Meta launches compute rental service to monetize AI infrastructure

Meta's announcement of Meta Compute reflects a deliberate strategy to offset the mounting cost of its AI ambitions by turning idle computing resources into a revenue-generating business. The company has invested heavily in building proprietary AI chips (Iris) and advanced models (Muse Spark 1.1, Muse Image), moves that have drawn scrutiny over rising capital expenditures. By renting out unused capacity, Meta aims to demonstrate that these investments are not pure R&D drag but can fuel growth in an enterprise segment that complements its advertising core.

However, the stakes are substantial. Meta's forward guidance projects $369.0 billion（約59兆円） revenue and $111.2 billion（約18兆円） earnings by 2029, requiring 19.7% yearly revenue growth and a $40.6 billion（約6.5兆円） earnings increase from current levels of $70.6 billion（約11兆円）. The calculus assumes that aggressive AI spending will not crush profit margins, which some analysts already expect to compress toward 29 percent even as absolute earnings climb. Meta Compute and the Iris chip ramp could either vindicate this optimism or force a fundamental rethink if capital intensity outpaces revenue generation.

A parallel concern shadows these announcements: EU regulatory actions and multibillion-dollar fine exposure could constrain Meta's ability to leverage user data and advertising insights for AI training and deployment. The success of Meta Compute depends not only on technical execution but also on regulatory permission to operate the data-driven business model that funds the AI buildout in the first place.