Evaporation of water ice above 100 K in the inner few 100 AU of low-mass embedded protostars (the so-called hot core) should produce quiescent water vapor abundances of ~10^-4 relative to H₂. Observational evidence so far points at abundances of only a few 10^-6. However, these values are based on spherical models, which are known from interferometric studies to be inaccurate on the relevant spatial scales. Are hot cores really that much drier than expected, or are the low abundances an artifact of the inaccurate physical models? We present deep velocity-resolved Herschel-HIFI spectra of the 3₁₂-3₀₃ lines of H_2^{16}O and H_2^{18}O (1097 GHz, E _u/k = 249 K) in the low-mass Class 0 protostar NGC 1333 IRAS2A. A spherical radiative transfer model with a power-law density profile is unable to reproduce both the HIFI data and existing interferometric data on the H_2^{18}O 3₁₃-2₂₀ line (203 GHz, E _u/k = 204 K). Instead, the HIFI spectra likely show optically thick emission from a hot core with a radius of about 100 AU. The mass of the hot core is estimated from the C¹⁸O J = 9-8 and 10-9 lines. We derive a lower limit to the hot water abundance of 2 × 10^-5, consistent with the theoretical predictions of ~10^-4. The revised HDO/H₂O abundance ratio is 1 × 10^-3, an order of magnitude lower than previously estimated. Herschel is an ESA space observatory with science instruments provided by European-led Principal Investigator consortia and with important participation from NASA.