Kevin Roose is AGI-pilled. Why aren’t you?

In an interview about The AGI Chronicles, Kevin Roose describes frontier AI firms racing to build AGI while relying on enterprise software sales to fund compute and research; he says OpenAI, Anthropic and Google are likely to endure, though their eventual success remains uncertain. He cites Anthropic’s growth to about 5,000 employees, compared with Dario Amodei’s 2021 estimate that roughly 50 people might suffice, and an early estimate of about $10 billion to build AGI. The discussion also highlights AI’s potential benefits alongside existential-risk concerns and the lack of public input and effective regulatory authority.

Analysis

Signal is low: this is an interview, and claims about AGI progress, inference demand, and which labs will win are opinions rather than independently verified operating data. Near term, it does not justify changing positions. The more investable tension is that frontier-model ambition can support sustained compute spending while enterprise monetization may lag; that creates different exposures for platform owners and capacity suppliers. Alphabet (GOOG) and Microsoft (MSFT) can potentially absorb AI investment across existing distribution and cloud businesses, but returns depend on incremental usage and pricing—not model capability alone. CoreWeave (CRWV) is more directly exposed to utilization, customer concentration, and financing conditions; verify those metrics before treating demand forecasts as earnings visibility. Salesforce (CRM) faces a two-sided effect: AI features may strengthen distribution, while capable agents could pressure seat-based software economics if customers reduce licenses. The YouTube-first creator strategy is a modest strategic positive for Alphabet’s platform, not a material NYT (NYT) earnings signal on this evidence.

Over 1–3 months, watch cloud capex versus disclosed AI revenue/usage, inference pricing and unit economics, and whether agentic tools complete messy real-world workflows—not just verifiable coding tasks. Over 6–18 months, reliable workflow automation could shift software spending and labor economics; brittleness or falling model prices without adequate utilization would weaken returns on compute. Contrarian point: the interview leans toward persistent frontier differentiation, but deployment reliability and customer ROI—not AGI belief—will determine durable pricing power. No trade is warranted from this source alone.

Key Decisions for Investors

• No immediate position change: the interview provides no new, verifiable company-level financial data and its industry forecasts are speculative.
• Put CRWV on a risk-monitor list; verify customer concentration, contracted utilization, financing needs, and realized inference economics before adding exposure. A thesis that demand converts to durable cash flow is falsified by weakening utilization or adverse financing terms.
• Track GOOG, MSFT, and META for evidence that AI usage contributes incremental revenue or retention relative to infrastructure and product costs; rising capex without improving disclosed returns would argue against paying for an AI-led re-rating.
• For CRM, monitor customer adoption and seat counts alongside AI feature monetization. Broad agent deployment that reduces paid seats without offsetting revenue would be a negative; limited adoption or measurable incremental spend would weaken the substitution thesis.