Google Regains AI Momentum With New Gemini Models and Antitrust Win

Google started September with a series of developments that could improve investor sentiment following a difficult summer. The company launched Gemini 3.8 Flash, introduced a specialized cybersecurity model and avoided a forced sale of its advertising exchange in a major U.S. antitrust case.

Gemini 3.8 Flash is Google’s third Flash model in six weeks, reflecting the rapid pace of development inside DeepMind. The model focuses heavily on coding, reasoning and agentic tasks, areas that AI companies increasingly view as critical to generating enterprise revenue.

Google says the latest model delivers substantial improvements over Gemini 3.7 Flash in software engineering and multi-step tasks. Smaller Flash models are also designed to operate faster and at lower cost than the company’s largest frontier systems while increasingly approaching their capabilities on certain workloads.

Pricing is central to Google’s competitive strategy. Gemini 3.8 Flash costs $0.75 per million input tokens and $3.75 per million output tokens, maintaining the introductory pricing of its predecessor despite improvements in performance.

Google is extending that strategy across Gemini Enterprise with pay-as-you-go pricing, token discounts of up to 20%, spending caps for AI agents and a zero-dollar base subscription option. The company is positioning this flexibility as an alternative to recurring seat fees and separate software licenses offered by competitors.

Scale could provide another advantage. Nearly three-quarters of Google Cloud customers already use its AI products, and Cloud CEO Thomas Kurian said those customers are spending around 50% more than their original commitments.

However, Google still faces questions about its competitive position. D.A. Davidson analyst Gil Luria said Gemini 3.8 Flash helps keep the company in the AI race but may not be enough to close the gap with OpenAI and Anthropic in the enterprise market.

Google is simultaneously extending Gemini into cybersecurity. Its new Gemini 3.8 Flash Cyber model is designed to identify and patch software vulnerabilities with frontier-level performance while operating faster and more cheaply than larger AI systems.

Because those capabilities could potentially be misused, access will initially be restricted to selected government and enterprise cybersecurity defenders through Google’s Fairwind Program. The approach highlights both the commercial potential and security risks associated with increasingly capable AI agents.

Alphabet also received a notable endorsement from Berkshire Hathaway CEO Greg Abel, who described Google as a significant AI player. Berkshire’s view is partly informed by how companies within its portfolio are already using AI and the benefits they are seeing from the technology.

The biggest financial question remains whether Google can generate sufficient returns on its enormous AI infrastructure investments. Its established businesses provide significant support, with Google’s advertising operation still growing 14% in the latest quarter.

That cash-generating business also received an important regulatory victory. A federal judge rejected the U.S. Justice Department’s request to force Google to sell its AdX advertising exchange, choosing behavioral remedies instead of the structural breakup sought by regulators.

The ruling follows another antitrust decision that rejected calls for Google to divest Chrome. Together, the outcomes reduce the immediate threat of regulators dismantling key parts of Alphabet just as competition across AI intensifies.

Alphabet shares have yet to stage a significant recovery, but September has started with a more favorable combination of product launches, enterprise traction and regulatory outcomes. The next challenge is converting those advantages into sustainable AI revenue and demonstrating that massive infrastructure spending can translate into stronger long-term growth.