Meta Platforms Inc. stunned Wall Street on Wednesday by raising its 2026 capital expenditure forecast to as much as $145 billion, a dramatic increase from the $72 billion spent last year. The news sent shares tumbling more than 7% in after-hours trading, even as the company reported its fastest revenue growth since 2021 at 33% for the latest quarter. The massive spending is driven almost entirely by the artificial intelligence boom, which has created unprecedented demand for high-bandwidth memory chips and data center infrastructure.
During the earnings call, CEO Mark Zuckerberg acknowledged that the increased spending is largely due to higher component costs, particularly memory pricing. The global memory chip supply has been constrained by the rapid buildout of AI data centers by tech giants like Google, Amazon, and Microsoft, as well as by Meta itself. This shortage has driven up prices for the valuable HBM (high-bandwidth memory) chips used in AI accelerators, and has also spilled over into consumer electronics, making laptops and smartphones more expensive.
Meta's $145 billion figure is more than double its capex from just two years ago and represents a bet-the-company move on AI. Zuckerberg is under immense pressure to deliver results after the company's previous big bet on the metaverse, through the Reality Labs division, has so far failed to generate meaningful revenue. In the same earnings report, Reality Labs posted an operating loss of over $4 billion on just $402 million in sales, adding to cumulative losses exceeding $80 billion since the division's inception.
The AI Catch-Up Plan
Meta has been perceived as lagging behind in the AI race, especially compared to Google, which has released a series of powerful generative AI models. Roughly ten months ago, Zuckerberg admitted the company was behind and announced a major catch-up effort. This included pouring billions into research and development, poaching top talent from rivals, and establishing a new unit called Meta Superintelligence Labs, led by Scale AI founder Alexandr Wang.
Earlier this month, Meta released its first major AI model from this lab: Muse Spark. The proprietary model is designed for multimodal tasks and is expected to be open-sourced in the future. Zuckerberg described it as a critical first step. "Now that we have a strong model, we can develop more novel products as well," he told investors on the call.
Two AI Agents for Personal and Business Use
Zuckerberg outlined plans to launch two AI agents: one for personal use and another for business customers. The business AI agent is already being tested, and Zuckerberg claimed that weekly conversations with it have grown tenfold since the start of 2026. These agents are part of Meta's strategy to integrate AI deeply into user experiences and enterprise tools, potentially creating new revenue streams beyond advertising.
The company is also using AI internally to improve its core products. Meta CFO Susan Li revealed that over half a billion users on Facebook and Instagram each week now watch videos translated and dubbed by AI. This feature has increased engagement and time spent on the platforms. Additionally, Meta is incorporating the new AI model into its recommendation systems to hyper-personalize feeds for users. Zuckerberg said, "Since our recommendation systems are operating at such large scale, we’ll phase in this new research and technology over time," but noted that the trend shows increasing returns from AI-enhanced personalization.
To help offset the massive investments, Meta is cutting costs through layoffs. The company is laying off 10% of its workforce and reportedly offering voluntary buyouts to 7% of its U.S. staff. While executives declined to directly attribute the job cuts to AI automation, Li mentioned that a "leaner operating model" would help offset the substantial investments. This follows a broader Silicon Valley trend of companies using AI to automate tasks and reduce headcount.
Broader Context: The Global Memory Crisis
The AI-driven data center buildout has caused a global shortage of memory chips, particularly HBM3 and HBM3e, which are essential for training large language models. Samsung, SK Hynix, and Micron have ramped up production, but demand continues to outstrip supply. This has led to price increases of up to 30% for some memory modules, impacting not only tech giants but also consumers who face higher prices for electronics. Meta's $145 billion figure includes a significant portion allocated to securing memory supplies for its data centers.
Analysts are divided on whether Meta's massive spending will pay off. Some argue that the company is making a necessary bet to catch up in AI, while others worry about repeating the mistakes of the metaverse investment. Zuckerberg, however, expressed confidence, stating, "This is the first release from Meta Superintelligence Labs, and it shows that our work is on track to build a leading lab." The coming months will be critical as Meta prepares to launch its AI agents and further integrate Muse Spark into its ecosystem.
In summary, Meta's earnings reveal a company in transition, willing to spend aggressively on AI infrastructure despite short-term market skepticism. The success of this strategy hinges on the practical deployment of AI agents and the ability to monetize new AI features, while managing the memory chip supply chain and the financial burden of Reality Labs' ongoing losses.
Source: Gizmodo News