Microtubules are self-assembling cytoskeletal protein structures composed of tubulin dimer subunits and are ubiquitous in mammalian cells, where they serve a multitude of functions such as maintaining cellular shape, providing pathways for intracellular transport, and driving neuronal outgrowth in early development. Their ability to elongate with the addition of tubulin, shorten by peeling away individual or oligomeric subunits, and switch between these two states is called dynamic instability and provides a focal point for cellular regulation of microtubule-driven processes. Indeed, many microtubule-associated proteins (MAPs) are known to bind microtubules and alter dynamic instability by promoting or suppressing either the polymerization or depolymerization of individual microtubules. One heavily studied MAP is tau, an intrinsically disordered protein primarily localized to the axonal compartment of mature neurons. Scientific focus on the properties and function of tau stems from the discovery many decades ago of neurofibrillary tangles as a hallmark of Alzheimer’s disease and, later of tau’s involvement in those tangles and in several other neurodegenerative diseases. Pathological and physiological roles for tau have been studied extensively. For example, we know that different tau isoforms are differentially expressed in development and maturity, that tau binds to and modifies microtubule dynamic instability, and that chemical alterations to tau leading to dysfunction are sufficient to induce disease-like pathology in cell cultures, animal studies, and humans. One poorly understood tau function is its ability to mediate the phase separation and regular spacing of microtubules in string-like arrays or bundles. Such microtubule structures are found at the axon initial segment of mature neurons, where they are referred to as “fascicles,” and have been re-produced with the addition of tau in non-neuronal cell cultures and in cell-free protein experiments involving only tubulin and recombinant human tau. Interestingly, the sorting of tau to the axon has been shown to rely on the structural integrity of the AIS, including microtubules, and is dependent on binding of tau to microtubules, implying a possible link between tau-mediated microtubule bundling and the subcellular localization of tau itself. Thus, imperative to understanding the physiological relevance of tau-mediated microtubule bundling is determining the mechanism by which tau mediates inter-microtubule interactions within bundles. Several models have been proposed, creating a dogma of tau-tau interactions driving bundling, but are incompatible with polyelectrolyte theory and the structural features of microtubule bundles, namely wall-to-wall distances between neighboring microtubules. To better understand the underlying mechanism of tau-mediated bundling, we sought to monitor the structures of microtubule bundles as a function of time in the presence of several factors or conditions we thought would alter the interactions underlying bundle formation. First, experiments were designed to probe the electrostatic-component of tau-mediated bundling. Several tau-tau models for microtubule bundling rely on the dipole-like distribution of charge along tau’s N-terminal tail, where a relative abundance of acidic residues are followed by an abundance of basic residues, providing a possible mechanism for two opposing tau molecules to overlap and favorably interact to hold adjacent microtubules together. By increasing the ionic strength of the buffers, we sought to weaken such charge-charge interactions, which would predict a decrease in bundling strength or an outright inability of tau to mediate bundle formation in high-salt buffers. Instead, our data show that wall-to-wall distances between microtubules is largely unaffected by increases in monovalent cation species. Interestingly, with the addition of excess divalent cations, Mg2+ or Ca2+, an unexpected transformation of the hexagonal bundle lattice was observed. Specifically, we found that above threshold concentrations of either divalent cation, a phase transition is induced as a function of time or increased cation content. The phase transition is marked by the sudden and simultaneous drop in average wall-to-wall spacing, increase in lattice parameter (bundle size, or number of microtubules per bundle), and proliferation of inverted tubulin rings. Based on these results, we propose and test a model where free tubulin oligomers participate in and are necessary for the bundling of microtubules by tau protein. In our model, tau binds to multiple species of tubulin to join together free (non-lattice-bound) tubulin oligomers and lattice-bound tubulin within microtubules. The multivalent nature of tau’s interactions with various tubulin oligomers and microtubules creates a network of tubulin and tau that cross-links microtubules within bundles. Within this model, an absence or depletion of free tubulin should prevent tau-mediated bundling, while increases to the free tubulin content should enhance bundle properties. Consistent with this model, experiments designed to induce rapid microtubule depolymerization by dropping the sample temperature reproduced the phase transitions observed in the depolymerization events induced by divalent cations. Similarly, samples prepared with increasing GTP content showed resistance to microtubule depolymerization over time, corresponding to the delay and in some cases elimination of the phase transition. Taken together, our results and the model we propose for tau-mediated microtubule bundling represent a novel role for tubulin in the bundling process. We are excited by the prospect of these new findings and their physiological relevance both to the understanding of developmental and mature tau function in human neurons but also to tau dysfunction in neurodegenerative diseases.