At Google, leaders are worried about falling behind in the race to offer AI coding tools, especially as rivals like Anthropic PBC offer more effective and popular tools to businesses, according to people familiar with the matter. The search giant is currently working to consolidate some of its coding initiatives under one banner to accelerate progress and capitalize on growing customer interest.
Concerns about the company’s position are growing at parts of Alphabet Inc.’s Google, particularly its AI research lab DeepMind, said current and former employees and executives, who asked not to be named because they are not authorized to speak publicly. Companies are realizing that with AI coding tools, anyone can instruct chatbots to build products. But Google doesn’t have a clear solution for them. The features of the company’s Gemini model are scattered across six different coding products with varying branding, illustrating how the company’s lack of focus and competing internal efforts are hampering its success, the people said.
Even within the company, some Google engineers prefer to use Anthropic’s Claude Code. Even more worrying, the people said, are engineers who are having a hard time implementing AI coding at all.
Google has made some efforts to reduce internal confusion over priorities. Chief AI architect Koray Kavukcuoglu will work with Google’s core engineering teams to integrate the company’s artificial intelligence coding tools under Antigravity, a platform released last year, in the coming weeks, a spokesperson said.
DeepMind has formed a new team led by research engineer Sebastian Borgeaud and is devoting more resources to AI coding, according to a former Google employee. The new team was previously reported by The Information. Google DeepMind CEO Demis Hassabis and John Jumper, who won the Nobel Prize in 2024, are also working on AI coding, according to people familiar with the matter.
Late last year, Google released Gemini 3, a model that outperformed competing services across a variety of benchmarks and was widely considered to be dominant in AI. But in recent months, Anthropic and OpenAI have gained business momentum by focusing on a lucrative market for products that speed up software development by streamlining the process of writing and debugging code.
“Coding is actually the easiest way to make money,” says Keith Zhai, co-founder of TinyFish, a startup that develops web agents. Many engineers in the Valley go back and forth between Claude Code and OpenAI’s Codex to see which program yields the best results, but Google is often not part of the conversation, he added.
Google still has plenty of reasons to be confident in its position. The company has made great strides in the quality of the underlying models underlying its coding tools, and has deep pockets and considerable computing power.
“In-house coding tools like Antigravity have seen tremendous adoption since we introduced them in recent years, and their use has accelerated the development of our models and AI tools,” a Google spokesperson said in a statement. Meanwhile, Google has been eager to tout the speed at which its internal culture is changing. Alphabet announced in February that about 50% of its new code is written by AI.
But Silicon Valley engineers are rapidly embracing AI coding, so even a moment’s delay in the market could have significant repercussions. Raj Gajwani, a former Google executive who is now chief business officer at startup OpenArt AI, says there is a growing belief in the industry that coding is not just a lucrative early application of AI, but the key to building software that rivals human capabilities. “From a computer science perspective, if you win in coding this year, you have the raw data you need to win in modeling ability next year,” he said.
Google’s focus on its own technology also complicates efforts to catch up. Most employees are prohibited from using competing tools such as Claude Code and Codex due to security concerns, but Googlers can request exceptions if they can prove there’s a business case, one former employee said. According to three former employees, some teams at DeepMind, including those working on Gemini models, internal applications, and open source models, use Claude code.
“Even within Google, we want the best people to use the best tools,” said one former employee.
Wired magazine reported that Anthropic cut off access to OpenAI’s models last year. Google is investing billions of dollars in Anthropic. An Anthropic spokesperson did not immediately respond to a request for comment.
In recent years, DeepMind has sought to gain more control over how AI breakthroughs are incorporated into Google products. Last year, Google appointed Kavukcuoglu to a new role as chief AI architect, where he is responsible for building generative AI into Google products. However, there is still confusion about who is taking the lead in AI coding. In addition to DeepMind, Google Cloud, Google Core, Google Labs and Android are all promoting AI coding in various ways, one of the people said.
Google released its Antigravity platform last year after acquiring the talent and technology from startup Windsurf in a $2.4 billion deal. It joins a motley lineup of Google AI coding tools that includes Gemini Code Assist, Gemini CLI, AI Studio, Firebase Studio, and Jules. Kathy Korevec, who oversaw Jules, left Google for OpenAI earlier this month, according to her LinkedIn profile.
In a post on Social Network Korebetz did not immediately respond to a request for comment.
Former employees say there is a philosophical clash within the Googleplex between AI researchers who want to transition as quickly as possible and traditional senior engineers who have strict standards for code quality. Former employees say the use of AI is taken into account in performance evaluations. But engineers seeking to use in-house AI coding tools often run into capacity constraints due to competition for computing power, former employees said.
Brian Saluzzo, one of the executives who oversaw Google’s internal AI coding efforts, recently left the company. Saluzzo did not immediately respond to a request for comment.
Businesses are still exploring how to best incorporate AI into their workflows, and the variety of products offered gives Google more opportunity to see what’s here to stay. But there are limits to the advantages of incumbents like Google, said Deepti Srivastava, a former Google executive and founder and CEO of AI startup Snow Leopard.
“The market moves too fast for big companies to think about it and then act,” Srivastava said. “Speed is the only moat.”
Love writes for Bloomberg.
