Meta Spends $72 Billion on AI as Stock Rises 9% on Earnings

Meta Platforms, Inc. is accelerating its artificial intelligence (AI) ambitions with a capital expenditure budget of $72 billion for 2025—surpassing previous forecasts of $60–$65 billion [1]. This aggressive funding is being directed toward AI infrastructure, including the development of massive computing clusters such as Prometheus and Hyperion, named after figures from Greek mythology. Prometheus is expected to become the world’s first 1GW+ cluster by 2026, while Hyperion is projected to scale up to 5GW over several years [1]. The investment is also driving talent recruitment, with Meta acquiring AI companies and poaching top-tier researchers and executives, including Alexandr Wang from Scale AI and Nat Friedman from GitHub [1].

The new Meta Superintelligence Labs (MSL), launched in June, is operating independently with a startup-like structure and has become a hub for AI innovation. MSL is working on successors to the Llama series, beginning with Llama 4.1 and 4.2, and is shifting toward a mixture-of-experts (MoE) architecture despite some developer concerns [1]. CEO Mark Zuckerberg emphasized AI as a core pillar during the Q2 2025 earnings call, dedicating the majority of the discussion to the company’s vision for “personal superintelligence.” This concept envisions AI that deeply understands individual users and helps them achieve their goals [1].

Meta is integrating AI across its platforms, with features already embedded into Instagram, WhatsApp, and its core app, reaching 1.5 billion active users. Its hardware division is also incorporating AI, with Ray-Ban Meta glasses featuring an embedded AI assistant [1]. While the company highlights the societal benefits of AI, Zuckerberg has acknowledged the risks and emphasized the importance of rigorous risk mitigation and careful open-sourcing decisions [1].

Financially, Meta’s Q2 2025 earnings showed a 22% revenue increase, and its stock price rose 9% following the earnings announcement, reflecting investor confidence in its AI strategy [1]. However, the company faces significant cost pressures, with year-over-year expenses rising by 9% to $24.76 billion in Q1 2025 [1]. The Reality Labs division, which previously focused on the metaverse, reported a $5.35 billion loss in Q2 2025, raising questions about the financial viability of its broader AI investments [1].

Public and industry reactions to Meta’s AI push have been mixed. Some express optimism about the team’s capabilities and the potential for rapid progress, while others remain skeptical given the company’s past high-profile ventures, such as the metaverse [1]. Meta’s approach to AI is now being benchmarked against rivals like OpenAI and Google, the latter of which is also committing $85 billion to AI-related investments for 2025 [1]. Unlike these competitors, Meta is pursuing a fully integrated AI ecosystem with end-to-end control over models, infrastructure, and applications [1].

The success of Meta’s AI gamble will depend on its ability to convert these investments into scalable, user-adopted products and measurable financial returns. While the company’s vision for personal superintelligence remains aspirational, the coming months will be critical in determining whether these efforts translate into tangible innovation or another high-risk, high-cost venture [1].

Source: [1] Meta's AI-Driven Transformation: Fueling Earnings and Reshaping the AI Arms Race (https://www.ainvest.com/news/meta-ai-driven-transformation-fueling-earnings-reshaping-ai-arms-race-2507/)

[2] Earnings call transcript: Meta's Q2 2025 results exceed expectations (https://www.investing.com/news/transcripts/earnings-call-transcript-metas-q2-2025-results-exceed-expectations-93CH-4161533)

[3] Meta Platforms, Inc. (META) Q2 2025 Earnings Call (https://seekingalpha.com/article/4806454-meta-platforms-inc-meta-q2-2025-earnings-call-transcript)