META · Q4 2025 Earnings

Narrative arc: Q2 "similarly significant 2026 capex" → Q3 "significantly faster 2026 expense growth" → Q4 "personal superintelligence as the operating model."

The 2026 framing has hardened from financial commitment to product manifesto in one quarter. In Q3, the multi-year capex super-cycle was justified by ROI evidence — value-weighted conversion rates, compute-constrained ad systems with positive-ROI use cases going unfunded. In Q4, Zuckerberg reframed the entire spend as architectural: "I expect 2026 to be a year where this wave accelerates even further on several fronts. We're starting to see agents really work. This will unlock the ability to build completely new products and transform how we work." The shift from ROI-defense to vision-statement is significant — when the spend ramp accelerates and the justification simultaneously moves up the abstraction ladder, the implicit signal is that the near-term ROI evidence isn't keeping pace with the spend curve.

Reality Labs framing pivoted from "uncertain trajectory" to "peak losses now." Across Q2 and Q3, RL was framed as a long-duration investment with no committed inflection point. This quarter Zuckerberg stated directly: "I expect reality labs losses this year to be similar to last year, and this will likely be the peak." The qualifier "likely" is doing meaningful work — and the guidance language ("remaining similar to 2025 levels") confirms losses are not narrowing in 2026. Combined with the AI glasses framing ("hard to imagine a world in several years where most glasses that people wear aren't AI glasses"), Zuckerberg is asking investors to extend the RL patience window by another year while pre-committing to a future peak that hasn't yet been demonstrated.

Ad monetization narrative shifted from "load + price" to "redistribution and timing." Susan Li disclosed: "In the second half of 2025, our initiatives on Facebook to redistribute ads across users and sessions delivered a nearly four times larger revenue impact than Facebook ad load increases." This is a material change from the Q2/Q3 framing of impressions and price as twin engines. The system has matured to where targeting precision outweighs load — which is healthier for engagement but caps the volume-only growth lever that drove Q4's +18% impressions print.

Capacity-constraint language remained, but now extends through "much of 2026." Eric Sheridan asked directly about constraints; Zuckerberg said Meta will "likely remain capacity constrained through much of 2026" despite the $115–135B capex commitment. In Q3, constraint language was framed as a 2025 issue being resolved. The forward extension of that constraint suggests demand for compute internally is growing faster than the buildout — which justifies the capex but also implies the spending ceiling hasn't been found.

The MSL frontier model deliverables remain deliberately vague. Doug Anmuth pushed for milestones; Zuckerberg deflected with "6 months into rebuild" and "positive early indicators" but explicitly positioned first models as showing "trajectory rather than peak capability." This is the same evasiveness flagged in Q3 — three quarters in, no concrete capability claim, no committed launch.