After closing out its longest monthly losing streak in over a decade, Alphabet has opened September with a flurry of developments that could help reset the narrative around its AI ambitions. The company rolled out a new AI model, scored a legal victory against the Justice Department, and received a high-profile endorsement from Berkshire Hathaway’s incoming CEO.
On Wednesday, Google introduced Gemini 3.8 Flash, its third Flash model in six weeks, touting it as the company’s best reasoning and coding model to date. The new model is optimized for software engineering and multi-step agentic tasks, which are increasingly seen as key battlegrounds for turning AI into enterprise revenue. Google also unveiled Gemini 3.8 Flash Cyber, a specialized cybersecurity model designed to detect and patch software vulnerabilities at “frontier-level performance” while running faster and at a lower cost than larger systems.
The product news comes alongside a federal judge’s decision to reject the Justice Department’s request to force Google to sell its AdX ad exchange. The ruling, which opts for behavioral remedies instead of a structural breakup, follows a separate antitrust case last year in which a judge declined to order the divestiture of Chrome. Wyatt Fore, an antitrust attorney at Schinder Cantor Lerner, called it a “big deal” that courts have now twice declined to break up the company, adding that Google heads into the AI race with “no hands tied behind its back.”
The stock itself has yet to stage a meaningful recovery. Shares rose 0.6% on Wednesday after dropping more than 1% Tuesday, leaving Alphabet slightly lower for the month so far. But the company’s actions suggest a more constructive path forward after a turbulent summer that included concerns about its standing in the AI model race, notable executive departures, and a restructuring inside DeepMind.
Smaller models, bigger strategy
Google is betting that its smaller Flash models — which are cheaper to run and quicker to iterate on than its largest frontier systems — can close the gap with leading rivals on select tasks. Tulsee Doshi, senior director of product management at Google DeepMind, told CNBC the recent Flash models have “really surprised us in positive ways in their performance” and “give us opportunities to lean into them.”
But analysts remain cautious about how much this shifts the competitive balance. Gil Luria of D.A. Davidson said that while the new model keeps Google in contention, it “probably won’t change the fact they are a distant third in the enterprise market,” trailing Anthropic and OpenAI. Luria, who rates the stock a hold, added that from a product standpoint Google appears to remain in the race — but without a dramatic leap forward.
Pricing for Gemini 3.8 Flash is set at 75 cents per million input tokens and $3.75 per million output tokens, matching the introductory rate of its predecessor despite the improvements. Meanwhile, Gemini Enterprise is introducing pay-as-you-go pricing, token discounts of up to 20%, monthly caps on agent spending, and a zero-dollar base subscription tier. In materials provided to CNBC, Google directly targeted Microsoft and Anthropic, arguing that their recurring seat fees and separate product licenses make their offerings more expensive and less flexible.
Scale and endorsements
Google is also leaning into its distribution advantage. Nearly three-quarters of Google Cloud customers are already using its AI products, and Google Cloud CEO Thomas Kurian told CNBC those customers are spending about 50% more than their initial commitments. DeepMind’s Demis Hassabis, speaking at the G20 Innovation meeting on Wednesday, described a future where Gemini acts as a general-purpose coordinator of cheaper, specialized models and agents — a strategy that could make breadth as important as having the single best model. It was Hassabis’s first public appearance since DeepMind’s reorganization last month, which saw him shift from CEO to chairman of the unit.
Google is also highlighting cost advantages in cybersecurity. Doshi said the company is “really excited about being able to provide an offering to defenders that is a fraction of the cost, much faster, while still showcasing that frontier-level performance.” Due to potential misuse, access to Gemini 3.8 Flash Cyber is initially limited to a small set of trusted government and enterprise defenders through Google’s new Fairwind Program.
Berkshire Hathaway CEO Greg Abel offered a public vote of confidence on Wednesday, telling CNBC’s Becky Quick that Alphabet is a “significant player” in AI. Abel said Berkshire’s portfolio companies provide visibility into how AI is being used and the benefits it delivers, which “brought incremental interest” in Alphabet as an AI winner.
The endorsement comes as Alphabet continues to spend heavily on AI infrastructure, betting that growth and market share gains will justify the outlay over the long term. Meanwhile, its core advertising business remains a reliable cash generator, growing 14% in the latest quarter — a foundation that could help fund the company’s AI ambitions without the distraction of a forced divestiture.
Source: www.cnbc.com — https://www.cnbc.com/2026/09/02/google-starts-september-with-ai-momentum-after-long-losing-streak.html
This article is for informational purposes only and does not constitute financial, investment, tax, or legal advice. Do your own research and consult a licensed professional before making financial decisions.
