI-generated code is linked to more outages, with GitHub flooded by inexperienced users, AWS having multiple outages tied to AI tools, and industry data pointing to AI-assisted coding as a cause [18:05]. Zitron says he believes Google weakened search in 2020 to increase queries and ad impressions, then placed generated answers at the top to stop users leaving [16:08]. He also says data-center power draws raise utility bills and chip demand pushes up electronics prices [03:50].
• Zitron says there was no pre-existing technical roadmap making generative AI inevitable in the way smaller GSM, Bluetooth, and Wi-Fi radios pointed to the iPhone [02:18]. Companies still use the broad 'AI' label to count protein folding, robotics, and autonomous systems as progress, even though those are not LLMs, according to the interview [09:21].