Google starts September with AI momentum after long losing streak

After ending Wall Street’s longest monthly losing streak in more than a decade. google started September by giving investors some reason to be optimistic.

The company on Wednesday launched Gemini 3.8 flash, its third flash model in six weeks, alongside new cybersecurity models aimed at trusted government and enterprise customers. Berkshire Hathaway CEO Greg Abel also offered a public vote of confidence in Alphabet’s AI position.

Meanwhile, a federal judge on Wednesday rejected a Justice Department push to force Google to sell its ad exchanges, marking another victory for the company against government antitrust laws.

Taken together, these developments put Alphabet in a more positive position after a difficult summer that saw Google lose momentum in the AI ​​model race, high-profile departures, and a major reorganization within DeepMind.

Stock prices have yet to show a major rebound. After falling more than 1% on Tuesday at the start of the month, it rose 0.6% on Wednesday, so it is still slightly down in September.

However, Google is enthusiastic about Gemini 3.8 Flash and focuses on coding and agent tasks. These are two areas where AI companies are increasingly trying to translate technological advances into corporate profits. Google calls it Gemini’s best reasoning and coding model to date, and it’s a significant improvement over 3.7 Flash for software engineering and multi-step tasks.

Tulsee Doshi, senior director of product management at Google DeepMind, told CNBC that recent Flash models have been “really surprising in a good way in terms of performance,” adding, “It’s given us an opportunity to rely on Flash models.”

Smaller Flash models are cheaper to run and iterate faster than the company’s largest frontier systems, but they’re getting closer and closer for some tasks. But that’s not enough to move the company closer to Anthropic and OpenAI, DA Davidson analyst Gil Luria said in an interview.

“From a product perspective, this model seems to keep Google competitive, but it probably won’t change the fact that Google is a distant third in the enterprise market,” said Luria, who recommends owning the stock.

Price is key to Google’s sales pitch.

Gemini 3.8 Flash costs 75 cents per million input tokens and $3.75 per million output tokens. This is the same introductory price as the last Flash model, even though the new model offers improvements in coding, agent tasks, and inference.

Gemini Enterprise adds pay-as-you-go pricing, up to 20% token discounts, monthly caps on agent spending, and a zero-dollar base subscription option. According to documents provided to CNBC, Google also directly targeted microsoft Anthropic claims that recurring seat fees and separate product licenses result in higher product prices and less flexibility.

Google also has economies of scale. Nearly three-quarters of Google Cloud customers are already using the company’s AI products, and Google Cloud CEO Thomas Kurian told CNBC that these customers are spending about 50% more than their original contract.

DeepMind’s Demis Hassabis similarly spoke about the future impact of his strategy. Hassabis told the G20 Innovation Conference on Wednesday that Gemini could increasingly play a role as a general-purpose layer coordinating cheaper specialized models or agents. In that world, breadth can be just as important as having the best model.

This is the first time Hassabis has been heard speaking publicly since DeepMind’s reorganization last month. As part of the announcement, Mr. Hassabis has been moved from DeepMind’s CEO role to chairman of the division.

Google also touts the cost of cybersecurity. The company says Gemini 3.8 Flash Cyber ​​can detect and patch software vulnerabilities with frontier-level performance, while running significantly faster and more affordably than larger models.

“We’re really excited to be able to serve defenders much faster at a fraction of the cost while delivering frontier-level performance,” Doshi said.

Because these features can also be exploited, Google is initially restricting access to a small group of trusted government and corporate defenders through the new Fairwind program.

There are trade-offs for investors. Alphabet is spending huge amounts of money on AI infrastructure, so it’s counting on growth and increased market share to deliver long-term benefits.

Berkshire’s Abel told CNBC’s Becky Quick on Wednesday that he sees Alphabet as a winner in the AI ​​space, based in part on how its portfolio companies are using its technology.

“We started getting a lot more information internally about how we were using AI and what benefits it was bringing us, so there was a lot of interest and we recognized Google as an important player,” Abel said.

While Google’s position in AI is considered shaky by some market experts, its advertising business remains strong, growing 14% in its most recent quarter.

This cash engine gained further momentum on Wednesday.

A federal judge overseeing the Justice Department’s ad tech antitrust case has ruled that Google doesn’t have to sell AdX Exchange, opting for behavioral relief rather than the structural dismantling that regulators had sought. The decision follows another antitrust ruling last year in which a judge rejected a request to force Google to sell Chrome.

Wyatt Foer, an antitrust lawyer and partner at Synder Kanter Lerner, told CNBC that the court’s decision to oppose the breakup in both cases was a “major event,” adding that Google is competing in the AI ​​race “without a shackle on its back.”

WATCH: Google’s Pixel 11 has Gemini Intelligence at the heart of the smartphone