Glucose-6-phosphate isomerase (GPI) has an essential function in both catabolic glycolysis and anabolic gluconeogenesis and is universally distributed among Eukaryotes, Bacteria, and some Archaea. In addition to the cytosolic GPI, land plant chloroplasts harbor a nuclear encoded isoenzyme of cyanobacterial origin that is indispensable for the oxidative pentose phosphate pathway (OPPP) and plastid starch accumulation. We established 12 new GPI sequences from rhodophytes, the glaucophyte Cyanophora paradoxa, a ciliate, and all orders of complex algae with red plastids (haptophytes, diatoms, cryptophytes, and dinoflagellates). Our comprehensive phylogenies do not support previous GPI-based speculations about a eukaryote-to-prokaryote horizontal gene transfer from metazoa to gamma-proteobacteria. The evolution of cytosolic GPI is largely in agreement with small subunit analyses, which indicates that it is a specific marker of the host cell. A distinct subtree comprising alveolates (ciliates, apicomplexa, Perkinsus, and dinoflagellates), stramenopiles (diatoms and Phytophthora [oomycete]), and Plantae (green plants, rhodophytes, and Cyanophora) might suggest a common origin of these superensembles. Finally, in contrast to land plants where the plastid GPI is of cyanobacterial origin, chlorophytes and rhodophytes independently recruited a duplicate of the cytosolic GPI that subsequently acquired a transit peptide for plastid import. A secondary loss of the cytosolic isoenzyme and the plastid localization of the single GPI in chlorophycean green algae is compatible with physiological studies. Our findings reveal the fundamental importance of the plastid OPPP for Plantae and document the plasticity of primary metabolism.