At the 2025 second quarter financial report meeting of Meta, the CEO and CFO introduced the company's current business performance, emphasizing the importance of over 3.4 billion daily active users and its AI investments. They discussed the company's progress in the field of AI, including self-improvement of AI systems, pursuit of superintelligence, and the newly established Meta Superintelligence Lab. In addition, they mentioned the company's development of multiple large-scale computing clusters to support AI research, and detailed the prospects for the application of AI in advertising, user experience, business information, AI devices, and other areas.

The strong performance of this quarter's results is mainly attributed to artificial intelligence (AI), which has unlocked higher efficiency and revenue in the advertising system. By expanding AI-driven advertising recommendation models, improving their performance, and introducing more signals and longer contexts, the conversion rates of ads on Instagram and Facebook have increased by approximately 5% and 3% respectively. Additionally, AI has made significant progress in improving user experience, with improvements in the recommendation system leading to a 5% and 6% increase in user time on Facebook and Instagram respectively. AI has also been used in video editing tools and business information exchange, helping businesses interact better with consumers. At the same time, the sales of AI devices such as Ray-Ban and Oakley Meta glasses have accelerated, showcasing the potential of AI applications in everyday life. In the field of virtual reality, the Meta Quest ecosystem continues to attract users, expanding applications such as cloud gaming and media browsing. Overall, AI has played a key role in driving business growth and innovation.

Meta Platforms Inc. had a total revenue of $47.5 billion in the second quarter of 2023, a 22% increase year-on-year, mainly driven by increased infrastructure costs and rising partner payment expenses. Total expenses for the quarter were $27.1 billion, a 12% increase, with research and development expenses increasing by 23% and marketing and sales expenses increasing by 9%. The total number of employees in the company was 75,900, a decrease of 1% in the quarter. Operating income was $20.4 billion, with an operating profit margin of 43%. Net profit was $18.3 billion, with earnings per share of $7.14. Capital expenditures were $17 billion, mainly used for investments in servers, data centers, and network infrastructure. Advertising revenue for the family of apps business was $46.6 billion, a 21% increase, mainly driven by strong growth in Europe and other regions worldwide.

In the second quarter (Q2), Meta company allocated the majority of its investments towards developing and operating its family applications, with expenses on family applications reaching $22.2 billion, accounting for 82% of total expenses, a 14% year-on-year increase. The operating income for its family applications and Reality Labs department were $25 billion and $37 million respectively, with the latter seeing a 5% year-on-year growth due to increased sales of AI glasses. Additionally, Meta continues to optimize its recommendation systems and product experiences to increase user engagement, especially in video content on Facebook and Instagram. The company is also exploring the use of AI technology to improve recommendation systems and content discovery, including providing services across multiple platforms with Meta AI. In terms of Reality Labs, demand for Ray-Ban Meta products continues to exceed supply, and the company is working to increase production to meet consumer demand. Furthermore, Meta has begun to introduce advertising in the updates labels for Threads and WhatsApp, even though the immediate contribution of these platforms to advertising revenue is minimal, the company remains optimistic about their long-term potential.

In the second quarter, Facebook significantly increased the relevance and personalization of its advertisements by innovating its ad system, including improving the Andromeda model architecture, enhancing the performance of the Generative Ads (Gem) system, and expanding the coverage of the Lattice model architecture. These improvements drove nearly 4% higher conversion rates on Facebook mobile Feed and Reels, increased ad conversion rates by about 5% on Instagram, and improved conversion rates by 3% on Facebook Feed and Reels. Additionally, Facebook's Advantage Plus AI-driven solution gained strong momentum, especially in the use of video generation, image animation, and text generation tools, attracting nearly 2 million advertisers. Furthermore, the company completed the deployment of global incrementality attribution capabilities and launched global omni-channel advertising to help advertisers optimize online and offline sales and reduce total purchasing costs.

The focus of this discussion is on the company's capital allocation strategy, mainly targeting investments in infrastructure and recruitment of talent, specifically in recruiting top talent in the field of AI and accelerating AI model and product development. It is expected that investments in infrastructure and talent will continue to grow in 2026, while emphasizing the significant returns of AI capabilities in advertising and organic engagement projects. In addition, revenue expectations for the third and fourth quarters of 2025 are provided, as well as expected ranges for total expenses and capital expenditures for the year, while also mentioning the potential impact of changes in tax laws and the regulatory environment in Europe.

In recent discussions, the company's senior management delved into the evolution of the artificial intelligence business in the past three to six months, sharing key learnings and how these learnings have influenced talent acquisition and the strategy adjustment of computing resources. They emphasized the initial effectiveness of AI in improving the quality of Facebook algorithms and user engagement, although the scale is currently small, showing an optimistic development trajectory. Additionally, the company insists on considering superintelligence as an important principle, believing that AI will profoundly change the company's systems and business models, thus driving the demand for top talent and leading computing capabilities to accelerate the widespread application and product integration of AI technology.

The company has not yet officially started the budget process for 2026, but has some preliminary expectations for key financial indicators next year. The main driver of expected total expenditure growth in 2026 will be infrastructure construction, particularly a significant increase in depreciation expenses due to the addition of new assets and services, as well as more capital expenditure on shorter-term assets. Additionally, with the recruitment of AI talent, employee salaries will also be another major factor contributing to expenditure growth. In terms of capital expenditure, the company plans to significantly increase AI capacity, especially in training capabilities, servers, networks, and data centers, while continuing to invest in core AI areas.

The dialogue focused on the challenges and goals in the development of super intelligent technology, particularly emphasizing the importance of self-improvement research, and optimizing the configuration of cutting-edge research in small and highly specialized teams. In addition, strategies and factors to improve user engagement through enhancing the core platform in the next 18 months were also discussed.

Meta company is focusing on multiple short-term goals in the future planning of its core recommendation engine, including enhancing the real-time adaptability of recommended content, optimizing the matching efficiency of small creators' content, deepening user interest exploration, and improving the advancedness and efficiency of the models. Long-term goals involve developing cross-service recommendation models, integrating LLDs (Low-Level Descriptors) more deeply, and optimizing system efficiency to ensure that expanding the recommendation system will not affect investment returns. At the same time, Meta continues to support open-source artificial intelligence, although it remains reserved about sharing very large-scale models mainly due to practicality and security issues. In terms of infrastructure investment, the company plans to continue bearing most of the capital expenditure on its own, while also considering the possibility of collaborating with financial partners to invest in data centers, in order to address the growing capital needs and maintain flexibility.

Meta company is currently focused on building infrastructure for internal use cases, including AI work, recommendation engines, organic content, and advertising ranking, among others. External applications of the infrastructure have not been considered yet. In terms of return on capital expenditure, the company sees strong returns in the core AI field, while in generative AI, it is still in the early stages and not expected to be a significant revenue driver in the short term. However, the company remains optimistic about long-term commercial opportunities, especially in five key areas. Additionally, Meta is building infrastructure with flexibility in mind to adapt to future changes in demand.

In the discussion, Meta emphasized key performance indicators (KPIs) to focus on in the pursuit of the vision for artificial intelligence (AI) superintelligence, including the quality of teams and models, the speed of improvements to AI systems, and the contributions of foundational models to the overall AI capabilities of the company. The company believes that advancements in AI technology will first be translated into new products and services for billions of users, followed by gradual commercialization. Despite the lag between investment and returns, Meta firmly believes that directly utilizing AI to provide high-quality services to customers will lead to higher returns, rather than just renting out hardware facilities. Additionally, the company stated that while large-scale investments at the current stage may impact short-term profit growth, in the long run, these investments will open up new opportunities for the company and strengthen core business, ultimately leading to sustained profit growth.

The dialogue discussed Meta AI's continuous improvement and upgrading of models (such as from Lambda 4 to Lambda 4.1) to enhance the personalized user experience and overall engagement of the WhatsApp platform. With Behemoth's launch, future plans will focus on improving AI performance in functions such as search and queries, to promote wider user adoption and potentially explore commercialization paths.

Meta discussed the progress of its AI glasses project, especially the Ray-Ban and Oakley Meta glasses, highlighting the increasing application of AI technology in glasses and the importance of glasses as an ideal form of AI interaction. Additionally, the company discussed how to minimize the issue of shareholder dilution caused by stock-based compensation (SBC) resulting from increasing employees through stock buybacks and quarterly cash dividends.