The 2015 Paris declaration obligated international development organizations to assess the climate compatibility of their projects. For irrigation projects, like those negotiated between the Agence Française de Développement, and the Cambodian government in the early 2020s, calculations of estimated greenhouse gas emissions have become important requirements. But how to estimate emissions from future rice fields and the effects of irrigation infrastructures that do not exist? To address this issue, emissions calculators have been developed as a means to bridge climate science and development knowledge infrastructures, so that data and forms of calculation from climate science can easily enter the world of development. However, by engaging in an infrastructural inversion, we argue that this understanding is flawed. Drawing on a case study of an irrigation project in Cambodia, we show that heterogeneous data concatenations are continuously transformed in the movement across infrastructures until referentiality breaks down. Emission calculators operate as a data wormhole, emitting extremely uncertain numbers that contribute to a problematic and speculative politics of anticipation. In contrast with the dominant politics of anticipation, which depends on futile efforts to neutralize uncertainty, infrastructural inversion makes it possible to envision a decentered politics attentive to distributed agency.