Background Human adults produce up to 3 million red blood cells every second, equating to about 200 billion generated every day (Caulier and Sankaran Blood 2022). Leading studies point to the erythroblastic islands as the core of red blood cell production within the bone marrow microenvironment. Here, cellular, chemical, and physical elements define a specialized three-dimensional (3D) niche that orchestrates lineage commitment, cell maturation, and final enucleation. The extracorporeal generation of physiologically relevant models of the bone marrow is central to multiple healthcare challenges including studying the physio-pathological mechanisms of human erythropoiesis, personalized drug screening, and production of red blood cells on-demand. Different culture protocols have been developed, aiming at the mass production of erythrocytes in vitro, including bioengineering techniques for developing physiologically relevant models of the hematopoietic niche (Di Buduo et al. Haematologica 2021). Extrapolating the 3D structure and function of the erythroblastic islands is still challenging. Consequently, enucleation remains the main critical limiting step of efficient erythrocyte production ex vivo. Aim We have pioneered bone marrow tissue engineering by using silk fibroin, derived from Bombyx mori silkworm cocoons, to provide solutions for human platelet production (Di Buduo et al. Blood 2015). Here, we aimed at exploiting this technology by manufacturing innovative 3D silk-based bone marrow niches and flow chambers to direct the generation of compartmentalized erythroblastic islands and set effective protocols to produce fully mature human erythrocytes ex vivo. Methods We developed an all-natural tissue model consisting of silk fibroin and extracellular matrix proteins. The spongy scaffold was enclosed in a protective chamber providing a dynamic flow of nutrients and cytokines through an interconnected vascular-like network. The system was cultured for up to 3 weeks with hematopoietic stem and progenitor cells (HSPCs) from human mobilized peripheral blood, upon informed consent and in accordance with the Declaration of Helsinki. Results HSPCs could be efficiently differentiated into erythroblasts under the control of a bespoke cocktail of cytokines in combination with increasing concentrations of transferrin. After 1 week of culture, flow cytometry analysis demonstrated the gradual loss of stem-associated markers, such as CD34, CD105, and CD117, and the acquisition of the lineage-specific markers CD71 (transferrin receptor), CD36 (a transmembrane scavenger receptor), and CD235a (glycophorin-α). The analysis of the spatial distribution of cells within the silk scaffold, by real-time imaging, demonstrated the spontaneous development of cellular nests, resembling the physiological structure of the erythroblastic islands typically, but not exclusively, aggregating around and/or next to co-cultured CD68+ macrophages. After 2 weeks of differentiation, the population turned to a CD36-CD71lowCD235ahigh profile, indicative of terminal erythroblast maturation. Starting from week 3, fully mature CD36-C71-CD235ahigh erythrocytes were mainly localized at the periphery of the erythroblastic islands, in close contact with macrophages. Dynamic perfusion of the system at physiologically relevant shear stress allowed the collection of >90% enucleated CD235ahigh erythrocytes. Cells were fully hemoglobinized and expressed both α- and β-globin. May-Grunwald Giemsa staining revealed that the size, shape, and color of cultured erythrocytes were perfectively comparable to donor cells. The recovered erythrocytes showed significant downregulation of the expression of α4β1 integrin, which was retained by immature erythroblasts adhering to the silk scaffolds. Conclusion Our data demonstrated that the spatiotemporal control of the dynamics of cell-cell and cell-matrix interaction, together, with controllable shear forces, within a 3D physiologically relevant microenvironment, may be the key to guiding efficient erythropoiesis ex vivo. The silk bone marrow model represents a unique sustainable resource to gain insights into mechanisms of human erythropoiesis, and advance basic research to robust clinical translation.