Sequoia Capital

Maintain a core position in NVIDIA (NVDA) as it remains the industry standard, with the upcoming Blackwell and Rubin architectures expected to deliver up to 30x performance improvements. For investors seeking value in custom silicon, Google (GOOGL) and Amazon (AMZN) offer high-conviction alternatives through their TPU and Trainium programs, which provide superior cost-efficiency for large-scale AI training. Monitor Broadcom (AVGO) as a key beneficiary of the "make vs. buy" trend, as they partner with hyperscalers to design these increasingly vital custom ASICs. High-growth opportunities exist in "NeoClouds" like CoreWeave or Nebius, which outperform traditional cloud providers by building data centers specifically optimized for AI workloads. To hedge against the looming power bottleneck, look toward energy infrastructure and companies capable of integrating high-bandwidth memory (HBM) directly onto logic chips to solve critical hardware constraints.