On July 24, according to news website Axios, six current and former Google DeepMind employees revealed,Low employee morale is one of the reasons why Google's DeepMind AI lab has delayed releasing models.Bloomberg previously reported that the most powerful model Google is developing, the Gemini 3.5 Pro, is months behind schedule. However, the company did recently release a series of smaller, more efficient models to mixed reactions.


Google

Rival executives were quick to sneer at Google's smaller models. Alexandr Wang, the head of Meta AI, posted on X, "Gemini? Never heard of it." Others asked Google executives when a more powerful model would be ready.

Although Google's latest earnings report on Wednesday beat market expectations, the company's free cash flow has turned negative.This is largely due to its AI investments: Google's total spending this year is expected to reach $190 billion.


Gemini

Bloomberg Intelligence analyst Mandeep Singh said that Google's cloud business revenue increased by 82%, but search business revenue was slightly lower than Wall Street expectations, which heightened market concerns about whether the company's AI capital expenditures can bring returns.

brain drain

Google AI's strength is considered to have declined by the outside world, just as the competition for AI talents is intensifying. This has the most serious impact on Google.

Currently, many of Google's top researchers have left their jobs to join competitors.

Among those departures is Noam Shazeer, an influential researcher and Gemini co-lead who now joins OpenAI. Nobel Prize winner in chemistry John Jumper, he switched to Anthropic.

Military deal sparks dissatisfaction

Several employees have publicly said they resigned over a deal Google reached with the U.S. Department of Defense in April. The agreement allows the U.S. military to use Google's technology.

The agreement is often brought up in employee exit interviews, according to people familiar with the matter who spoke to Axios on condition of anonymity.

“We’re falling behind,” one employee told Axios, referring to the gap between the company’s model capabilities and those of its competitors.

One person familiar with the matter said it was an "ongoing battle" that led to "emotional burnout" for those who were morally opposed to the Pentagon deal.

Former DeepMind research scientist Alex Turner resigned over Google's military contract. He told Axios that while DeepMind CEO Demis Hassabis released an AI security framework last week, his presence within the company has not been as consistent and frequent as that of rival CEOs Sam Altman and Dario Amodei.

Falling behind

Some people familiar with the matter attributed recent declines in morale among Google employees to the company's military protocols, but others said the burnout stemmed from feeling always one step behind competitors.

A DeepMind employee who participated in model training said that as Google raced against time to protect its search business and respond to the challenge of ChatGPT, Google did not prioritize the development of agent programming.

In the ranking of large language models tracked by OpenRouter, the Gemini series models failed to rank among the top ten.

respond

Google denies that employee morale issues are causing its AI models to lag, or that a large number of employees are leaving due to the company's contract with the Pentagon.

Google said its AI talent attrition rate in the first half of this year was lower than the same period last year. At the same time, more than 90% of those hired for Google AI positions accepted the position.

While the new models Google released this week aren't class-leading in terms of performance, launching a more cost-effective model does fill a clear need in the market.

"We're pleased with the Flash family of models released this week, our future product roadmap, and the overwhelming demand for our models," a Google spokesperson told Axios.

In the AI ​​competition, it is normal for model providers to surpass each other. Leads tend to change quickly and last only a short time.

Just last year, Google was still in the spotlight. However, with the introduction of Mythos, Anthropic seemed unbeatable. Now, OpenAI looks set to be next in line, with plans to introduce its new model to lawmakers next week.

However, Google can afford to be briefly overtaken by competitors, but it cannot afford to lose talent, nor can it afford investors to lose patience.