Meta’s $135 Billion AI Spending Plan Raises the Stakes for Big Tech—and the Power Grid

Meta Platforms Inc. posted another quarter of strong growth—but it was the company’s spending plans, not its earnings, that reset expectations for Big Tech’s artificial-intelligence race.

A capital budget that looks like a utility’s

Meta said it expects 2026 capital expenditures of $115 billion to $135 billion, alongside total expenses of $162 billion to $169 billion, largely to fund data centers, servers and hiring for AI. The scale of the forecast—more typical of an oil major or power utility than an advertising-driven social-media company—helped push the stock higher in trading following the announcement.

Meta reported fourth-quarter 2025 revenue of $59.9 billion, up 24% year over year, and net income of $22.8 billion, up 9%. Diluted earnings per share rose to $8.88, topping analyst estimates. For the full year, Meta reported $201 billion in revenue, up 22% from 2024.

Chief Financial Officer Susan Li said the majority of expense growth would come from infrastructure, including “third-party cloud spend, higher depreciation and higher infrastructure operating expenses.” She said employee compensation—“particularly AI” technical talent—would be the second-largest driver. Despite higher costs, Li told analysts that Meta expects to generate more operating income in 2026 than in 2025, in absolute dollars.

Why Meta says it has to spend so much

Chief Executive Mark Zuckerberg framed the surge as necessary to keep pace in what he called a “major AI acceleration.”

“In 2025, we rebuilt the foundations of our AI program,” Zuckerberg said. “Our vision is building personal superintelligence. To do that, we need to significantly expand our data center and compute capacity.”

Meta said the investment will support new and expanded data centers, custom servers optimized for AI workloads, networking equipment and power infrastructure. The company also expects to keep paying for capacity from outside cloud providers while waiting for its own facilities to come online.

Li said Meta remains “capacity constrained,” with demand for compute growing faster than availability. She expects constraints to persist “through much of 2026 until additional capacity from our own facilities comes online.”

Much of the work is centered in Meta Superintelligence Labs, formed in 2025 to consolidate frontier-model research and development. The group oversees the Llama foundation models, the Meta AI assistant integrated across the company’s apps, and large-scale recommendation and ad-ranking systems used in Facebook and Instagram.

Li said Meta doubled the GPU cluster used to train its GEM recommendation model in 2025 and plans to “meaningfully scale up” training again in 2026.

How Meta’s spending compares with other hyperscalers

Meta’s plan lands it in a small group of companies spending at unprecedented levels to build AI infrastructure:

• Microsoft Corp. has reported tens of billions of dollars in recent quarterly data-center and AI-related investment and signaled elevated spending will continue.
• Alphabet Inc. has told investors it expects about $75 billion in 2025 capital expenditures, largely for AI and data centers; some analysts project 2026 capex could approach $140 billion.
• Amazon.com Inc. has outlined plans to invest about $118 billion in AI-related initiatives in 2025, much of it through Amazon Web Services.

What stands out about Meta is the speed of its trajectory and its business mix: Meta is not primarily a cloud provider, and the vast majority of its revenue still comes from advertising across Facebook, Instagram and WhatsApp. If the company hits its 2026 guidance, its capital budget would rival—or exceed—companies whose core business is renting out compute power.

Why Wall Street seems more comfortable with Meta’s AI bet

Meta’s shares jumped roughly 9% to 11% in after-hours and early trading following the results and guidance. The response contrasted with the market’s reaction to Microsoft’s recent results, after which the software maker’s stock fell about 10%, wiping out more than $400 billion in market value.

Analysts and investors have pointed to a key difference: Meta is emphasizing more direct links between AI infrastructure and near-term revenue drivers, particularly improvements in ad targeting and engagement.

“We continue to see strong returns from our AI investments in our ads business,” Li said, citing higher click-through rates and better performance for advertisers. Meta is also expanding AI assistants in messaging and building tools that let businesses use automated agents to handle customer queries inside WhatsApp and Messenger.

Ripple effects: chips, data centers—and electricity

Meta’s guidance has implications well beyond social media.

Chipmakers and the silicon mix

AI-accelerator makers and high-end GPU suppliers are direct beneficiaries of hyperscale spending. Nvidia Corp. shares have previously moved on reports of higher Meta AI capex, reflecting expectations of sustained demand for data-center chips.

But Meta’s scale is also raising questions about vendor concentration. People familiar with the company’s plans say Meta has explored using Alphabet’s Tensor Processing Units (TPUs) at greater scale from 2026 onward. Analysts say shifting even 10% to 15% of Meta’s annual AI hardware budget toward alternative silicon could reduce Nvidia’s future data-center revenue potential by several billion dollars a year.

Data-center operators and equipment suppliers

Meta has indicated it will continue using third-party cloud capacity and may pursue joint ventures or co-developed facilities to supplement its buildout, which could benefit colocation firms and infrastructure owners.

Large AI campus announcements have also driven investor interest in suppliers tied to backup power, cooling and electrical equipment, as hyperscalers expand orders for power management and thermal systems.

Pressure on the power grid

Meta has discussed expansive new sites in several U.S. states, including a proposed multibillion-dollar AI facility in Louisiana, as it seeks long-term energy supplies and land. Such projects add to strain on regional grids facing rising industrial demand.

Utilities and independent power producers have cited AI data centers as a major expected source of load growth. Developers of low-carbon baseload options—such as nuclear operators and backers of small modular reactors—have highlighted AI and cloud computing as potential customers.

The scale of demand is also drawing political attention. Lawmakers and local officials in some states have raised concerns about water and electricity use by large data centers, and some jurisdictions have moved to slow approvals while studying impacts on grids and communities.

Legal and regulatory headwinds remain

Meta’s AI pivot is unfolding alongside ongoing legal and regulatory challenges to its core social-media business. The company faces lawsuits in the U.S. alleging its platforms harm young users and foster addiction—claims Meta has denied.

Legal and regulatory costs contributed to a 24% increase in total expenses in 2025, to $117.7 billion. Li said Meta expects “trials scheduled for this year in the U.S. … which may ultimately result in a material loss.”

In Europe, Meta is adjusting advertising systems in response to privacy rulings and new digital regulations. The company has begun offering “less personalized” ad options in the European Union and said the changes will be a headwind to revenue growth in 2026.

From metaverse to AI infrastructure

Meta’s surge in AI spending coincides with a step back from the all-encompassing “metaverse” push that followed the company’s 2021 name change. Reality Labs generated $955 million in fourth-quarter revenue but posted an operating loss of about $6 billion. Meta has cut staff in the division and is emphasizing products that blend AI with hardware, including Ray-Ban-branded smart glasses.

Zuckerberg now describes Meta as a “deep technology company” whose long-term advantage comes from controlling both the models and the infrastructure that run them. He has argued that as AI systems become more capable, leading models may be less likely to be widely available through low-cost APIs, increasing the strategic value of building in-house compute.

What Meta’s bet could mean for the next decade

If Meta executes its 2026 plan, Meta, Microsoft, Alphabet and Amazon could collectively spend more than half a trillion dollars on capital projects in a single year, much of it tied to AI. That concentration is consolidating control over advanced computing infrastructure among a small set of firms—even as regulators debate how to oversee the technology.

Whether Meta’s $135 billion wager pays off will depend on more than ad performance. The company must secure chips, land and energy, manage a tougher regulatory climate, and convince users and policymakers that increasingly powerful AI embedded in social platforms will do more good than harm. For now, investors appear willing to fund the experiment—and suppliers across the economy are positioning for a world in which AI infrastructure, not social networking alone, defines Meta’s next era.