Amazon's Kiva-Era Fulfillment Legacy
Cytronic
The key implication is that Amazon still optimizes fulfillment by moving shelves to people, not by rebuilding the workflow around item level automation. Kiva gave Amazon a huge speed advantage, but it was designed for broad catalog storage inside giant buildings. That makes it powerful for mixed SKU volume, yet harder to match a newer system built only for small parcel DTC picking, packing, and carrier sortation.
-
Amazon bought Kiva in 2012 and kept scaling that architecture across its network. Even in 2025, Amazon described robots moving tall stacks of inventory to associates, which shows how deeply the original goods to person model still sits inside the fulfillment system.
-
That legacy does not mean Amazon is unautomated. Amazon says it has more than 1 million robots and more than 200 US fulfillment centers. The point is that much of this footprint was built for maximum SKU breadth and marketplace scale, not for a narrow low cost DTC service with fewer product edge cases.
-
Locus shows the alternative path. Instead of rebuilding a warehouse around fixed Kiva style infrastructure, Locus drops robots into existing sites in 4 to 6 weeks and charges about $2,000 per robot per month. That improves labor productivity inside incumbent 3PLs, but it still layers onto the old operator model rather than replacing it.
The next phase of fulfillment competition is a split between incumbents modernizing huge legacy networks and new operators designing warehouses around one narrow job from day one. Amazon will keep adding robots, but the strongest challengers will win by choosing simpler SKU sets, denser automation, and a service model built around lower per order cost rather than maximum assortment breadth.