AI’s Next Big Leap

Tom Wig and Stephen Byrd discuss the nonlinear improvement in AI frontier models and its implications for investors. Stephen explains that scaling laws show capabilities doubling when compute increases 10x, leading to a massive compute shortage. He is bullish on AI infrastructure plays and hyperscalers, citing excellent token economics and pricing power, with revenue inflection expected in 2026. - AI models are improving at a nonlinear rate, outpacing linear market expectations. - Scaling law: 10x more training compute yields 2x more capability. - Expect multiple powerful models this spring and summer at lower cost. - Token economics show strong returns for hyperscalers selling compute. - Massive compute shortage gives hyperscalers pricing power. - Agentic AI will cause a 10x increase in token usage across all knowledge work. - Fast adopters are already demonstrating huge economic benefits. - Revenue inflection for hyperscalers is likely a 2026 event.