When ChatGPT burst onto the scene in late 2022, it surprised the public and the tech world, including the corner office of Google’s headquarters. In a recent interview at Google’s Manhattan offices, CEO Sundar Pichai opened up about the moment he realised the AI race had just started, and despite having declared Google an “AI-first” company back in 2015, Pichai admits the speed of the shift was a surprise.While talking to Fast Company, Pichai remembered his instinctive reaction after seeing OpenAI’s ChatGPT chatbot for the first time: “Wow, this technology is going to diffuse earlier and faster than we were expecting.”He described the feeling as “uncomfortably exciting.” Notably, Google’s own research labs had all the building blocks needed to create an AI chatbot but it was OpenAI that was the first to package it into a product that quickly grabbed the headlines.
‘Code Red ’ and Gemini 3
At that time, reports said that Google had issued ‘Code Red’ and the employees “scrambled” to work on it. Pichai described that the “scramble” eventually led to the creation of Gemini 3, Google’s latest and most powerful series of AI models. Released in late 2025, Gemini 3 Pro has already begun beating rivals from OpenAI and Anthropic in several industry benchmarks.The impact was significant and this time, OpenAI CEO Sam Altman reportedly issued a ‘Code Red’, warned his staff in an internal memo that “the vibes out there” would be “rough for a bit” as Google regained its momentum.
Pichai on Google’s strategy
This is not the first time that Pichai has talked about ChatGPT launch. Speaking at Salesforce’s Dreamforce conference last year, Pichai acknowledged that OpenAI deserved credit for releasing their chatbot first, even though Google had been developing similar technology internally.“We knew in a different world, we would’ve probably launched our chatbot maybe a few months down the line,” he added, noting the product hadn’t reached Google’s quality standards at the time.Pichai explained that Google always worked towards a “full-stack” approach to innovation, improving everything from infrastructure to model training and test-time compute. “If you were on the outside, it would look like we were quiet, or we were behind, but we were putting all the building blocks in place, and then executing on top of it,” he said.
