Anthropic’s Growth is Slowing: Is AI Capex Next? | The Weekly Wrap

Episode Overview

• Host Steve Eisman discusses recent tech and market developments, focusing on Anthropic's S-1 filing, French sovereign debt, a massive wave of corporate earnings, and constraints on AI data centers.
• Explains how the end of token maxing is slowing down growth for major AI firms like Anthropic.
• Contrasts the performance of various market sectors, noting the distinct split between tech stock corrections and strong showings in energy, infrastructure, and healthcare.
• Examines the real-world operational challenges of AI data centers, specifically power grid limitations.

Key Concepts

• Token Maxing and the AI Growth Slowdown: Token maxing refers to the initial phase where enterprises pushed employees to use AI tools regardless of cost. As companies face rising expenses and shift to cheaper open-weight models, sequential quarter-on-quarter growth for top LLM providers is starting to decelerate.
• Sovereign Debt Scale (US vs. France): French sovereign debt is significantly smaller than US sovereign debt. Despite similar debt-to-GDP ratios, France's smaller economic footprint means it has less leverage in global capital markets, making political threats to default highly impractical.
• The SaaS-pocalypse and Customer Retention: The fear that AI will completely replace software-as-a-service (SaaS) companies often resurfaces during market corrections. However, many SaaS companies are deeply embedded in enterprise infrastructure, meaning any disruption will be gradual rather than immediate.
• Infrastructure Constraints over Capital: While tech giants have massive capital expenditure budgets for AI, the actual physical deployment of data centers is increasingly bottlenecked by physical constraints like electricity, transformer availability, and power substation construction timelines rather than lack of funding.

Quotes

• At 3:32 - "Anthropic finally dropped its S-1, and there were some very important data points." - Explains the initiation of the core analysis of Anthropic's financials and future commitments.
• At 6:08 - "But that ended this summer, when costs got out of control and companies now had the option to switch to cheaper open-weight models." - Highlighting the shift in enterprise behavior that is cooling down initial AI revenue growth.
• At 13:20 - "The point is that CoreWeave's business model means that they will be building more AI data centers, and to build them they will have to raise new debt." - Clarifying how rising debt costs impact the viability of specialized AI cloud providers.
• At 15:48 - "2027+ will be all about who has electricity and who doesn't. GPUs, HBM (high bandwidth) memory and money are not the problems" - Pointing out the shift from hardware supply constraints to utility and power grid constraints.
• At 16:33 - "Announced capex sounds great in the media and politically, but stranded capex is what we see because so many still don't have power..." - Detailing the gap between corporate financial commitments and real-world infrastructure readiness.

Takeaways

• Evaluate AI investments not just by announced corporate capex, but by the physical capacity of the utility grid and infrastructure partners to deliver power to new data centers.
• Factor in the transition from high-margin capital-light business models (like traditional web search) to high-capex, lower-moat models when valuing tech giants undergoing AI transitions.
• Look past short-term market hype regarding sovereign debt defaults or rapid tech disruptions by comparing absolute market scale and integration depth of the entities involved.