News

Meta CFO Susan Li says the company is still catching up on AI infrastructure capacity due to underestimated demand

Friday, March 6, 2026 at 02:52 PM

Meta is actively working to bridge the gap between its AI infrastructure capacity and actual demand, admitting that previous capacity forecasts from two to three years ago underestimated current requirements. The company confirms it is still in a phase of 'catching up' regarding its data center and hardware footprint through the current year.

Context

During Meta's fourth-quarter 2025 results follow-up call on January 28, 2026, CFO Susan Li addressed the company's persistent infrastructure constraints, stating that the firm is still playing catch-up because they underestimated demand for AI capacity 24 to 36 months ago. This supply-demand imbalance comes as Meta pivots aggressively toward personal superintelligence, a shift that has seen capital expenditures nearly double from $72 billion in 2025 to a projected $115–$135 billion for 2026. This massive infrastructure buildout is designed to support the next generation of LLMs and recommendation systems. To bridge the capacity gap, Meta recently announced a multiyear strategic partnership with NVIDIA for large-scale GPU deployment. While the company's core advertising revenue reached a record $201 billion in 2025, investors are closely monitoring whether this historic level of spending will translate into improved ad targeting and user engagement as the firm navigates this high-stakes AI arms race.

Related Companies

Meta
Meta
META
US