AppLovin 2026 Conference Call Summary Company Overview - **Company**: AppLovin (NasdaqGS: APP) - **Date**: March 04, 2026 - **Speakers**: Adam Foroughi (CEO), Matt Stumpf (CFO), Matt Kost (Morgan Stanley U.S. Internet team) Key Points Industry and Market Dynamics - AppLovin operates in the mobile gaming and advertising industry, focusing on gaming ads and expanding into e-commerce and web advertising [6][8] - The company has a target growth rate of 20%-30% for gaming ads, which has been exceeded in recent years, indicating strong market potential [5][6] Growth and Technology - AppLovin's technology is described as nascent but rapidly improving due to ongoing enhancements and data accumulation [6][7] - The company is leveraging its existing user base of over 1 billion daily active users to expand into e-commerce, which is seen as a more complex but lucrative market [12][15] - The introduction of new advertising products, such as universal campaigns and new customer campaigns, aims to drive customer acquisition and engagement [13][14] Advertising Performance - AppLovin currently has a 1.3% conversion rate on ads served, with potential to increase to over 5% as the technology improves and more advertisers join the platform [20][23] - The company emphasizes the importance of a powerful recommendation system to personalize ad content and improve conversion rates [22][42] Financial Management and Investment - AppLovin maintains a lean operational structure with around 900 employees, focusing on efficiency and disciplined investment in technology and headcount [30][31] - The company plans to add a small number of employees to support growth in e-commerce and web advertising without significantly impacting the overall cost structure [31][32] Competitive Landscape - AppLovin views competition as an opportunity to expand the advertising market rather than a threat, leveraging its unique position and technology to differentiate from larger platforms like Meta and Google [37][38] - The company has successfully captured a significant market share in mediation with its MAX product, which is designed to provide unbiased and transparent auction processes for publishers [52][53] Future Outlook - AppLovin is focused on long-term growth, aiming to build a robust platform that can scale significantly over the next 5 to 10 years [15][16] - The company recognizes the need to improve its marketing efforts to raise awareness of its capabilities and attract more customers [65] Challenges - Acknowledges the challenge of effectively marketing its business and technology to potential customers, which is crucial for future growth [65] Additional Insights - The company is exploring the use of generative AI for ad creative, which could enhance the efficiency and effectiveness of advertising campaigns [47][50] - AppLovin's recommendation system is expected to evolve alongside advancements in AI, potentially doubling its predictive capabilities in the coming years [63][64] This summary encapsulates the key insights and strategic directions discussed during the AppLovin conference call, highlighting the company's growth potential, technological advancements, and market positioning.

Model Building and Training - LinkedIn leverages large language models (LLMs) for personalization and ranking tasks, aiming to use one model for all tasks [2][3] - The process involves converting user information into prompts, a method called "promptification" [8] - LinkedIn builds a large foundation model, Blue XL, with 150 billion parameters, then distills it to smaller, more efficient models like a 3B model for production [12] - Distillation from a large model is more effective than training a small model from scratch [14] - Increasing data, model size (up to 8x22B), and context length can improve model performance, but longer contexts may require model adjustments [17][18][19] Model Performance and Generalization - The model improves performance for cold start users, showing a growing gap compared to production models as interactions decrease [21] - The model demonstrates generalization to new domains, performing on par with or better than task-specific production models in out-of-domain tasks [23] Model Serving and Optimization - LinkedIn focuses on model specification, pruning, and quantization to improve throughput and reduce latency for production [26] - Gradual pruning and distillation are more effective than aggressive pruning, minimizing information loss [29][30] - Mixed precision, including FP8 for activations and model parameters but FP32 for the LM head, is crucial for maintaining prediction precision [31][32] - Sparsifying attention scores can reduce latency by allowing multiple item recommendations without each item attending to each other [34][35] - LinkedIn achieved a 7x reduction in latency and a 30x increase in throughput per GPU through these optimization techniques [36]