Capital Expenditures: The Big-Ticket Item
The most direct signal of Meta's AI ambitions is found in one line item: Capital Expenditures, or CapEx. This figure represents spending on long-term physical assets, and for Meta, it overwhelmingly means servers, data centers, and network infrastructure.
In its Q3 2026 earnings, Meta reported a staggering $34.91 billion in quarterly CapEx, a number that outstripped its cash from operations. This spending caused its free cash flow to turn negative, to the tune of minus $2.12 billion, for the first time since 2012. When this number surges, as it has, it’s a clear indication that the company is aggressively building the physical backbone required for developing and deploying advanced AI models. Analysts watch Meta's full-year CapEx guidance—now projected at $135 to $145 billion for 2026—as the primary barometer of its AI investment intensity.
The Earnings Call: Where Numbers Get Their Story
Numbers alone don't tell the full story. The real color comes from the commentary provided by CEO Mark Zuckerberg and other executives during the quarterly earnings call. This is where the 'why' behind the 'what' is explained. For instance, after reporting a negative free cash flow, Zuckerberg framed it as a necessary cost of competition, pointing to the launch of the company's first frontier models, Muse and Muse Wonder, as a key milestone enabled by these investments. Analysts and investors scrutinize these calls for mentions of GPU (Graphics Processing Unit) procurement, data center timelines, and the expected returns on AI investments. Management's tone and specific justifications for the astronomical spending help investors determine whether the outlays are a sign of strategic strength or a potential cash drain.
Beyond CapEx: Finding Other AI Costs
While CapEx is the main event, other parts of the financial statements reveal more of the AI picture. The 'Cost of Revenue' includes massive, ongoing operational expenses like the electricity needed to run the power-hungry data centers. The 'Research and Development' (R&D) budget reflects the human side of the equation: the salaries of the thousands of highly paid engineers and data scientists building the AI models. Furthermore, accounting practices themselves can offer clues. For example, in early 2026, Meta extended the depreciable useful life of its servers, a move that reduces annual depreciation expenses and can make profitability appear stronger in the short term, even as spending soars. Understanding these secondary costs gives a more holistic view of the total investment.
Reality Labs: Untangling AI From the Metaverse
One of the key challenges for investors is separating AI spending from the company's other major bet: the metaverse, housed within the Reality Labs division. Both initiatives require massive computational power. However, Meta has started providing clearer signals. For 2027, the company stated it expects losses in Reality Labs to decrease as it reallocates investment toward 'other priority areas, primarily AI'. The Reality Labs segment reported a revenue of $443 million against ongoing losses, which are projected to be around $15 billion in 2027. By tracking the declared shifts in investment priorities between these two divisions, investors can better estimate how much capital is being funneled specifically into the AI arms race versus the development of virtual worlds and hardware like the Quest headsets.
