In this paper, we present a new asynchronous seizure detector that is part of an implantable integrated device intended to identify electrographic seizure onset and trigger a focal treatment to block the seizure progression. The proposed system has a low-power front-end bioamplifier and a seizure detector with intelligent mechanism to reduce power dissipation. This system eliminates the unnecessary clock gating during normal neural activity monitoring mode and reduces power dissipation in the seizure detector; as a result, this device is suitable for long-term implantable applications. The proposed system includes analog and digital building blocks with programmable parameters for extracting electrographic seizure onset information from real-time EEG recordings. Sensitivity of the detector is enhanced by optimizing the variable parameters based on specific electrographic seizure onset activities of each patient. The detection algorithm was validated using Matlab tools and implemented in standard 0.13 μm CMOS process with total die area of 1.5 × 1.5 mm². The fabricated chip is validated offline using intracranial EEG recordings from two patients with refractory epilepsy. Total power consumption of the chip is 9 μW and average detection delay is 13.7 s after seizure onset, well before the onset of clinical manifestation. The proposed system achieves an accurate detection performance with 100% sensitivity and no false alarms during the analyses of 15 seizures and 19 non-seizure datasets.