The raw data (all Neuropixels channels sampled at 30 kHz) is available as part of an AWS Public Dataset. The easiest way to interact with this data is by mounting the S3 bucket in an AWS SageMaker instance. It’s also possible to download these files via the AWS Command-Line Interface, but this will be quite slow, as there is ~1.2 TB of data for each experiment.
See this thread for information on how to align the raw data with the spike times in the NWB files.