The first wave of warehouse automation was about replacing conveyor belts with more sophisticated conveyor belts. Systems from companies like Dematic, Swisslog, and Honeywell built highly efficient fixed paths for moving goods through distribution centers — impressive at scale, but expensive to install and difficult to reconfigure when product mixes changed.
The second wave, which is accelerating through 2026, is fundamentally different in architecture: instead of fixed infrastructure, it deploys fleets of autonomous mobile robots (AMRs) that navigate dynamically, pick flexibly, and can be redeployed as demand patterns shift. The economics are also different: AMR systems typically cost 40-60% less to install than equivalent fixed-automation systems and can be operational within weeks rather than the 12-18 months required for large fixed automation projects.
Amazon Robotics — the division that emerged from the 2012 acquisition of Kiva Systems — operates the world's largest AMR fleet, with over 750,000 robots deployed across its network. The latest generation, the Sequoia system, combines high-density storage pods with robotic arms capable of picking individual items into orders, a task that earlier systems could not handle. Amazon reports that Sequoia sites complete orders 25% faster than facilities using the previous generation of systems.
Ocado, the British online grocery operator turned technology company, represents a different model: the customer fulfillment center (CFC) as a product. Ocado's grid-based robotic system — where hundreds of bots crawl over a three-dimensional grid of storage, retrieving totes on demand — has been licensed to grocery retailers including Kroger (US), Sobeys (Canada), and Lotte (South Korea). The Ocado system excels at the specific demands of grocery: high SKU counts, mix of ambient, chilled, and frozen products, and the need to assemble individual customer orders precisely.
AutoStore, a Norwegian company, has taken a similar grid concept to a broader range of retail and pharmaceutical fulfillment. Its cubic storage system can achieve storage densities four times higher than traditional shelving, making it valuable in urban distribution centers where square footage is expensive. Customers include Swisscom, Lufthansa, and specialty pharmaceutical distributors where pick accuracy is regulated.
The remaining frontier for warehouse automation is piece picking — the ability for a robotic arm to reliably grasp arbitrary product shapes and sizes at commercial throughput speeds. Companies like Mujin, Nimble Robotics, and Covariant are attacking this problem with robot arms trained on large datasets of real warehouse images. Accuracy rates for piece picking have improved dramatically, from roughly 85% in 2022 to over 98% in controlled conditions by 2025, but the gap between controlled conditions and the actual variety of a real distribution center remains a challenge.
What is clear is that the total number of AMRs deployed globally is growing at roughly 25-30% per year, and the technology is moving from early-adopter logistics companies to mainstream retail and manufacturing. By 2028, analysts at IDC expect more than 4 million AMRs to be operating in commercial environments worldwide.