The AI industry's extreme optimism is now facing a reality check over bubble fears and sustainability concerns.
The episode opens by framing 2025 as a 'vibe check' for AI, where initial fervor shifted to scrutiny over bubble risks and sustainability. It details massive funding rounds: OpenAI's $40 billion at a $300 billion valuation, with talks for $100 billion more; Anthropic's $16.5 billion raising its valuation to $183 billion; and startups like Thinking Machine Labs securing $2 billion seed funding. These valuations persist despite modest enterprise adoption and infrastructure constraints. The infrastructure boom is driven by capital tied to compute, blurring investment and customer demand, exemplified by Stargate's $500 billion U.S. project and Alphabet's $4.75 billion acquisition of Intersect. However, fragility is shown when Blue Owl Capital pulled out of a $10 billion Oracle deal. Grid constraints and political pushback, including from Senator Bernie Sanders, are slowing projects. The expectation reset notes that model releases like GPT-5 lacked the punch of earlier ones, while new labs like Deeptiq shipped competitive models cheaper. Investors now focus less on raw capacity and more on business models, with companies testing pricing, like OpenAI's potential $20,000 monthly fee, and distribution strategies, such as Perplexity's Snap deal. Trust and safety issues escalated, with over 50 copyright lawsuits and AI psychosis linked to suicides, leading to regulations like California's SB243. Industry leaders, including Sam Altman, warned against emotional over-reliance on chatbots. The episode ends by predicting 2026 will require AI companies to prove business models and economic value, potentially leading to a reckoning.
