How Google put together the pieces for its AI comeback

When ChatGPT launches in 2022 Google It’s caught flat-footed, but the launch of Gemini 3 and the Ironwood AI chip this month has experts raving about Alphabet’s AI comeback.

Google kicked off November by unveiling Ironwood, the seventh generation of tensor processing units (TPUs) that it says allow customers to “run and scale the largest, most data-intensive models available.” Last week, Google launched its latest AI model, Gemini 3, saying it requires “less direction” and provides smarter responses than previous models.

Salesforce CEO Marc Benioff captured the excitement of Gemini 3 with his speech on Sunday to mail Despite using OpenAI’s ChatGPT on X every day for three years, he said he wouldn’t go back after using Gemini 3 for two hours.

“This jump is crazy,” wrote Benioff, whose company has partnerships Google, OpenAI And other frontier AI model providers. “Everything is sharper and faster. It’s like the world has changed again.”

Most tech stocks started the week lower; except one: Alphabet.

Shares of Google’s parent company rose more than 5% on Monday, adding to last week’s gain of more than 8%. Warren Buffett’s Berkshire Hathaway disclosed earlier this month that it had a $4.3 billion stake in Alphabet as of the end of the third quarter.

Alphabet shares are up nearly 70% this year and Meta’s It’s up more than 50 percentage points this year and surpassed Alphabet’s market cap last week Microsoft’s.

Despite all this he came Nvidia reported stronger-than-expected revenue and guidance in its third-quarter earnings last week.

“You may be asking why nearly all of the AI ​​stocks we cover have sold off after such good news from Nvidia,” Melius Research analyst Ben Reitzes wrote in a note Monday, citing Nvidia’s positive third-quarter earnings last week. he said. “There’s a real reason to worry, and it’s Alphabet’s ‘AI comeback.'”

However, experts say that although Google seems to have regained the upper hand, its leadership against its rivals in the competitive artificial intelligence market is still very weak.

“Three years ago, they were seen as sort of lost, and there were heated claims that they had lost their way. Sundar is a failure” said Nathanson. “They have a huge advantage now.”

The company suffered a series of AI product setbacks in its early attempts to catch up with OpenAI. In 2024 alone, Google was forced to discontinue its rendering product Imagen 2 for several months after users discovered a series of historical inaccuracies. The launch of AI Overview caused a similar backlash when users realized it gave inaccurate advice, which the company later fixed with additional guardrails.

“There was a lot of incompetence and they were scrambling,” said Gil Luria, managing director of technology research firm DA Davidson.. “But there was technology in the pantry, and it was just a matter of putting it all together and shipping it.”

It is especially notable how quickly Google released Gemini 3 after the spring release of Gemini 2.5, which was already considered an impressive model. Nano Banana’s hyper-realistic rendering features are another notch in Google’s belt. Gemini dethroned ChatGPT to top the Apple App Store in September, after the company first released its rendering tool.

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

Google’s ownership of YouTube and all content on the video platform gives the company an advantage when it comes to training models for creating images and videos.

“The amount of video and current 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 get around this.”

Additionally, Google has successfully incorporated AI models into its enterprise products, increasing sales of the company’s cloud unit. In last month’s third-quarter earnings results, Google reached its first $100 billion quarter, driven by growth in the cloud. The company’s cloud unit, which hosts AI services, showed solid growth and a $155 billion backlog from customers.

And it’s not just AI models. Google also attracts attention with its artificial intelligence chips.

Google says Ironwood is almost 30 times more efficient than its first TPU from 2018. Google’s ASIC chips are emerging as the company’s secret weapon in the AI ​​wars, and have helped the company close recent multibillion-dollar deals with customers like Anthropic.

after a while report Nvidia saw its shares fall 3% on Tuesday after Meta said it could strike a deal with Google to use its TPUs for the social media company’s data centers, prompting the chipmaker to post a response on social media.

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

“The advantage of having the entire stack is that you can optimize your model to run particularly well on a TPU chip and render everything in a more optimally designed way,” Luria said.

The company’s ability to serve AI enterprise customers with TPUs and Google Cloud offerings, as well as its incorporation of Gemini 3 into its consumer products, is fueling Wall Street’s excitement.

Experts who spoke to CNBC said the competitive landscape is broader than a single AI winner, but they added that it’s becoming increasingly expensive for multiple companies to prove success.

Despite these gains, Google is still in fierce competition with other artificial intelligence companies, experts said.

“Having a high-end model for a few days doesn’t mean they’re making as much money as the stock market implies,” Luria said, pointing to Anthropic’s new Opus 4.5 model. It started on Monday.

Earlier this month OpenAI also announced The company said it made two updates to the GPT-5 model to make it “more efficient and easier to understand in everyday use” as well as being “warmer and more chatty by default.”

“The frontier models still look neck-and-neck in some respects,” said Forrester Research’s Gualtieri.

Experts said that given the expense of the AI ​​race, competitive advantage will likely go to companies willing to spend more money. Alphabet, Meta, Microsoft and Amazon removed their capital spending guidance in last month’s earnings reports. Collectively, they expect that figure to rise to more than $380 billion this year.

“These companies spend a lot of money assuming there will be a winner-take-all, when in reality we could end up with frontier models becoming a commodity and many of them becoming interchangeable,” Luria said.

For Google, maintaining leadership in artificial intelligence will not be without its challenges.

Company executives told employees earlier this month that Google needed to double its service capacity every six months to meet demand for AI services and operate its leading models, CNBC reported last week.

“Competition in AI infrastructure is the most critical and also the most expensive part of the AI ​​race,” Google Cloud VP Amin Vahdat told employees.

Although Google’s in-house TPUs are gaining more traction as alternatives to Nvidia’s Blackwell chips, Nvidia still holds over 90% of the AI ​​chip market.

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

Despite getting Salesforce’s Benioff to move to Gemini, Google also has a lot of work to do on its consumer chat product, experts said, citing hallucinations and lower user numbers than OpenAI.

According to Google’s statement last month, the Gemini application has 650 million monthly active users and AI Overview has 2 billion monthly users. By comparison, OpenAI said in August that ChatGPT reached 700 million users per week.

“Yes, Google took action,” Luria said. “But that doesn’t mean they won.”

WRISTWATCH: Plexo Capital’s Lo Toney says the AI ​​narrative is shifting towards Google with its full stack