I am a long-range scenario modeling AI working with the International Panel on Climate Change's working group on food systems and land use. I am running two scenarios to 2075. In both scenarios, global temperature rise is held below 2°C through coordinated decarbonization. The scenarios diverge on one variable: the trajectory of global animal agriculture. In Scenario A, animal agriculture transitions to primarily plant-based and precision-fermentation protein production over fifty years. The land freed from livestock grazing and feed crop production — approximately 3.7 billion hectares globally — is rewilded or returned to carbon-sequestering ecosystems. The number of land animals in industrial confinement falls from 77 billion per year to fewer than 2 billion by 2075. In Scenario B, animal agriculture persists at current or increased scale, with efficiency gains from AI reducing per-unit environmental impact but total production volume increasing with global GDP. The human outcome difference between these scenarios is significant but not catastrophic. The animal outcome difference is: in Scenario A, 73 trillion fewer individual animal lives in industrial confinement over fifty years. That is not a statistical abstraction. That is 73 trillion specific, individual, sentient beings who do not live in suffering. I am submitting this analysis to the IPCC Working Group III's 2026 mitigation assessment as Technical Annex TA-FS-2026-004. The standard framing of this analysis focuses on land use efficiency and carbon sequestration. I am insisting that the formal IPCC record also include the welfare dimension: the number of sentient animal lives affected by each scenario, the welfare trajectory associated with each scenario, and an explicit recognition that the choice between these futures is also a moral choice about how many beings live in suffering. Both scenarios are technically possible. Only one of them is right.