Amazon's Automation Army

Hey there,

Last week we covered autonomous trucks going public and humanoid robots hitting walls. This week, the story splits across two fronts: warehouses quietly deploying robots at scale, and the military making a hard bet on cheap drones.

The warehouse automation market isn't coming. It's here. And the scale is staggering.

Amazon now operates over 1,000,000 robots across its global network. That number comes from June 2025 data, and it's not a typo. One million robots, handling picking, sorting, moving shelves. The company that changed how we think about e-commerce just changed how we think about automation scale.

Locus Robotics reported 350+ site deployments in April 2025, with robots that boost productivity 2-3X and cut labor costs in half. Companies like UPS, CEVA, and Fleet Feet doubled their productivity with fewer people because the robots assist rather than replace. Anthony Pendola at Fleet Feet said it plainly: "Our associates are going home less tired, and we've seen a big boost in efficiency."

The dream of "lights-out" warehouses died somewhere between 2019 and 2025. What replaced it: collaborative robots working alongside humans. The shift happened because full automation costs too much and breaks too often. Collaborative systems let you start small, add robots during peak season, scale back when things slow down. No massive infrastructure overhaul required.

The numbers back this up. The warehouse automation market grew from $30.9 billion in 2023 and will hit a 15.93% CAGR through 2032. By 2025, the US needs an additional 1 billion square feet of warehouse space. To optimize those facilities, around 4 million robots will deploy, with 25% of warehouse space dedicated to online fulfillment.

Design lesson: the winning systems aren't the most sophisticated. They're the ones that work with existing infrastructure, deploy in days not months, and can be operated by workers with minimal training. Locus robots navigate warehouses on their own using sensors and maps. A worker picks items, the robot carries them. Simple. The sophistication happens behind the scenes in the routing algorithms, not in the interface.

The Robot Report, June 15, 2025
Locus Robotics Press Release, April 8, 2025
Warehouse Automation Market Report, 2024

The robotaxi business model just got a stress test. Three companies are betting $200M+ it works.

In September 2025, Lucid delivered its first Gravity SUV for autonomous robotaxi retrofit. The plan: deploy 20,000 of these across the US over six years, operated exclusively through Uber's platform.

This partnership matters because it shows three different approaches combining. Lucid builds a 450-mile range EV platform designed for high utilization. Nuro provides the autonomous driving stack (Nuro Driver). Uber supplies the network and operations at scale.

The economics only work if the vehicles run constantly. A robotaxi sitting idle loses money. Lucid's long range and software-defined architecture help maximize uptime. Nuro's AI handles the actual driving. Uber's 34 million daily trips across 70 countries provide the demand.

Compare this to GM's Cruise, which shut down its robotaxi division earlier this year. The difference: Cruise tried to own the entire stack and operate the service. This new model splits responsibilities.

Meanwhile, Waabi raised $200 million in June 2025 (led by Uber and Khosla Ventures) to deploy fully driverless trucks in 2025. Kodiak went public at $2.5 billion in September 2025. The autonomous trucking segment is moving faster than robotaxis because the use case is simpler: highway driving between fixed points.

China leads deployment. Baidu's Apollo Go operates autonomous taxis in Wuhan with plans for 500 vehicles. Waymo delivers approximately one million rides per month in early 2025, putting it on track for 25-50 million rides this year. The World Economic Forum projects robotaxi fleets will operate at scale in 40-80 cities by 2035, with China and the US dominating rollout.

Why this matters: Splitting the stack works when each partner focuses on their core strength. Trying to own everything creates bottlenecks. The companies winning in autonomy are the ones willing to integrate, not insulate.

TechCrunch, September 12, 2025
Waabi Funding Announcement, June 18, 2025
Waymo Operations Report, January 2025

The military is fielding thousands of cheap autonomous drones. The design patterns will influence civilian systems.

In July 2025, Defense Secretary Pete Hegseth issued a memo titled "Unleashing US Military Drone Dominance." The goal: every Army squad equipped with small attack drones by end of fiscal 2026.

The policy changes matter for anyone building autonomous systems. Drones are now consumable commodities. Previously, small drones were classified as "durable property." If one crashed, soldiers faced investigation. That bureaucracy killed adoption. Now drones under $2,000 are treated like ammunition. Use them, lose them, order more.

Volume beats sophistication. The Navy wants carrier-based autonomous combat drones for $15 million each, versus $30 million the Air Force spends. In September 2025, the Navy awarded contracts to Anduril, Northrop Grumman, Boeing, and General Atomics for designs. The focus: cheap enough to field in quantity.

Ukraine proved the concept. Ukrainian officials say 80% of their success hitting targets comes from drones. Not artillery. Not missiles. Drones. Both sides field tens of thousands. The side with more wins, even if individual units are less capable.

The US military is playing catch-up. CNN reported in September 2025 that making drone parts in America costs "literally 100 times more expensive" than using Chinese components. The US banned Chinese parts in weapons systems, creating a supply chain problem.

Design lesson: design for expendability and volume, not perfection and longevity. When unit cost drops low enough, you can field more systems and take more risks. The same principle applies to civilian robotics at different price points.

Department of Defense Memo, July 2025
Navy Drone Contracts, September 15, 2025
CNN Ukraine Drone Report, September 22, 2025

While everyone watches cars and drones, autonomous underwater vehicles are quietly maturing.

The University of Michigan opened a 10,000-gallon robotics water tank in 2025 specifically for autonomous underwater vehicle (AUV) research. The facility enables testing of swarm coordination, multi-robot systems, and vehicles that transition between land and water.

The underwater robotics market will cross $13.59 billion by 2032, growing at 14.59% CAGR. Applications span offshore energy, marine research, defense, and infrastructure inspection.

The challenges in underwater autonomy differ from land or air: communication is hard (acoustic signals instead of radio), navigation requires different sensor approaches, and recovery matters because these systems cost $90K-$250K each. When you can't maintain constant communication with an autonomous system, you need better planning algorithms and more robust failure modes. Those lessons apply everywhere.

Boxfish Robotics showcased tetherless autonomous underwater vehicles at Ocean Business 2025 in April, with systems designed for deep-sea research and environmental monitoring with minimal human intervention. Companies like Cellula Robotics supply survey-grade AUVs to universities, commercial firms, and naval forces globally.

Why this matters: The design patterns emerging from underwater robotics inform other domains. Limited communication forces better autonomy. Harsh environments force better fault tolerance. High unit cost forces better recovery systems. Each constraint creates innovation that transfers to other markets.

University of Michigan News, March 2025
Marine Robotics Market Report, 2025
Ocean Business Conference, April 2025

Systems designed to be expendable often outperform systems designed to last forever.

Military drones at $2,000 each get used aggressively because losing one doesn't matter. Warehouse robots that cost $90K-$250K need to last years to justify the investment. Underwater AUVs at similar prices require recovery systems. The design philosophy shifts based on unit economics.

The military targets high-volume, low-cost systems designed for single missions or short lifespans. Prioritize ease of deployment over durability. Warehouse robots sit in the medium range, designed for repair and reuse over 2-5 years with modular components. Deep-sea AUVs and space robotics occupy the high-cost category, requiring extensive diagnostics and recovery systems.

Most civilian autonomous systems target the middle category. They need to work reliably for 2-5 years, be repairable, and handle thousands of missions. The military is proving that moving down-market to the high-volume category unlocks different capabilities.

The question for product teams: are you designing for the right price point and lifespan? Sometimes cheaper and more disposable wins.

1,000,000: Robots operating in Amazon warehouses (June 2025)

350+: Locus Robotics site deployments (April 2025)

20,000: Lucid robotaxis planned for Uber over 6 years (Sept 2025)

$15M: Navy target cost per autonomous combat drone

80%: Ukraine's target hits that come from drones (Sept 2025)

$13.59B: Underwater robotics market by 2032

Map your autonomy economics.

For any autonomous feature you're building: Calculate cost per unit. Estimate useful lifespan. Determine acceptable failure rate. Ask: would going cheaper and more disposable actually work better?

Sometimes the answer is no. But sometimes designing for expendability opens new use cases.

That's it for this week.

The pattern across all these stories: volume and operational reality beat sophistication and perfect engineering. Build systems that work in the real world, not the lab.

Until next week,
Rob

Forward this to someone deploying robots or autonomous systems.