How Google put the pieces together for an AI resurgence

When ChatGPT launches in 2022, google failed, but with the launch of Gemini 3 and Ironwood AI chips this month, experts are praising Alphabet’s AI resurgence.

In early November, Google announced Ironwood, the seventh generation of its tensor processing units (TPUs), which it says will enable customers to “run and scale the largest and most data-intensive models in existence.” And last week, Google announced its latest artificial intelligence model, Gemini 3, which it said provides “fewer prompts” and smarter answers than previous models.

Salesforce CEO Marc Benioff captured the Gemini 3 excitement in a post on Sunday’s X, saying that despite using OpenAI’s ChatGPT every day for three years, he would never go back after using Gemini 3 for two hours.

“This leap is insane,” Benioff wrote. The company has partnerships with Google, OpenAI, and other frontier AI model providers. “Everything is sharper and faster. It feels like the world has changed again.”

Most tech stocks have been down since the beginning of the week, with the exception of Alphabet.

Shares of Google’s parent company soared more than 5% on Monday, adding to last week’s more than 8% rise. Warren Buffett’s Berkshire Hathaway announced earlier this month that it owned $4.3 billion in Alphabet stock at the end of the third quarter.

Alphabet stock has outperformed, up nearly 70% this year. meta It’s up more than 50 percentage points this year and surpassed Alphabet’s market capitalization last week. microsoft’s.

All this happened despite Nvidia Last week, the company reported better-than-expected earnings and guidance for its third quarter.

“With such good news from Nvidia, you may be wondering why nearly all of the AI ​​stocks we cover are selling,” Ben Reitzes, an analyst at Melius Research, wrote in a note Monday, noting Nvidia’s strong third-quarter results last week. “There’s one real reason to be worried, and that’s Alphabet’s ‘AI comeback.'”

But while Google appears to have regained an edge, experts say its lead over rivals in the highly competitive AI market remains narrow.

Michael Nathanson, co-founder of equity research firm Moffett Nathanson, said that with Gemini 3 and Ironwood, Google CEO Sundar Pichai finally seems ready for his company’s AI products. Google serves a wide range of customers, from consumers to businesses, but the company initially struggled with ChatGPT.

“Three years ago, they were kind of seen as lost, and there was a lot of enthusiasm that they had lost their way, that the Thunder were failures,” Nathanson said. “Now they’re stepping up in a big way.”

The company experienced many failures with its AI products in its first attempts to catch up with OpenAI. In 2024 alone, Google had to shut down its image generation product Imagen 2 for several months after users discovered numerous historical inaccuracies. The release of AI Overviews sparked a similar reaction when users realized it was providing incorrect advice, but the company later fixed this with additional guardrails.

“There was a lot of fumbling and confusion,” said Gil Luria, managing director of technology research firm DA Davidson. “But they had the technology in the pantry, so they just put it all together and shipped it.”

Particularly noteworthy is how quickly Google launched Gemini 3 after the spring release of Gemini 2.5, which was already considered a great model. Nano Banana’s ultra-realistic image generation capabilities are on another level from Google. After the company first launched its image generation tool, Gemini shot to the top of the Apple App Store in September, dethroning ChatGPT.

And after the launch of Gemini 3, Google released Nano Banana Pro last week.

Google owns all the content on YouTube and the video platform, giving the company an advantage when it comes to training models for image and video generation.

“The amount of video and up-to-date data that Google has is a really big competitive advantage,” said Mike Gualtieri, vice president and principal analyst at Forrester Research. “I don’t see how OpenAI and Anthropic can overcome that.”

Additionally, Google has successfully incorporated its AI models into enterprise products, boosting sales for its cloud division. Last month’s third-quarter results showed Google reaching $100 billion in its first quarter, fueled by cloud growth. The company’s cloud division, which provides AI services, showed strong growth, with customer backlog reaching $155 billion.

And it’s not just AI models. Google is also attracting attention for its AI chips.

Google says Ironwood is nearly 30 times more power efficient than the first TPU from 2018. Google’s ASIC chips have emerged as the company’s secret weapon in the AI ​​wars, helping it score billions of dollars worth of deals in recent days with customers like Anthropic.

Nvidia stock fell 3% on Tuesday after reports that Meta could strike a deal with Google to use its TPUs in the social media company’s data centers, and the chipmaker posted a response on social media.

With the rise of Google’s TPUs, Nvidia may no longer corner the AI ​​chip market.

“The advantage of having the whole stack is that we can optimize our models to work particularly well on TPU chips, and we can build everything into a more optimal design,” Luria said.

The company’s ability to serve AI enterprise customers with TPU and Google Cloud services, as well as its embedding of Gemini 3 across its consumer offerings, is fueling enthusiasm on Wall Street.

Experts who spoke to CNBC said the competitive landscape is broader than just one AI winner, but added that it is becoming increasingly costly for multiple companies to prove success.

Despite these wins, Google remains in fierce competition with other AI companies, experts said.

“Holding a cutting-edge model for a few days doesn’t mean they’re as winning as the stock market would suggest,” Luria said, referring to Anthropic’s new Opus 4.5 model, which went on sale Monday.

Earlier this month, OpenAI also announced two updates that make GPT-5 models “warmer and more conversational by default,” as well as “more efficient and easier to understand in everyday use,” the company said.

“The Frontier model still appears to be in some ways evenly matched,” said Forrester Research’s Gualtieri.

Given the cost of the AI ​​race, the competitive edge is likely to go to companies that spend more, experts say. Alphabet, Meta, Microsoft, and Amazon each raised their outlook for capital spending in their earnings reports last month. They expect this number to reach more than $380 billion this year.

“These companies are spending huge sums of money on a winner-take-all basis, but the reality is that frontier models may become a commodity and some may become fungible,” Luria said.

For Google, maintaining its lead in the AI ​​space is not without its challenges.

CNBC reported last week that company executives told employees earlier this month that Google would need to double its service delivery capacity every six months to meet demand for AI services and execute its frontier model.

“Competition in AI infrastructure is the most important and most expensive part of the AI ​​race,” Google Cloud Vice President Amin Vahdat told employees.

Google’s homegrown TPUs are emerging as a viable alternative to Nvidia’s Blackwell chips, but Nvidia still controls more than 90% of the AI ​​chip market.

Nvidia noted in a post Tuesday that its chips are more flexible and powerful than ASIC chips, such as Google’s Ironwood, which are typically designed for a single company or function.

And despite prompting Salesforce’s Benioff to switch to Gemini, experts say Google also has a lot of catching up to do with its consumer chat product, citing illusions and a smaller user base than OpenAI.

Last month, Google announced that its Gemini app has 650 million monthly active users and AI Overviews has 2 billion monthly active users. In comparison, OpenAI announced in August that ChatGPT had reached 700 million users per week.

“Yes, Google is working together,” Luria said. “But that doesn’t mean they won.”

WATCH: AI story is shifting to Google with complete stack, says Plexo Capital’s Lo Toney