The AI Boom Is Entering A New Phase -- And Nvidia May Not Be The Biggest Winner - Yahoo Finance

As the artificial intelligence (AI) industry continues to evolve, the market is entering a new phase where inference—rather than training—is becoming the dominant use case for AI models. This shift pr... Read original →

The surge in AI inference is triggering a paradigm shift in semiconductor architecture. Technologically, 3nm and EUV processes are increasingly optimized for ASICs, as GPU’s general-purpose design struggles with the power-efficiency demands of hyperscale data centers focused on total cost of ownership. Geopolitically, U.S. export controls on advanced chips paradoxically accelerate Big Tech’s in-house ASIC development—bypassing restrictions via customization—but extend supply chain validation cycles and inflate upfront capex. In the competitive arena, while NVIDIA clings to its CUDA moat, Broadcom is leveraging custom AI accelerators for Google and Meta to penetrate the data center core; hitting its $100B AI chip revenue target by 2027 would upend pricing dynamics. Over the next 18 months, every 10% gain in ASIC adoption could compress GPU gross margins by 2–3 points—not just a product substitution, but a fundamental recalibration of compute economics.