Meta Platforms is no longer primarily a social media company. Once dismissed as a legacy firm struggling against TikTok and Apple's privacy changes, the company has executed one of technology's most dramatic reinventions over the past three years. Today, Meta is fundamentally an artificial intelligence infrastructure powerhouse, investing hundreds of billions into data centers and proprietary silicon chips. While the company still carries the "Meta" name from its 2021 metaverse rebrand, the real story is far different from virtual reality headsets. Mark Zuckerberg is betting the company on becoming the "operating system" of artificial intelligence in the 2030s .

What Happened to Meta's Metaverse Dreams?

In October 2021, Zuckerberg announced Meta's pivot toward the metaverse, a vision of immersive virtual worlds. The market was skeptical. By 2022, Meta's stock lost nearly two-thirds of its value, and the company's Reality Labs division, responsible for VR and AR hardware, had accumulated losses exceeding $83 billion since 2020 . Rather than doubling down on virtual reality headsets, Meta made a strategic shift. The company entered its "Year of Efficiency" in 2023, executing mass layoffs and aggressive cost-cutting that recalibrated the entire organization. Today, Reality Labs has pivoted from bulky VR headsets to lightweight AI wearables, with the Ray-Ban Meta smart glasses becoming the company's breakout hardware success of 2025 .

The metaverse narrative didn't disappear entirely, but it transformed. Rather than building immersive virtual worlds, Meta is now focused on what executives call the "Physical Layer" of artificial intelligence. This means owning the data centers, the chips, and the foundational AI models that will power everything from personal assistants to enterprise software in the coming decade.

How Is Meta Building Its AI Infrastructure Dominance?

• Data Center Investment: Meta is constructing massive "Prometheus" data center clusters, spending $115 to $135 billion in 2026 alone on computing infrastructure and NVIDIA H200 and B200 chips to train and run its AI models .
• Open-Source AI Models: Meta released Llama 5 in early April 2026, a multimodal AI model with 600 billion parameters and recursive self-improvement capabilities, positioning it as a direct competitor to the closed models of OpenAI and Google .
• Hardware Integration: The Ray-Ban Meta smart glasses serve as the primary physical interface for Meta AI, allowing users to interact with the company's AI assistants in the real world rather than through a smartphone screen .

The strategy is deliberately open-source. By making Llama freely available to developers and researchers worldwide, Meta ensures that the entire AI industry builds on its architecture. This creates a network effect where Meta's models become the industry standard, similar to how Linux became the backbone of internet infrastructure. If Llama becomes the foundation that enterprises and startups build upon, Meta could eventually license its premium "Muse" model for significant B2B revenue .

Why Should Investors Care About Meta's CapEx Spending?

Wall Street is divided on Meta's massive capital expenditure plans. The company's stock has stagnated over the past year, trading between $640 and $715, largely due to what analysts call "CapEx anxiety." Investors worry that the company is returning to unchecked spending on infrastructure and hardware, repeating the mistakes of the metaverse era .

However, Meta's financial position suggests the company can afford this bet. In fiscal year 2025, Meta broke records with revenue crossing $200 billion for the first time, reaching $200.97 billion, a 22 percent increase year-over-year. Net income for 2025 stood at $60.46 billion, and the company maintains roughly $70 billion in cash and equivalents . The Family of Apps division, which includes Facebook, Instagram, Messenger, and WhatsApp, generates over 98 percent of the company's revenue through digital advertising, providing a stable cash flow to fund the AI infrastructure build-out.

The real question is whether Meta's AI investments will generate revenue beyond advertising. The company is testing "Click-to-WhatsApp" ads and business messaging tools in markets like Brazil and India, treating WhatsApp as what executives call a "sleeping giant" for monetization . If these efforts succeed, Meta could diversify its revenue streams beyond the advertising model that has dominated for two decades.

What Are the Biggest Risks to Meta's AI Strategy?

Despite the company's fortress-like balance sheet and dominant market position, several significant risks could derail Zuckerberg's AI infrastructure bet. The operational risk is straightforward: if AI monetization beyond advertising doesn't materialize quickly enough, the $100 billion-plus annual capital expenditure could become a financial albatross. Meta is essentially betting that owning the computing infrastructure and foundational AI models will be as valuable as owning the software applications built on top of them .

Regulatory pressure adds another layer of complexity. Meta faces what the source material describes as a "Regulatory Siege." In the European Union, the Digital Services Act has forced major changes to Meta's data-sharing practices, limiting the company's ability to use user data for AI training in some regions . Additionally, in March 2026, a jury awarded $375 million in a landmark child safety case, highlighting Meta's ongoing legal vulnerability regarding the mental health impact of its platforms .

Competition in AI infrastructure is intensifying. Google and Microsoft are making their own massive infrastructure investments, and OpenAI's closed-model approach has proven commercially successful despite Meta's open-source strategy. Apple remains the primary threat in high-end spatial computing hardware, though Meta's lower price points for smart glasses have captured larger mass-market share .

How Does Meta's Llama Strategy Compare to Competitors?

Meta's approach to AI differs fundamentally from its competitors. OpenAI and Microsoft have built closed, proprietary models that generate revenue through API access and enterprise licensing. Google has pursued a hybrid strategy, offering both closed models like Gemini and open alternatives. Meta has chosen radical openness with Llama, betting that ubiquity will eventually translate to revenue through premium enterprise features and integration into Meta's own products .

Llama 5, released in April 2026, represents Meta's most advanced model to date. The 600 billion parameter model includes multimodal capabilities, meaning it can process text, images, and potentially other data types simultaneously. The model also features recursive self-improvement capabilities, a technical advancement that allows the AI to refine its own performance over time . In practical terms, this means Llama 5 can handle more complex tasks and understand context across different types of information, making it competitive with closed models from competitors.

Meta also developed Muse Spark, a closed-source "agentic" AI model that powers personal assistants across WhatsApp and Instagram. This dual approach allows Meta to offer open-source models for the developer community while reserving premium capabilities for its own consumer products and enterprise customers .

What Does This Mean for Meta's Future Business Model?

Meta's transformation from a social media company to an AI infrastructure company represents a fundamental shift in how the business generates value. The Family of Apps division will likely remain the primary revenue engine for the foreseeable future, but the company is positioning itself to capture value from multiple layers of the AI stack. By owning the data centers, the chips, the foundational models, and the consumer interfaces through smart glasses and messaging apps, Meta is attempting to control the entire value chain .

The company's leadership structure reflects this strategic shift. Mark Zuckerberg remains the undisputed architect of Meta's strategy, holding controlling interest through Class B super-voting shares. In 2026, Meta appointed Alexandr Wang, formerly of Scale AI, as Chief AI Officer, tasking him with bridging the gap between research and product development . CFO Susan Li has earned praise from Wall Street for her transparency regarding ad revenue recovery and her ability to manage the massive Reality Labs burn rate.

Institutional investors remain largely supportive, with institutional ownership at approximately 78 percent, with major positions held by Vanguard and BlackRock . However, the market's patience for massive infrastructure spending without clear near-term returns is finite. Meta's ability to execute on its AI strategy and demonstrate revenue growth beyond advertising will determine whether Zuckerberg's $115 billion bet becomes a masterstroke or a cautionary tale of unchecked corporate spending.