ByteDance is developing custom CPUs for its AI infrastructure to mitigate surging chip costs, potentially reducing Nvidia GPU demand. Alphabet’s TPUs target about 20% of the AI accelerator market and AWS is deploying Trainium and Graviton chips alongside its $150 billion GPU run rate, intensifying competition.
ByteDance is engineering its own CPUs for deployment in servers and data centers to support a massive rollout of agent-based AI products and alleviate surging chip prices and supply constraints, potentially reducing its reliance on Nvidia GPUs.
Alphabet has tripled its stake share to nearly 7% of its $332 billion portfolio while positioning its custom Tensor Processing Units to win deals at Anthropic and Meta and secure roughly 20% of the AI accelerator market, challenging Nvidia’s GPU dominance.
Amazon Web Services, generating a $150 billion GPU run rate, is rolling out its Trainium AI accelerators and Graviton CPUs to cut capital expenditures by tens of billions annually, yet continues to integrate Nvidia GPUs for high-performance workloads.
A leading mega‐cap growth ETF allocates 45.8% of its 59-stock portfolio to Nvidia, Apple, Alphabet and Microsoft, driving significant passive inflows into Nvidia but raising volatility risks if AI growth disappoints.
Fool
