Cathie Wood on Collapsing AI Innovation Costs and Market Shifts

Collapsing AI Innovation Costs

• 📉 The cost of AI innovation is collapsing, with training costs dropping 75% per year and inference costs 85-90%.
• 💡 This shift throws more weight towards inference chips, a more competitive space, compared to training chips where Nvidia is highly exalted.
• 🚀 Breakthroughs like DeepSeek demonstrate that AI models can be trained for significantly less capital, reinforcing the thesis of rapidly declining costs.

Tech Stack Evolution and Key Beneficiaries

• 🎯 Platform as a Service (e.g., Palantir) is expected to gain significant share within the evolving tech stack.
• 📈 While Infrastructure as a Service will stabilize and Software as a Service may see a share decrease, all segments are projected to experience rapid growth.
• 🏥 Key AI use cases include autonomous mobility (cars, drones) and the healthcare space, where the convergence of sequencing, AI, and CRISPR gene editing is expected to cure more diseases.

Global Competition and Market Dynamics

• 🇨🇳 DeepSeek represents significant competition, challenging the notion of US tech exceptionalism and prompting tech titans to speak of it glowingly.
• 🌍 The collapse in innovation costs is beneficial globally, but China's push for "new productive forces" highlights its growing innovation.
• 📊 Historic market concentration in mega-cap stocks is anticipated to give way to a broader-based bull market, rewarding smaller-cap stocks.

Regulatory Environment and US-China Relations

• ⚠️ US AI regulations are seen as potentially holding back innovation in a nascent technology, compared to the internet in the early 90s.
• 🤝 The discussion draws an analogy to Nixon in China, suggesting that President Trump might favor competition and less regulation in AI, despite past executive orders.
• 💰 The drastically reduced cost of AI development means that firms do not need billions in cash to excel, changing the competitive landscape.

Investment Implications and Market Outlook

• ✅ A broad-based bull market is considered stronger than one favoring only a few mega-cap stocks, indicating a coming shift in market dynamics.
• 🌱 The neglect of the rest of the market (beyond hyperscalers like Meta and Amazon) has been extreme, but this is expected to change towards diversification.
• 📈 Historically, periods of extreme market concentration, like during the Great Depression, eventually led to smaller-cap stocks outperforming as the market broadened out.