Google’s AI comeback gains speed with Gemini 3

• Google gains ground as Gemini 3 wins praise and its silicon draws interest.
• Google’s AI stack gets higher TPU demand, company enters new partnerships.

Analysts and engineers have spent the past few years saying Google was falling behind in the push to build advanced artificial intelligence. Even some of its own former leaders shared that view. But the story has shifted.

As reported by BloombergGoogle has rolled out new AI software and signed major partnerships, including a chip deal with Anthropic PBC, that have calmed earlier worries about it losing ground to OpenAI and other competitors. Its newest model, the Google Gemini 3, has drawn praise for handling reasoning and coding tasks, and for performing well on the kinds of tricky prompts that often confuse chatbots. Google’s cloud unit is also gaining momentum, supported by global interest in AI development and the heavy demand for computing power.

There are also signs that Google’s custom AI chips are attracting more attention. They are one of the few large-scale alternatives to Nvidia’s hardware. A report that Meta Platforms is in talks to use Google chips helped push Alphabet’s stock higher this week. The company has gained nearly US$1 trillion (RM4.1 trillion) in market value since mid-October, helped by both Warren Buffett’s US$4.9 billion (RM20.3 billion) investment and Wall Street’s confidence in Google’s AI plans.

On Tuesday, Alphabet shares rose 1.5% to US$323.44 (RM1,337), bringing its market value close to US$4 trillion (RM16.54 trillion). SoftBank Group, a key OpenAI investor, dropped 10% on concerns about rising competition from Gemini. Nvidia shares slipped 2.6%, wiping out US$115 billion (RM475.5 billion) in value.

“Google has arguably always been the dark horse in this AI race,” said Neil Shah, analyst and co-founder of Counterpoint Research. He called it “a sleeping giant that is now fully awake.”

A revived push in Google’s stack

For years, Google leaders said their deep investment in research would help them protect their search business and build the next wave of computing. That confidence was shaken when ChatGPT arrived and created the first real challenge to Google search in a long time – even though much of the core technology began at Google. But the company still has advantages: vast amounts of training data, strong profits, and the ability to run its own large computing systems.

“We’ve taken a full, deep, full-stack approach to AI,” Sundar Pichai, CEO of Google and Alphabet, told investors last quarter. “And that really plays out.”

Concerns about regulatory pressure are easing too. Google avoided the harshest outcome in a recent US antitrust case, in part because the rise of new AI competitors weakened arguments that it held unbreakable control over search. At the same time, Google is showing gradual progress in expanding beyond search. Waymo, its autonomous driving unit, is moving into more cities and recently added freeway trips to its robotaxi service – a milestone made possible by heavy research and investment.

A major part of Google’s advantage comes from how much of the computing stack it controls. It builds the apps people use, like its Nano Banana image generator, designs AI models, runs cloud infrastructure, and it makes its own chips. Google also holds massive amounts of data from search, YouTube, and Android devices – training material the company rarely shares. That gives it more control over its AI products and reduces its reliance on outside suppliers.

Rising interest in Google’s chips

Other companies, including Microsoft and OpenAI, have been trying to develop or source their own chips to reduce their dependence on Nvidia. Google was long the primary user of its homegrown tensor processing units, or TPUs. The chips were created more than ten years ago to speed up search results and now support modern AI workloads. That is changing as outside companies begin using Google hardware. In October, Anthropic said it could use up to one million Google TPUs in a deal worth tens of billions of dollars.

This week, The Information reported that Meta plans to run Google chips in its data centres starting in 2027. Google did not confirm the report but said its cloud business is seeing rising demand for both TPUs and Nvidia GPUs. “We are committed to supporting both, as we have for years,” a spokesperson said.

Nvidia struck a friendly tone. “We’re delighted by Google’s success,” a spokesperson said Tuesday. “They’ve made great advances in AI, and we continue to supply to Google.” The spokesperson added that Nvidia remains “a generation ahead of the industry.”

Analysts said the Meta news shows Google’s chip strategy is working. Ben Barringer, head of technology research at Quilter Cheviot, wrote that “many others have failed in their quest to build custom chips, but Google can clearly add another string to its bow here.”

What’s lifting Google’s AI standing now

Google’s AI unit reached this point by taking risks. In early 2023, the company consolidated its AI teams under Demis Hassabis, who leads DeepMind. The shift had rough moments, including a flawed release of an image-generation tool. DeepMind had also spent years on research projects – like protein-folding – that shaped new business opportunities and earned a Nobel Prize, but did little to lift Google’s profits. Under the restructuring, the group is now tightly focused on foundational models that can compete with OpenAI, Microsoft, and others.

Hassabis has helped convince top AI engineers to stay despite large offers from competitors. Pichai has also been willing to spend heavily to keep talent in-house.

The Google Gemini 3 Pro now sits near the top of major AI rankings, like LMArena and Humanity’s Last Exam, and OpenAI co-founder Andrej Karpathy called Gemini “clearly a tier 1 LLM.” Google has framed the model as one that can handle advanced science and maths problems, and fix long-standing issues, including spelling mistakes in image-text outputs.

Measuring interest from everyday users is more difficult. Google said last week that 650 million people use the Gemini app, yet OpenAI reported recently that ChatGPT has 800 million weekly users. Sensor Tower estimates that Gemini had 73 million monthly downloads in October, compared with 93 million for ChatGPT.

Stronger footing in the AI race

Google remains dominant in advertising, but finding other reliable revenue streams has been difficult. Its cloud business brought in US$15.2 billion (RM63 billion) in the third quarter, up 34% year over year. Even so, it still trails behind Amazon Web Services and Microsoft Azure, which both brought in more than twice that amount. Shah of Counterpoint Research said Google’s AI adoption in large companies still lags Microsoft and Anthropic.

OpenAI, meanwhile, is building hoping to turn a profit by selling paid versions of ChatGPT and related tools. It has also formed chip partnerships with Broadcom, AMD, and Nvidia.

For now, Google’s TPUs appeal most to companies with huge computing needs, like Meta and Anthropic, said Meryem Arik, CEO of Doubleword. She also noted that the AI chip market isn’t a winner-take-all arena.

The catch is that Google’s chips are tied to Google Cloud, and developers have more flexibility when they use Nvidia GPUs. “As soon as you use TPUs, you’re locked into the Google cloud ecosystem,” Arik said. That kind of lock-in might have been a red flag in the past. Today, Google’s progress in AI has made it less of a concern.

“It’s definitely fair to say that Google is back in the game with Gemini 3,” said Thomas Husson, analyst at Forrester. “To paraphrase a quote attributed to Mark Twain, reports of Google’s death have been widely exaggerated.”

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events, click here for more information.

AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.