Why Employees Are Leaving Google: Beyond Salary

Why Employees Are Leaving Google: Beyond Salary

For decades, a Google offer letter was the closest thing tech had to a gold star. The perks were famous, the pay was rich, and the company still had the aura of a place where the smartest people in the room all happened to work. That aura is thinner now. The keyword question, why employees are leaving Google, has a fairly unromantic answer: money still matters, but speed, influence and mission matter more than they used to.

The market is changing around Google, not just inside it. Anthropic and OpenAI together account for 51% of all interviewing.io coaching requests, while OpenAI alone draws 16% of candidates, roughly the same as Google’s 17% share, according to Pragmatic Engineer in June. AI engineering job openings also rose 60% over the past year at top companies, versus 7% growth for software engineering roles overall, Pragmatic Engineer reported in June. That does not mean Google has fallen apart. It means the field around it has become more crowded, and more interesting.

Why top talent is leaving Google for AI startups

The easiest version of the story is a pay story. It is not wrong, just incomplete. Senior engineers in the US now routinely clear $300,000 in base pay at the 80th percentile, Pragmatic Engineer said in June, so compensation alone does less of the persuading it once did.

Meta reportedly offered some OpenAI and Google DeepMind employees packages worth more than $100 million, and OpenAI CEO Sam Altman said on a podcast that his team had received giant offers but, so far, none of OpenAI’s best people had taken them up, TechCrunch reported in June 2025. Altman’s view was that mission and the chance to reach AGI were doing as much work as the paycheck. That may be true, or at least true enough for the people saying no.

Google can still pay at the top of the market. Pragmatic Engineer said in September 2025 that Google generated $115 billion in profit after all costs and taxes last year, and that staff-level engineers in the US can earn $550,000 to $700,000 in total compensation. The company is not losing talent because it suddenly got cheap. It is losing some talent because other firms now offer a different bargain.

That bargain looks less like a legacy platform job and more like a shot at being inside the next thing before it hardens into the old thing. Pragmatic Engineer found that Anthropic’s two-year retention rate is 80%, while OpenAI’s is 67%. Engineers are not just taking these jobs. They are staying.

Why Google no longer feels like the default dream job

What changed is not only the market for AI researchers. It is the texture of work itself. Former Google product leader Praveen Seshadri described a company where more than 175,000 capable, well-paid employees get very little done because they are stuck inside approval chains, launch procedures, legal reviews, performance cycles, executive reviews, documentation, meetings and reorganizations, Seoul Economic Daily reported in June, citing his February 2023 blog post. Seshadri’s argument was blunt: Google’s core value had shifted from innovation for users to risk avoidance.

That critique matters because it matches the complaint many ambitious engineers quietly make about large companies once the shine wears off. The problem is not that Google lacks smart people or money. It is that a lot of smart people there seem to spend too much time managing the machinery around the work instead of doing the work.

Analysts have made a similar point from the outside. DA Davidson’s Gil Luria said Google was briefly recognized as a leader in AI by holding a cutting-edge model for a few weeks last year, before the gap widened again, Seoul Economic Daily reported in June. That is a telling distinction. Missing the lead entirely is one problem. Losing it after touching it is another.

Mustafa Suleyman, a co-founder of Google DeepMind, made the same basic case after moving to Microsoft. He said Microsoft’s AI division offered a more flexible, startup-like environment than DeepMind, and Seoul Economic Daily reported in June that he recruited at least 24 Google personnel over several months last year. Again, this is not really a salary story. It is a story about where people think they can move faster and matter more.

The Google talent exodus is also a story about missed timing

The cleanest example may be the Transformer saga. After the 2017 paper that introduced the architecture, Noam Shazeer reportedly proposed that Google use Transformers to train a massive network and rethink how the company organizes information, but it did not work, Seoul Economic Daily reported in June, citing Wired. OpenAI then assigned the idea to Alec Radford, which helped lead to the GPT products that kicked off the generative AI frenzy.

That sequence is hard to ignore, even if it should not be flattened into myth. Google did not invent the technology and then blithely walk away from it. But the story does suggest something about how ideas travel once they leave the laboratory and enter a large organization. A promising architecture is one thing. Getting a giant company to move around it is another.

The departure list around that work reinforces the point. Seoul Economic Daily reported in June that, with Shazeer’s departure, all eight original Transformer researchers had left Google. That does not prove causation. It does show how much of the generative AI era was built by people who eventually decided their best work would happen somewhere else.

Why Google still has real advantages

The talent exodus narrative is real, but it is not the whole chart. Google is still one of the strongest places in tech to build a career, and the numbers support that.

Digit/FYI reported in August 2025, citing SignalFire, that nearly 70% of Google engineers stay past the four-year mark, one of the best retention rates in the industry. The same report said Google, Snowflake and Nvidia have been among the most successful companies at keeping engineering talent, while engineers elsewhere often leave faster than they arrive. For most people, Google is still an excellent employer. That is worth saying plainly.

The company is also showing signs that it knows the bureaucracy critique has teeth. CNBC reported in December that Google had cut more than one-third of managers overseeing small teams as part of a broader effort to remove bureaucracy, and that it had conducted employee buyouts as well. CNBC also reported that Google co-founder Sergey Brin has at times personally reached out to prospective candidates, and that the company saw a jump in AI researchers coming from major competitors compared with 2024.

Google is not standing still, in other words. It is trying to make itself feel more nimble without losing the scale that makes it formidable.

There is also a small but useful counterpoint to the exodus story: some of the talent is coming back. CNBC reported in December that 20% of Google’s software AI engineer hires in 2025 were boomerang employees, with the company confirming that the statistic remained accurate and noting the rise in researchers arriving from competitors. That does not erase the departures. It does suggest that Google still has a pull of its own, especially for people who want its infrastructure without its overhead.

The real competition is over how work feels

The companies winning this talent race are not simply paying more. They are selling a particular working life. Digit/FYI quoted SignalFire as saying engineers are drawn to firms offering “growth without chaos,” places that combine technical challenge, support and peer quality. That phrase gets at the part compensation tables miss.

Google still scores well on resources, infrastructure and peer density. CNBC reported in December that one Google spokesperson said engineers want to work there to keep building groundbreaking products, and Seoul Economic Daily noted that Google is expected to chase OpenAI and Anthropic with world-class cloud infrastructure. Those are real advantages. Nobody serious is pretending otherwise.

But frontier labs have another selling point: the sense that the most consequential AI work is happening there now, not somewhere down the road in a committee packet. That perception can be self-reinforcing. Once enough engineers believe the center of gravity has moved, the rest of the talent market starts to behave accordingly.

What Google’s talent problem really means

The sensible conclusion is not that Google has stopped mattering. It plainly has not. It still pays well, still retains most of its engineers past four years, and still has the kind of infrastructure many rivals can only envy. The company remains a destination.

The sharper conclusion is that Google is no longer the default destination for everyone who wants the best technical career. The rise of OpenAI, Anthropic and other AI startups has given top engineers something they did not have at scale before: a credible alternative to Big Tech that feels faster, tighter and closer to the edge of the field. For some people, that matters more than the brand on the badge.

That is why the current talent movement is bigger than a compensation war. Salary gets people in the door. Culture, speed and the chance to shape the work decide whether they stay. Google still has plenty going for it. It just no longer gets the benefit of the doubt.