The caseins, main milk proteins, interact with each other to form spherical aggregates called micelles. The water mobility in casein and caseinate systems has been previously described in suspension and gel. The sensitivity of water relaxation to casein structure is known in milk and for concentrated dairy matrix at least up to 15% in dry matter. Nevertheless the water mobility and the casein-casein interaction for highly concentrated casein system above the close packing limit, is unknown. All information on such system will be helpful to improve diffusion driven processes as for example filtration process especially when cake filters are formed and reduced the flow. These study report for the first time results of the change in casein structure and water diffusion in casein and caseinate system for very high concentration up to 450 g/l.. Samples were prepared from native casein powder NCP and sodium caseinate powder rehydrated in UF permeates from milk. For 2.5 g/l up to 480 g/ l the casein suspensions were obtained after a rehydration during 12 hours at 35 °C for NCP and at 50°C for caseinate. For higher concentration, the casein suspensions were obtained by the osmotic stress technique. The NMR measurements were performed with a Bruker Minispec operating at 20 Mhz and equipped with a pulsed field gradient probe which delivered 5 T/m.T2 we measured with a (CPMG) sequence and a FID. The water diffusion coefficient was measured with a Stejskal-Tanner spin echo sequence. The T2 relaxation presented a multi-exponential behaviour. The major component due to water protons and exchangeable protons was already identified. Two other components were observed; these minor components presented a relaxation time of the order of the millisecond. In studying casein dispersions in D2O these two minor components were assigned to non-exchangeable protein protons. The variation of the water relaxation 1/T2 as a function of the casein micellar and sodium caseinate concentration in g/g of water was correctly described by a linear regression for all the range of concentration. Theses results demonstrated that no structural variation of the casein was observed during dehydration despite the liquid-solid transition observed at 200 g/l. However, a large difference between caseinate and micellar casein were reported. This difference was explained by the casein structure between native phosphocasein (NPC) which are spherical aggregates (~ 150 nm of main diameter) with calcium phosphate nanocluster and Na-caseinate which are smaller structure (~10 nm of diameter) without the nanoclusters. The Na-caseinate relaxation rate was lower than the NPC relaxation rate at the same concentration. In the opposite, we showed that the water diffusion coefficient was insensitive to the casein structures. No difference in water diffusion was observed in NPC and caseinate systems. The variation of the water diffusion according to the casein concentration was discussed and compared to published model. These results demonstrated that at high concentration in casein a stronger reduction of the water diffusion than expected was observed. Theses results highlighted the specificity of such high concentrated matrix.