https://www.geeknetic.es/Noticia/36676/El-CEO-de-Salesforce-abandona-ChatGPT-tras-tres-anos-de-uso-la-llegada-de-Gemini-3-marca-el-punto-de-inflexion-para-Google.html

OpenAI’s hegemony in the generative artificial intelligence market, undisputed since the launch of ChatGPT, appears to be facing his first real big challenge. Marc Benioff, CEO of Salesforce and undoubtedly an influential voice in the business technology sector, has publicly announced that he has stopped using the OpenAI chatbot after three years of daily use. Its new flagship tool is Gemini 3, the latest model presented by Google.

Benioff’s change is not merely anecdotal, but reflects a tectonic shift in the industry. According to the executive, the difference in performance between both models is abysmal, going so far as to describe it as an “incredible leap.” Among the determining factors for this migration, Benioff highlights the superior reasoning abilityresponse speed and advanced multimodal capabilities, especially in video and image processing, areas where Google’s new model has demonstrated surprising technical solvency.

Google recovers lost ground with its own hardware

This move validates Google’s long-term strategy, which seems to have woken up from its initial slumber. It is important to remember that, although OpenAI was the one who capitalized on the boom in generative AI, the base technology on which these models are based, the Transformer architecturewas originally developed by Google researchers. For the past two years, the Mountain View company has had to row against the current to match a market that they helped found, but the launch of Gemini 3 suggests that they have not only caught up with their rival, but in key ways they could be surpassing it.

Google’s competitive advantage in this new stage is not limited only to the software or the quality of its algorithms. The company has a fundamental strategic asset, namely its hardware infrastructure. Unlike OpenAI, which relies on third-party infrastructure, Google has been designing and perfecting its own Tensor Processing Units, known as TPUs, for years. These chips, designed specifically for machine learning, allow Google to train and run its models with efficiency and vertical integration that its competitors have difficulty replicating.