This paper proposes a novel channel estimator for low-complexity MIMO channel estimation using diffusion models (DMs), a top-rated generative model. Unlike previous works that utilize generative priors, the proposed estimator employs a lightweight convolutional neural network (CNN) with positional embedding of the signal-to-noise ratio (SNR) information to learn the channel distribution in the sparse angular domain. The estimator avoids stochastic resampling and truncates reverse diffusion steps for lower SNR, resulting in low complexity and memory overhead. Numerical results show that the proposed DM-based estimator outperforms state-of-the-art channel estimators using generative priors, achieving better performance in both 3GPP and QuaDRiGa channel models. The estimator's performance is close to the optimal conditional mean estimator (CME) and demonstrates asymptotic optimality with a moderate number of diffusion timesteps. The lightweight CNN architecture and efficient estimation strategy make the proposed method practical for real-time applications in massive MIMO systems.This paper proposes a novel channel estimator for low-complexity MIMO channel estimation using diffusion models (DMs), a top-rated generative model. Unlike previous works that utilize generative priors, the proposed estimator employs a lightweight convolutional neural network (CNN) with positional embedding of the signal-to-noise ratio (SNR) information to learn the channel distribution in the sparse angular domain. The estimator avoids stochastic resampling and truncates reverse diffusion steps for lower SNR, resulting in low complexity and memory overhead. Numerical results show that the proposed DM-based estimator outperforms state-of-the-art channel estimators using generative priors, achieving better performance in both 3GPP and QuaDRiGa channel models. The estimator's performance is close to the optimal conditional mean estimator (CME) and demonstrates asymptotic optimality with a moderate number of diffusion timesteps. The lightweight CNN architecture and efficient estimation strategy make the proposed method practical for real-time applications in massive MIMO systems.