Yahoo Finance·2025-12-14 15:40Core Theme - The total cost of ownership, which includes chip price, energy usage, and performance, is crucial in determining whether Google's TPUs can undercut NVIDIA's offerings [1][15]. TPU vs. NVIDIA - TPUs are application-specific chips that are more energy-efficient and potentially cheaper than NVIDIA's general-purpose hardware, which could lead to a significant market shift as Google considers external sales of TPUs [2][5]. - Broadcom's agreement to sell TPU chips to Anthropic is projected to reach $21 billion by the end of 2026, highlighting the financial impact of TPU adoption [2]. Market Sentiment and Performance - Companies linked to the TPU supply chain, such as Broadcom, Celestica, and TTM Technologies, have significantly outperformed those dependent on NVIDIA as expectations for TPU deployments rise [3]. - Since last November, companies with exposure to Google and TPU infrastructure have seen an index value increase to 265, while those tied to NVIDIA have only reached 134, indicating a 34% gain versus a 164% gain [17][18]. Financial Implications - The AI compute market is projected to be worth over $500 billion, and companies with a 40-50% margin on TPUs could gain a competitive advantage by being the lowest-cost producers [12][13]. - An incremental order of 1 million TPUs could add $10 billion in revenue to Broadcom, $800 million to TTM Technologies, and $500 million to Celestica, showcasing the substantial financial benefits of TPU demand [22][23]. Competitive Landscape - NVIDIA currently commands 70% margins, making it a costly option for companies like Meta, which spends heavily on compute resources [11]. - Google's TPUs, with lower margins, present a more cost-effective alternative, potentially reshaping the competitive dynamics in the AI hardware market [12][24].