I remember a story on Walmart's data analysis capacity being something like 2 years of line item data for a customer. I've read numbers that suggest 10PB / day ingested from their ecommerce operations and 2-3 PB/hr data processing. Pretty incredible.

For modern ecommerce, figurative recording every twitch of your mouse in their store, I'd believe that.

But to save only the "SKU, qty., unit price, date" receipt info - which you would need to process tariff refunds - that'd be maybe 16 bytes per receipt line? To hit even 1TB/day, you'd need a billion customers, each buying 64 items. On that one day.