Remember when ChatGPT dropped in 2022 and everyone said Google was toast? The tech giant fumbled hard. AI image tools generated historically inaccurate pictures. Its AI Overviews feature told people to eat rocks.

Fast forward to today. Google just launched two game-changing products that have Wall Street buzzing and rivals scrambling. Alphabet shares surged 5% Monday while most tech stocks tanked. Plus, Warren Buffett quietly bought $4.3 billion worth of the company’s stock.

So what changed? Let’s break down how Google assembled the pieces for its AI comeback—and why the battle’s still far from won.

Gemini 3 and Ironwood Changed Everything

Google rolled out two major products this month that signal it’s back in the fight.

First came Ironwood, the company’s seventh-generation tensor processing unit. These AI chips let customers run massive models that would choke competing hardware. Google claims Ironwood runs 30 times more efficiently than its first TPU from 2018.

Then Google launched Gemini 3 last week. The new AI model needs less hand-holding than older versions and delivers sharper answers. Salesforce CEO Marc Benioff captured the excitement perfectly. After using ChatGPT daily for three years, he switched to Gemini 3 and declared he wasn’t going back.

“The leap is insane,” Benioff wrote on X. “Everything is sharper and faster. It feels like the world just changed, again.”

That’s not marketing hype from a Google employee. That’s the CEO of a massive enterprise software company whose business partners with OpenAI, Google, and other AI providers.

Wall Street Finally Sees What Google’s Been Building

Ironwood runs 30 times more efficiently than first TPU from 2018

Alphabet shares jumped nearly 70% this year. The stock outperformed Meta by more than 50 percentage points and recently pushed past Microsoft in market capitalization.

Investors are betting Google finally cracked the code. But it took a while to get here.

“Three years ago, they were seen as kind of lost and there were all these hot takes saying they lost their way,” said Michael Nathanson, co-founder of equity research firm Moffett Nathanson. “Now, they have a huge leg up.”

The pieces were always there. Google just struggled to assemble them into products people actually wanted to use. The company owns YouTube and all its training data. It runs Google Cloud, which serves enterprise customers. Plus, it had been developing AI chips for years.

However, shipping functional AI products proved harder than expected. Google pulled its Imagen 2 image generator for months after users discovered it created historically inaccurate pictures. AI Overviews gave terrible advice that went viral for all the wrong reasons.

“There was a lot of fumbling, and they were scrambling,” said Gil Luria, managing director at technology research firm DA Davidson. “But they had the tech in the pantry, and it was just a matter of getting it all together and shipped.”

YouTube Data Gives Google a Secret Weapon

Google’s ownership of YouTube matters more than most people realize. The platform hosts billions of videos covering nearly every topic imaginable. That’s training data competitors can’t easily replicate.

“The amount of video and current data that Google has, that’s really a huge competitive advantage,” said Mike Gualtieri, vice president and principal analyst for Forrester Research. “I don’t see how OpenAI and Anthropic can overcome that.”

This advantage shows up clearly in image and video generation capabilities. After Google launched its image generation tool Nano Banana, the Gemini app shot to number one in Apple’s App Store. It dethroned ChatGPT in September and held that position through the launch of Nano Banana Pro last week.

Salesforce CEO switched from ChatGPT to Gemini 3 after three years

Meanwhile, Google’s cloud business is booming. The company hit its first $100 billion quarter last month, boosted by cloud growth. Google Cloud’s AI services showed solid expansion and customers signed deals creating a $155 billion backlog.

That’s enterprise customers betting big on Google’s AI infrastructure.

Google’s Chips Challenge Nvidia’s Dominance

Ironwood represents more than just another chip upgrade. Google’s tensor processing units are emerging as legitimate alternatives to Nvidia’s GPUs.

These custom chips optimize specifically for Google’s AI models. That gives the company control over the entire stack from silicon to software. Plus, customers like Anthropic signed multi-billion dollar deals to use Google’s TPUs instead of Nvidia hardware.

Reports suggesting Meta might strike a TPU deal with Google sent Nvidia’s stock down 3% Tuesday. The chipmaker felt threatened enough to post a defensive response on social media.

“The advantage of having the whole stack is you can optimize your model to work specifically well on a TPU chip and you’re building everything to a more optimally designed,” Luria said.

Nvidia still controls over 90% of the AI chip market. But Google’s gaining ground fast. And unlike Nvidia’s general-purpose GPUs, Google’s chips integrate tightly with its cloud services and AI models.

That integration matters for customers tired of stitching together tools from multiple vendors.

The Race Remains Brutally Competitive

Google struggled to assemble pieces into products people wanted to use

Google’s recent wins don’t mean it’s won the AI war. Not even close.

OpenAI updated its GPT-5 model this month to make it “warmer by default and more conversational.” Anthropic launched its new Opus 4.5 model Monday, directly competing with Gemini 3’s capabilities.

“Having the state of the art model for a few days doesn’t mean they’ve won to the extent that the stock market is implying,” Luria warned.

The frontier AI models remain “neck and neck,” according to Gualtieri. One company might edge ahead temporarily, then another leapfrogs with its next release. This pattern repeats every few months.

Moreover, the costs are astronomical. Alphabet, Meta, Microsoft, and Amazon collectively expect to spend more than $380 billion on AI infrastructure this year. Each company lifted its capital expenditure guidance last month.

“These companies are spending a lot of money assuming there’s gonna be a winner take all when in reality we may end up with frontier models being a commodity and several will be interchangeable,” Luria said.

Google faces specific challenges maintaining its momentum. Company executives told employees Google must double its serving capacity every six months to meet AI demand. That’s an expensive treadmill to stay on.

“The competition in AI infrastructure is the most critical and also the most expensive part of the AI race,” Google Cloud Vice President Amin Vahdat told employees.

Consumer AI Still Belongs to ChatGPT

Despite winning over enterprise customers and Salesforce’s CEO, Google trails badly in consumer chat adoption.

Alphabet shares jumped 70 percent and outperformed Meta by 50 points

The Gemini app has 650 million monthly active users. AI Overviews reaches 2 billion monthly users. Those sound like big numbers until you compare them to ChatGPT’s 700 million users per week.

OpenAI owns the consumer mindset around AI chat. When normal people want to ask an AI something, they go to ChatGPT. Breaking that habit will take more than a technically superior model.

Plus, Gemini still hallucinates and makes mistakes. The gap between Google and OpenAI narrowed significantly, but it hasn’t disappeared.

Nvidia made this point in its Tuesday social media post. While defending its chips against TPU competition, Nvidia emphasized that its hardware remains more flexible and powerful for most AI workloads.

Google’s ASIC chips like Ironwood optimize for specific tasks. That’s great for Google’s own models but limits usefulness for customers running different AI systems.

Google Assembled the Pieces. Now Comes the Hard Part

Google proved it can compete in the AI race. The company shipped impressive products that work well and serve both enterprise and consumer customers. Investors rewarded this progress with surging stock prices and Warren Buffett’s vote of confidence.

But the finish line keeps moving. Every AI company is spending billions to train larger models, build more chips, and acquire more customers. Nobody’s pulling away from the pack yet.

Google’s YouTube data, TPU chips, and cloud infrastructure create real competitive advantages. However, maintaining technological leadership requires constant innovation at tremendous cost.

The company that stumbled out of the gate in 2022 is now running strong. Whether it can maintain this pace for years remains the billion-dollar question.