Gemini 3 and Ironwood mark a shift after early stumbles in the AI race
Google has reasserted itself in the artificial intelligence market with the release of its Gemini 3 model and its seventh-generation Ironwood tensor processing unit, drawing sharp reactions from industry leaders and analysts. When OpenAI launched ChatGPT in 2022, Google was widely viewed as unprepared. This month’s developments, however, prompted renewed attention to Alphabet’s AI strategy.
Google began November by introducing Ironwood, a new TPU designed to run and scale large, data-intensive models. Last week, the company released Gemini 3, which it said requires less prompting and delivers more accurate responses than earlier versions. Salesforce CEO Marc Benioff wrote on X that after two hours with Gemini 3, he was no longer using ChatGPT, calling the change “insane.”
Alphabet shares rose more than 5% on Monday following an 8% gain the week before. Berkshire Hathaway disclosed a $4.3 billion position in Alphabet earlier this month. Alphabet’s stock is up nearly 70% this year, outpacing Meta by more than 50 percentage points, and briefly surpassed Microsoft’s market cap last week.
Analysts call Google’s gains significant but caution the lead is narrow
Melius Research analyst Ben Reitzes wrote that a primary concern for investors is Google’s “AI comeback.” Michael Nathanson of Moffett Nathanson said CEO Sundar Pichai has now assembled Google’s AI components effectively, positioning the company to serve both consumers and enterprises. Analysts said Google’s early missteps — including issues with the Imagen 2 image generator and the initial launch of AI Overviews — slowed its response to rivals.
Gil Luria of DA Davidson said that although Google “was scrambling,” it already possessed the underlying technology and needed to align its releases. The rapid turnaround from Gemini 2.5 in the spring to Gemini 3, along with the introduction of Nano Banana image features and last week’s launch of Nano Banana Pro, helped fuel market enthusiasm.
YouTube data and cloud performance strengthen Google’s positioning
Google’s ownership of YouTube provides substantial training data for image and video generation. Mike Gualtieri of Forrester Research said the volume of video and recent content gives Google an advantage that competitors may find difficult to match.
Google’s AI models have also been integrated into its enterprise products, driving cloud adoption. Alphabet reported its first $100 billion quarter last month, supported by Google Cloud’s growth and its AI services backlog of $155 billion.
Ironwood TPUs gain traction amid industry-wide chip shortages
Google said Ironwood is nearly 30 times more power-efficient than its first TPU from 2018. The ASIC-based chips have contributed to multibillion-dollar deals, including with Anthropic. Reports that Meta may use Google’s TPUs caused Nvidia’s stock to fall 3% on Tuesday before the company issued a public response.
The rise of competing AI chips comes at a time when Nvidia still holds more than 90% of the market. Nvidia noted that its chips remain more flexible than ASIC designs, which typically serve specific workloads.
Competition remains intense across frontier models
Analysts said Google’s consumer chat experience still trails OpenAI’s on user metrics. Google said Gemini has 650 million monthly active users and AI Overviews has 2 billion monthly users. OpenAI reported in August that ChatGPT reached 700 million weekly users.
Anthropic launched its Opus 4.5 model this week, and OpenAI recently updated its GPT-5 model to improve conversational behavior and ease of use. Analysts said top models remain close in capability and that competing companies are increasing capital expenditures to sustain the race. Alphabet, Meta, Microsoft, and Amazon expect combined capital spending to exceed $380 billion this year.
Infrastructure demands present ongoing challenges
Google executives told employees this month that the company must double serving capacity every six months to support demand for AI services and run its frontier models. Google Cloud Vice President Amin Vahdat said infrastructure is the most resource-intensive aspect of the AI competition.
Luria said Google has made significant progress but emphasized that this does not guarantee long-term leadership.
Featured image credits: cottonbro studio via Pexels
For more stories like it, click the +Follow button at the top of this page to follow us.