openai explores google's tpus: nvidia's ai dominance at risk?

2025-07-01 nvidia

San Francisco, Tuesday, 1 July 2025.
openai is testing google’s tpus for chatgpt, aiming for cost savings. this move challenges nvidia’s gpu dominance in ai. while openai denies a full shift, tpu tests show 20-30% lower costs per query. openai still uses nvidia and amd chips, plus it is developing its own silicon. google’s tpus are also used by apple and ai startups. a significant shift could impact nvidia’s revenue, as openai is a major gpu consumer. google is not offering its most powerful tpus to openai.

nvidia’s market position

Nvidia has been a dominant player in the AI chip market, with its GPUs widely used for both training and inference [1][5]. OpenAI’s potential shift to Google’s TPUs raises concerns about Nvidia’s competitive advantage [5]. Experts suggest that if Google successfully attracts more cloud providers to adopt TPUs, the AI chip landscape could evolve from a single dominant player to a more multi-polar competitive environment [5]. Meta is also reportedly considering adopting TPUs, further indicating a possible shift in the chip landscape [5].

revenue and cost dynamics

OpenAI’s move is driven by the high costs and supply constraints associated with Nvidia’s GPUs [5]. OpenAI’s AI server expenses exceeded $4 billion in 2024 and are projected to approach $14 billion in 2025 [5]. By using Google Cloud TPUs, OpenAI anticipates reducing inference costs [2][6]. Early tests indicate that TPUs can handle ChatGPT-scale workloads with 20–30% lower costs per query compared to Nvidia GPUs [1]. This cost efficiency could make TPUs a more attractive alternative for AI applications [6].

google’s tpu strategy

Google has been expanding the external availability of its Tensor Processing Units (TPUs), which were initially designed for internal use [3][6]. These TPUs have attracted customers including Apple, Anthropic, and Safe Superintelligence [2][3]. Google is offering TPUs to OpenAI, though not the most powerful versions, retaining the cutting-edge chips for its own Gemini models [2][3]. This strategy allows Google to compete more effectively in the AI infrastructure market while supporting its internal AI initiatives [1][6].

analyst perspectives on alphabet

Analysts have a positive outlook on Alphabet (GOOGL), with a ‘strong buy’ consensus rating [2]. The average price target for GOOGL is $200.06, suggesting a 12.1% upside potential [2]. Morgan Stanley has also released a research report supporting Google, suggesting that a confirmed agreement with OpenAI would demonstrate Google’s confidence in its long-term search business and accelerate the development of Google Cloud [6]. Year-to-date, GOOGL stock is down about 6% [2].

Bronnen

nvidia ai chips