The tech takeoff: Altimeter's Brad Gerstner on Nvidia, the Cerebras IPO and the future of AI

Summary

Brad Gerstner discusses Nvidia's valuation catch-up, the Cerebras IPO, and the shift from training to inference in AI. He argues Nvidia is undervalued at 14-15x earnings despite massive demand, and sees Cerebras as a promising inference-focused chipmaker with execution challenges. The conversation highlights the growing importance of inference compute and the potential for multiple winners in the AI hardware ecosystem.

• Nvidia's stock broke out after six months of stagnation, driven by valuation catch-up rather than China news.
• Gerstner emphasizes Nvidia trades at a low multiple relative to peers despite strong earnings growth.
• Cerebras is going public with a wafer-scale chip optimized for low-latency inference.
• The world is moving from pre-training (FLOPs) to inference (token production), changing compute demands.
• Inference demand is nearly unlimited, but power and data center capacity are becoming constraints.
• Multiple companies, including Nvidia, Broadcom, and Cerebras, can coexist in the AI chip market.
• Enterprise users are now citing compute costs as a material factor in AI profitability.
• Altimeter Capital holds Nvidia as its largest position at ~20% of the portfolio.