Podcast, Interview, Fireside Chat
Jensen's Open-Weights Letter | Google Cloud Grows 82% But The Market Tanks
- A revenue shortfall of approximately $100 billion for frontier labs is viewed as beneficial to existing business models, while the open weights trend is expected to clarify its market size over the next six to 12 months.
- NVIDIA faces a strategic choice between maximizing current high-margin pricing in 2024-2025 or aiming for a larger revenue scale by 2027, though open weight models may bypass their hardware while they continue as a component manufacturer.
- Regulatory restrictions on chip sales to China, combined with distillation penalties and model approval processes, are predicted to act as regulatory capture that restricts competition from open-weight Chinese models.
- Anthropic's advocacy for regulatory steps is interpreted as a strategy to impose a de facto ban or slowdown on open weight models, despite not explicitly calling for a prohibition.
- Cyber risks involving LLM agents are projected to affect every company within the next 24 months, with 20 Fortune 500 companies allegedly experiencing undisclosed breaches involving leaked confidential information.
- Security breaches caused by LLM agents are expected to occur at an unprecedented scale, potentially resulting in CIO firings, while companies may cut security corners due to rising costs and the pressure to use cheaper models.
- The assumption that open source code is inherently safer due to public auditing is challenged regarding open-weight models, as the internal weights remain unverified black boxes where undisclosed reinforcement learning could trigger specific actions.
- The inference chip market is projected to become the largest chip market globally, creating opportunities for specialized chips like those from Etsch, which could potentially reach a $100 billion valuation if integrated with allowed open-weight models.
- Google Cloud experienced an 82% year-on-year growth rate but posted its first negative free cash flow, raising market concerns over the $200 billion capital expenditure spending and its potential return within two to three years.
- Market volatility is anticipated to increase, with predictions of further crashes similar to a 28% drop in Korean markets, alongside a projected clampdown on AI spending next year as companies shift from experimentation to strict budgeting.
- A market dynamic is expected where a small percentage of "token maxed" companies cut costs, while the vast majority of new entrants expand spending, potentially offsetting reductions and swamping the market.
- A "ceiling" for AI companies is predicted to appear next year, marking the first real material constraints on spending as the industry moves beyond initial experimentation phases.
- Iconic founders like Travis Kalanick and Jeff Bezos are expected to attract significant capital for large-scale ventures, while the traditional SaaS turnaround playbook is seen as failing due to five years of price increases and limited new growth.
- Private equity investment targets are expected to be limited to companies growing at 40% or higher with agentic products, while the trend of founders leaving profitable firms for AI startups is viewed as a net negative.
- Acquisition negotiations may involve strategic leaks to generate pseudo second offers, with big companies expected to enter deal mode within days to create paper counter-offers before actual deals close after a week or more.
- Revolut is considered a more attractive investment target than Stripe due to its larger Total Addressable Market within 500 million exploited Europeans, despite Stripe's superior execution and potentially overestimated network effects.