Meta could soon lease computing power to Anthropic

This is best viewed as a capital-efficiency story, not a new-core-business story. If Meta can externalize excess AI infrastructure, the market may start discounting a portion of its AI spend as revenue-generating rather than purely strategic capex, which supports multiple expansion at the margin. The immediate winners are META holders and, secondarily, GPU/server suppliers if the arrangement signals that frontier model demand is still outrunning internal buildouts.

The more interesting second-order effect is on competitor economics. A material lease deal would imply that a non-cloud platform can arbitrage its balance sheet into AI rent, which could pressure AWS and Azure on model-lab hosting if Anthropic diversifies away from incumbent clouds. It also raises the bar for peers like GOOGL and AMZN to show that their own AI capex is translating into monetizable utilization rather than stranded depreciation.

The risk is that the market overstates the earnings impact before the terms are known. A lease can be high headline value but low near-term P&L contribution if it is phased over years, heavily customized, or simply offsets internal underutilization; in that case the rally fades once investors realize this is mostly optics. Time horizon: momentum can persist for days, but the real catalyst is 1-3 months of disclosure around capex, utilization, and whether this becomes a repeatable channel. Falsifier: a capex guide increase with no offsetting revenue commentary, or a clarification that the arrangement is non-binding / immaterial to 2025 results.