--- title: "The Ambitions and Hidden Worries Behind Meta's Soaring Performance and Stock Price" type: "Topics" locale: "zh-HK" url: "https://longbridge.com/zh-HK/topics/32492981.md" description: "Facebook's parent company $Meta Platforms(META.US) recently sparked a talent war in Silicon Valley. Since the beginning of this year, Meta has been poaching top AI talent from companies like $Apple(AAPL.US) and OpenAI, primarily through high salaries, stock options, and opportunities to participate in key projects. In May, Meta acquired a 49% stake in the startup Scale AI for $14.3 billion..." datetime: "2025-08-01T02:50:15.000Z" locales: - [en](https://longbridge.com/en/topics/32492981.md) - [zh-CN](https://longbridge.com/zh-CN/topics/32492981.md) - [zh-HK](https://longbridge.com/zh-HK/topics/32492981.md) author: "[财华社](https://longbridge.com/zh-HK/profiles/11651030.md)" --- > 支持的語言: [English](https://longbridge.com/en/topics/32492981.md) | [简体中文](https://longbridge.com/zh-CN/topics/32492981.md) # The Ambitions and Hidden Worries Behind Meta's Soaring Performance and Stock Price Facebook's parent company$Meta Platforms(META.US) has recently sparked a talent war in Silicon Valley. Since the beginning of this year, Meta has been assembling a super-intelligent team, primarily by luring top AI talent from companies like$Apple(AAPL.US) and OpenAI with high salaries, stock options, and opportunities to participate in key projects. In May, Meta acquired a 49% stake in the startup Scale AI for $14.3 billion, bringing its CEO Alexandr Wang on board. A week later, it poached Daniel Gross, co-founder and CEO of Safe Superintelligence (SSI). It then spent $1 billion to partially acquire NFDG Venture Capital, securing the services of former GitHub CEO Nat Friedman. But this was just the beginning. In June, Meta poached several core researchers from OpenAI, including Liu Lu, the lead developer of GPT-4o's "Ghibli-style" image generation feature, renowned multimodal AI architecture expert Allan Jabri, and three top visual AI experts from OpenAI's Zurich office: Lucas Beyer, Alexander Kolesnikov, and Xiaohua Zhai. Additionally, at least eight other senior researchers joined, including Bi Shuchao, head of deep generative models; Zhao Shengjia, head of lightweight models; Yu Jiahui, leader of the perception team; and Ren Hongyu, head of post-training. In July, Meta recruited four key researchers from Apple's Fundamental Models Team (AFM), including multimodal AI expert Zhang Bowen, former team leader Pang Ruoming, and senior members Tom Gunter and Mark Lee. Notably, Meta offered Pang Ruoming a compensation package exceeding $200 million over five years—far surpassing Apple's top executives and nearly triple the annual salary of Apple CEO Tim Cook. Whether these hires will deliver ROI beyond their hefty paychecks remains to be seen, but Meta's latest quarterly results show that AI is indeed boosting profitability. During the earnings call, Meta emphasized that its talent acquisition spree will continue, with capital expenditures set to rise further. This has left the market both excited and wary, though optimism seems to prevail for now: Meta's pre-market stock surged 11.53% post-earnings, rivaling$Microsoft(MSFT.US) , which is nearing a $4 trillion market cap. **Meta's Q2 Margins Improve, but Capex Rises Again** In Q2 2025, Meta's revenue grew 21.61% YoY to $47.516 billion. Operating costs rose 16.19% to $8.491 billion, driven by infrastructure expenses and partner payments. R&D spending increased 22.82%, primarily due to higher employee compensation and infrastructure costs. Operating profit reached $20.441 billion, up 37.68% YoY, with operating margins expanding 5.02 percentage points to 43.02%. Net income grew 36.18% to $18.337 billion, with diluted EPS at $7.14. The "Family of Apps" segment continued to grow, with management reporting over 3.4 billion daily active users in June. Segment revenue rose 21.77% to $47.146 billion, led by a 21.48% increase in ad revenue ($46.563 billion). Europe and Rest of World saw the strongest growth at 24% (23% FX-neutral), followed by North America (21%) and Asia-Pacific (18%). Segment operating profit jumped 29.15% to $24.971 billion, with margins up 3.03 points to 52.97%. Reality Labs revenue grew 4.82% to $370 million, buoyed by AI glasses sales offsetting Quest declines. The segment's operating loss widened 0.94% to $4.53 billion due to higher non-employee tech development costs. Meta's Q2 capex hit $17 billion, driven by server, data center, and network investments. Free cash flow was $8.55 billion. Meta repurchased $9.8 billion of Class A shares and paid $1.3 billion in dividends. **Why the Hiring Spree?** Mark Zuckerberg stated during the earnings call that Meta formed the Superintelligence Labs to realize its vision of "everyone having a personal superintelligence." The initiative combines foundational research, product development, and FAIR (Facebook AI Research), with a new lab dedicated to next-gen models. To staff these efforts, Meta has aggressively recruited senior researchers from OpenAI, Apple, and others at significant cost. He noted progress on Llama 4.1 and 4.2, with parallel development of next-gen models expected to push boundaries in 1–2 years. Meta is building an elite team: Alexandr Wang oversees overall management, Nat Friedman leads AI product/applied research, and Zhao Shengjia serves as chief scientist. Meta is constructing multiple multi-gigawatt clusters. The Prometheus cluster, touted as the world's first 1+ GW cluster, will launch next year. The Hyperion cluster will scale to 5 GW in coming years, with other massive clusters in development. Zuckerberg highlighted five priorities last quarter: ad optimization, user experience, commerce messaging, Meta AI, and AI hardware. For ads, Meta expanded its AI-driven recommendation model, boosting Instagram ad conversion by ~5% and Facebook's by 3%. For user experience, AI enhances content recommendations and quality to increase engagement. Commerce messaging integrates AI into Facebook, Instagram, and e-commerce sites. Meta AI now has over 1 billion MAUs, with ongoing model refinements to deepen engagement. In hardware, Ray-Ban Meta smart glasses sales are strong. A new Oakley Meta HSTN model offers longer battery life and higher-resolution cameras for sports. The Meta Quest 3S Xbox Edition launched last month, with cloud gaming interest at record highs. Zuckerberg sees broader applications beyond gaming. **Capex Surge Continues** CFO Susan Li cited two revenue drivers: community engagement and monetization efficiency. Meta is boosting engagement via AI-optimized experiences. Reality Labs' Ray-Ban Meta glasses face supply shortages despite ramped production. For monetization, Meta is refining ad loads, improving targeting, and upgrading its ad platform. **In capex allocation, management prioritizes infrastructure and talent. Hiring will focus on top priorities, with headcount growth expected in 2025–2026. Meta Superintelligence Labs will aggressively recruit top AI talent.** **Infrastructure investments will scale significantly in 2026, particularly in AI compute.** Li projects Q3 revenue of $47.5–$50.5 billion, with Q4 growth slowing due to tough comps. 2025 operating expenses are now guided at $114.8–$118 billion (up from $113–$118B), a 20–24% YoY increase. She expects 2026 opex growth to accelerate, led by infrastructure costs (from AI investments) and higher compensation (from continued hiring). 2025 capex is projected at $66–$72 billion (up from prior $64–$72B), marking a 68%+ YoY jump from 2024's $39.225 billion. 2026 capex growth will mirror 2025's, driven by generative AI scale-up. Meta will self-fund much of this but is exploring partnerships for data centers (no deals finalized). Li stated that compute resources will first meet internal needs (e.g., content recommendations, ad ranking, and next-gen model training), with no near-term plans for external use. On ROI, core AI investments show strong returns, but generative AI remains early-stage, unlikely to materially impact revenue until beyond 2026. **Conclusion** Meta's aggressive AI bets—from talent wars to massive capex—underscore its superintelligence ambitions, earning market approval for now. However, soaring capex (up 68%+ in 2025, staying high in 2026) could become a volatility trigger: if AI pays off, growth may absorb costs; if not, efficiency concerns may weigh on shares. Author: Mao Ting ### 相關股票 - [Apple (AAPL.US)](https://longbridge.com/zh-HK/quote/AAPL.US.md) - [Meta Platforms (META.US)](https://longbridge.com/zh-HK/quote/META.US.md) - [OpenAI (OpenAI.NA)](https://longbridge.com/zh-HK/quote/OpenAI.NA.md) - [Microsoft (MSFT.US)](https://longbridge.com/zh-HK/quote/MSFT.US.md)