Why Alphabet’s Gemini 3 Pro is shaking up AI
Archived article: Please note that tax, investment, pension and ISA rules can change and the information and any views contained in this article may now be inaccurate.
Google-owner Alphabet released its latest AI chatbot Gemini 3 Pro to rave reviews on 18 November, raising the stakes in the AI race and challenging the dominance of OpenAI’s ChatGPT.
Salesforce CEO Mark Benioff took to social media to say he is never going back to ChatGPT after using Gemini for just two hours. "The leap in insane. It feels like the world just changed again," enthused Benioff.
The news gave Alphabet’s shares a further boost and follows on from gains related to Berkshire Hathaway announcing a $4.9 billion stake, representing a rare foray for the Omaha-based conglomerate into the technology sector.
Observers suggested Gemini 3 Pro demonstrated superior results compared with rival AI models in its handling of text, images, videos and code while also demonstrating strong advanced reasoning.
In a note to investors Alphabet CEO Sundar Pichai said the Gemini app now surpasses 650 million users per month and is used by more than 70% of its Google Cloud customers.
For context ChatGPT is estimated to have garnered more than 800 million weekly users.
Google is integrating Gemini 3 across all platforms including Search, YouTube, Google Cloud and Android.
Industry experts are hoping Gemini 3 represents a potential inflection point which forces OpenAI and others to innovate faster, leading to a rapid cycle of “leapfrogging” in capabilities, benefiting users.
Alphabet makes push for custom-made AI chips
An important difference with Gemini is that it was built using Alphabet’s own custom-made computer chips, co-designed with Broadcom, specifically made to accelerate machine learning in large language models.
As the chart shows the market is picking up on this new development in the artificial intelligence space with AI infrastructure stocks linked to Alphabet, including Broadcom outstripping those, like Nvidia and Microsoft, with links to OpenAI.
This represents a departure from existing AI models which have so far relied on advanced AI chips designed by Nvidia. Should other large technology companies follow Alphabet’s custom design approach it could see the AI infrastructure market evolve into two distinct groups with Google and Broadcom on one side and Nvidia, AMD, Microsoft and Oracle on the other.
An early sign of the industry moving in this direction was seen on 25 November when reports suggested Meta Platforms was in talks with Alphabet to spend billions of dollars on Alphabet’s custom-made chips for use in its data centres from 2027.
According to Reuters, the conversations involved the possibility of Meta renting chips from Google Cloud as early as 2026.
This would mark a strategic shift for Alphabet as it makes a push to get customers to adopt its custom-made chips, potentially becoming a fully-fledged competitor to Nvidia.
