Bloomberg Intelligence Nvidia-Google AI Chip Rivalry Escalates on Report of Meta Talks
11 snips
Nov 25, 2025 Join Lindsay Dutch, a savvy consumer hardlines analyst, who dives into the ups and downs of Dick's Sporting Goods and Best Buy's strong earnings. Mary Ross Gilbert, a retail expert, shares insights on Kohl's renewed focus on private labels and the contrasting fortunes of Abercrombie and Hollister. Diana Rosero Pena discusses JM Smucker's turbulent results in the packaged food sector and the ongoing tariff impacts. Together, they dissect the latest trends reshaping retail, from sneaker strategies to evolving consumer demands.
AI Snips
Chapters
Transcript
Episode notes
Meta Diversifies Away From Nvidia
- Meta is seeking to diversify AI infrastructure away from sole reliance on Nvidia by exploring Google's TPUs.
- Google’s Gemini model plus TPU efficiency makes Google a viable lower-cost alternative for large-scale inferencing.
Google TPUs Catch Up Technically
- Google’s TPU development reached a seventh generation and performance parity with Nvidia GPUs for many workloads.
- Gemini’s rapid model improvements made Google’s stack attractive for cost-effective token generation and deployment.
Nvidia Retains Training Dominance
- Nvidia still dominates training with superior tokens-per-watt economics and massive demand per gigawatt.
- Hyperscalers seek inferencing alternatives to lower total infrastructure cost despite Nvidia’s lead in training.