Recent LHCb data shows that the direct CP asymmetries of the decay modes D 0 → π + π − and D 0 → K + K − have the same sign, violating an improved U -spin limit sum rule in an unexpected way at 2.1σ. From the new data, we determine for the first time the imaginary part of the CKMsubleading, U -spin breaking ∆U = 1 correction to the U -spin limit ∆U = 0 amplitude. The imaginary part of the ∆U = 0 amplitude is determined by ∆a dir CP . The corresponding strong phases are yet unknown and could be extracted in the future from time-dependent measurements.Assuming O(1) strong phases due to non-perturbative rescattering, we find the ratio of U -spin breaking to U -spin limit contributions to the CKM-subleading amplitudes to be ( 173+85 −74 )%. This highly exceeds the Standard Model (SM) expectation of ∼ 30% U -spin breaking, with a significance of 1.95σ. If this puzzle is confirmed with more data in the future, in the SM it would imply the breakdown of the U -spin expansion in CKM-subleading amplitudes of charm decays. The other solution are new physics models that generate an additional ∆U = 1 operator, leaving the U -spin power expansion intact. Examples for the latter option are an extended scalar sector or flavorful Z models.