Meta's AI Push Intensifies in Q2 2025: Rapid Advancements Ahead
In Q2 2025, Meta's advertising business experienced a significant transformation, thanks to the deployment of three advanced AI architectures - Andromeda, GEM, and Lattice. These innovations enhanced ad targeting, relevance, and performance, leading to increased user engagement and robust financial gains.
Andromeda, used for ad retrieval, sifted through tens of millions of ads to select the few thousand most relevant candidates for each user. Enhancements to Andromeda in Q2 enabled more precise and personalized ad candidate selection, extending coverage to Facebook Reels and mobile feeds. This resulted in a 4% increase in ad conversions on Facebook [2][4].
GEM handles the ranking stage, deciding which ads to show from the candidates Andromeda provides. By leveraging organic engagement data and doubling the length of user event sequences analyzed, GEM led to a 5% increase in ad conversions on Instagram and a 3% increase on Facebook across feed and reels [2][4].
The Lattice architecture replaced many specialized models with a unified system learning across all objectives, enabling continuous self-improvement. Early implementation in Q2 contributed to nearly a 4% uplift in conversions and demonstrated Meta’s growing AI capabilities [2][4].
These AI-driven improvements boosted user engagement, with increased time spent on Instagram and Facebook. The efficiency and effectiveness of advertising allowed marketers to achieve better returns on investment even amid economic uncertainties. Meta reported a 22% year-over-year revenue growth to $47.5 billion in Q2 2025, with $46.6 billion from advertising [1][2][3][5].
Europe led with a 24% growth in ad revenue, despite regulatory headwinds. Instagram video time surged over 20% year-over-year, and Facebook video engagement in the US jumped 20%. Two million advertisers now use Meta's AI video generation tools [1][2][3][5].
However, Meta's financial success was tempered by a $4.5 billion operating loss incurred by its Reality Labs division. Susan Li, Meta's CFO, anticipates a "sharp acceleration in depreciation expense growth" in 2026, with annual depreciation exceeding $20 billion due to Meta's large capital expenditures. Meta plans to spend $66-72 billion on capital expenditures in 2025, with estimates of $100 billion in 2026 [1][2][3][5].
Meta expects minimal revenue from generative AI through 2026, but its emphasis on building "elite, talent-dense teams" for AI research suggests a winner-take-all dynamic in AI researcher recruitment. Meta's first gigawatt-scale cluster, Prometheus, comes online in 2026 [1][2][3][5].
In conclusion, Meta's AI architectures transformed its ad business by improving ad relevance, increasing conversions by around 3-5% depending on the platform, boosting user engagement, and driving robust financial gains with higher advertising revenue and margins [1][2][4][5].
- The scale of Meta's business expanded in Q2 2025, as demonstrated by a 22% year-over-year revenue growth, with $46.6 billion from advertising.
- This growth was driven by innovations in artificial intelligence, such as Andromeda, GEM, and Lattice, which enhanced ad targeting and performance.
- Investment in technology has been instrumental in driving the business growth, with plans to spend $66-72 billion on capital expenditures in 2025 and estimates of $100 billion in 2026.
- Marketers have benefitted from these advancements, achieving better returns on investment due to the efficiency and effectiveness of advertising.
- Meta's competitors in the advertising business are under pressure to respond to these innovations, as the company leads the way in AI research and development.
- The leadership at Meta, including its CFO, have acknowledged challenges in other areas of the business, such as the $4.5 billion operating loss incurred by the Reality Labs division.
- Despite regulatory headwinds, regions like Europe have shown significant growth in ad revenue, with Instagram video time and Facebook video engagement showing strong increases.
- Future revenue streams, such as those from generative AI, are expected to be minimal until 2026, but Meta is focusing on building "elite, talent-dense teams" to remain competitive in the race for AI dominance.
