As the AI race intensifies, Meta Platforms (Facebook's parent company) is in early talks with Alphabet's Google Cloud to explore using Google's Gemini model to optimize its advertising business. According to The Information, Meta employees proposed fine‑tuning Gemini and the open‑source Gemma model on the company's ad data to improve ad summarization and recommendation accuracy. The negotiations are at an early stage and may come to nothing, but the report highlights bottlenecks in Meta's AI scaling—despite pouring hundreds of billions of dollars, its in‑house Llama series still lags competitors. Meta and Alphabet did not immediately respond to Reuters requests for comment; the development could heighten cooperation and competition tensions between the two ad giants.
Advertising is Meta's lifeline: ad revenue totaled $38.3 billion in Q2 2025, accounting for 97.5% of total revenue. But amid the generative AI wave, advertisers demand smarter personalization, such as dynamically generated ads based on real‑time user behavior. The Information notes Meta faces internal AI scaling challenges: despite investments in infrastructure and talent, its Llama 4 model performs modestly on ad content understanding and cannot efficiently process massive user data. Last month, Meta executives considered partnering with OpenAI to bolster its Meta AI chatbot and app features; the turn to Gemini reflects a pragmatic move beyond its 'open‑source first' strategy.
Negotiation details: Fine‑tuning Gemini to target ad pain points
People familiar with the matter said Meta's team and Google Cloud are focused on integrating Gemini into ad recommendation systems. Proposed measures include fine‑tuning Gemini on Meta's ad datasets to enhance 'content understanding' capabilities—such as automatically summarizing user posts, predicting interest tags, or generating targeted ad creative. This could address a current pain point: Meta's AI trails Google's Gemini 1.5 in accuracy when handling complex social contexts (such as Instagram Stories or WhatsApp group chats), with Gemini 1.5 supporting multimodality at a 95% rate.
The talks are in an early phase and may be limited to a pilot rather than a full license. Investing.com analyzed that if realized, this would mark Meta's first large‑scale borrowing of a rival's AI, similar to its $14.3 billion investment in Scale AI (for a 49% stake), aimed at addressing data‑labeling shortfalls. On X (formerly Twitter), investor Beth Kindig posted: 'Meta turning to Gemini exposes Llama's scaling problems but could unlock new ad growth.' Another user, @MunshiPremChnd, predicted this could kick off a 'Meta‑Google ad alliance' that increases industry efficiency.
Meta's AI ambitions are clear: CEO Mark Zuckerberg announced last month that 2025 capital expenditures will reach $66‑72 billion, mainly for AI data centers. But after internal restructuring, the generative AI team's powers have been curtailed, and new hires (including former OpenAI executives) have complained of bureaucracy. The Hindu reported that Meta views external partnerships with Google or OpenAI as a 'temporary bridge' until Llama 5 matures and it returns to in‑house development.
The ad giants' love‑hate relationship: cooperation or competition?
Meta and Google are the two leaders in digital advertising, with a combined market share exceeding 50% in 2025. Both say AI investment is driving ad growth: Meta's Q2 ad revenue rose 22% year‑on‑year, aided by AI‑generated image tools, while Google uses Gemini to optimize YouTube recommendations. But cooperation is sensitive: Meta has accused Google of search‑ad dominance, and turning to Gemini may be seen as a plea for help.
Similar precedents have emerged. In August, Meta executives discussed using OpenAI models to enhance Meta AI conversations and app features. TechCrunch reported that after Meta's Scale AI investment, OpenAI and Google paused some projects, highlighting ecosystem tensions. Seeking Alpha analyst Chris Ciaccio predicted that if a Meta‑Google deal goes through, ad targeting accuracy could improve by 10–15%, but data‑sharing risks would need to be managed.
Market reaction was muted: on September 25 after‑hours Alphabet shares rose 1.5%, while Meta slipped 0.56%. X user @Bearlovesbull posted, 'Meta borrows Gemini — a new era of ad precision?' But Hacker News discussions expressed worry: 'This exposes Meta AI's lag; has Zuckerberg's open‑source bet failed?'
Broader impact: the AI advertising arms race
The talks reflect an industry trend: AI is shifting from an auxiliary tool to a core engine. The Wall Street Journal reported Meta aims for fully AI‑automated advertising by 2026, but brand safety and creative control remain pain points. Capacity Media analyzes that Meta trails OpenAI (ChatGPT Enterprise) and Google (Gemini integrated with Azure) and needs external help to catch up. The Drum noted that in the 2025 AI browser race, OpenAI is challenging Google Chrome while Meta is betting on Superintelligence Labs.
For advertisers, this could mean more precise targeting and a 10% lift in ROI. But privacy regulation is tightening: the EU's GDPR could scrutinize data fine‑tuning. MLQ.ai predicts that if the partnership is realized, it would accelerate an 'AI ad alliance' and reshape the trillion‑dollar market.
Will Meta's strategy of 'leveraging outside strengths' mark a turning point? In the AI race, collaborations are often short‑term expedients; long‑term success still requires in‑house development. Zuckerberg's next move is worth watching.