For years, automation procurement asked about payload, uptime, cycle time, integration, and ROI.

Now add another question:

Where does the robot come from, and what happens if that suddenly matters?

🌍 Robotics Just Became a Supply-Chain Risk

This week offered two very different pictures of the same robotics market.

In Beijing, more than 300 companies gathered for the World Robot Conference, with over 2,000 exhibits and more than 150 product launches. The focus is shifting quickly from robot demonstrations toward real work: parcel sorting, assembly, material handling, retail tasks, and other repetitive operations.

At the same time, Unitree made a spectacular Shanghai market debut.

Its shares closed 460% above the IPO price, valuing the company around $50 billion. Unitree raised roughly $900 million and is already profitable, although relatively few of its robots are operating in commercial environments.

That alone would make this a remarkable robotics week.

But there is another side to the story.

In July, the U.S. Federal Communications Commission added foreign-made humanoids, quadrupeds, and other mobile robots to its covered equipment list. New affected products cannot receive normal FCC equipment authorization without conditional approval.

The restriction applies broadly to foreign suppliers, not only Chinese manufacturers. Existing authorized products are not automatically removed from the market.

For warehouse and manufacturing leaders, this changes robotics procurement.

Until now, most automation evaluations treated country of origin primarily as a sourcing consideration.

It is becoming an architecture consideration.

A connected robot is not simply steel, motors, batteries, and sensors.

It contains cameras. Maps facilities. Connects to wireless networks. Receives software updates. Generates operational data. Communicates with fleet-management systems. In some cases, it may eventually run foundation models trained or updated remotely.

That means procurement teams need to understand more than price and performance.

Ask:

• Where is operational data stored?
• Can the vendor remotely access the system?
• Which components depend on foreign suppliers?
• What happens if future models cannot be imported?
• Can software updates continue during a regulatory dispute?
• Are spare parts available from multiple regions?
• Can the fleet operate if cloud connectivity disappears?
• How difficult would it be to replace the platform?

These questions may sound like cybersecurity or procurement questions.

They are becoming operations questions.

The risk is not theoretical.

China represents an estimated 82% of global humanoid shipments, according to IDC figures cited by Reuters. Meanwhile, U.S. companies already use mobile robots from China and Europe partly because imported equipment can be less expensive.

This creates an uncomfortable reality.

The fastest-moving robotics ecosystem may not always be the easiest ecosystem to procure from.

And the safest sourcing decision may not always deliver the most advanced hardware.

The answer is not to avoid international robotics vendors.

It is to stop treating automation equipment like an isolated capital purchase.

Robotics now belongs inside supply-chain resilience planning.

For every strategically important automation platform, understand your second source, your data exposure, your replacement path, and your dependency on the vendor’s software stack.

The next robot you buy may work perfectly.

The bigger question is whether your operation can keep working if the environment around that robot changes.

🧠 Supporting Insights

📦 Amazon Is Standardizing Another Automation Layer

​Amazon has entered a global framework agreement with AutoStore covering potential procurement of its automated storage and retrieval technology.

The agreement does not include purchasing commitments, but it establishes terms for future systems and services.

The significance is less about one AS/RS vendor and more about architecture.

Amazon has historically developed substantial automation internally. A global framework with an established outside platform suggests that even the most technically capable operators do not need to invent every automation layer themselves.

The strategic question is where proprietary technology creates differentiation, and where standard platforms create speed.

Vertical integration is valuable until it becomes reinvention.

🧠 AI Productivity Is Starting to Become Measurable

​Kimberly-Clark is combining procurement software, AI agents, and workforce development as part of its supply-chain transformation.

One AI-powered manufacturing-agent pilot produced a reported 40% to 50% productivity improvement in knowledge-search tasks.

That metric is interesting because it moves the AI conversation away from generic “copilot” adoption.

Operations leaders should measure agents around a specific unit of work:

How long did the task take before, and how long does it take now?

Knowledge retrieval, troubleshooting, supplier research, maintenance diagnostics, and exception resolution are good places to start.

AI becomes operational when the productivity gain can be measured without mentioning AI.

🛒 The Store Is Becoming the Distribution Center

​Home Depot has rolled out three-hour-or-less express delivery nationwide, using more than 2,000 U.S. stores alongside its broader supply-chain network.

The company says most express deliveries are already occurring in under one hour in applicable markets.

Meanwhile, Walmart expanded its 30-minute-or-less service to 38 U.S. markets and says customers using faster delivery shop more frequently.

This is not simply a last-mile story.

It changes inventory positioning.

When stores become fulfillment nodes, inventory accuracy, picking productivity, replenishment, and local automation become part of the delivery promise.

The warehouse is moving closer to the customer, sometimes all the way into the store.

🚁 Thirty Minutes Is Becoming a Logistics Category

​Amazon plans to expand Prime Air drone delivery to roughly 500 U.S. locations by year-end, up dramatically from its current footprint.

Each hub serves customers within a limited radius, and the system targets lightweight orders that can arrive in roughly 30 minutes.

Drone delivery will remain a niche compared with vans for some time.

But strategically, that misses the point.

The delivery network is segmenting.

Truckloads, parcels, same-day vans, store fulfillment, gig delivery, lockers, autonomous vehicles, and drones can each serve different order profiles.

The future last mile will not have one winning transportation mode.

It will have an orchestration problem deciding which mode should fulfill each order.

⚡️ Snippets

• ​Gravis Robotics raised $200 million to expand autonomous construction equipment. Physical AI is spreading beyond warehouses into messy, unstructured environments where perception and reliability matter more than perfect factory conditions.
• ​Generalist is showing how human demonstrations can become reusable robot behaviors. The real robotics data race is not about collecting more video, it is about collecting physical interactions that teach machines what successful work looks like.
• ​AiMOGA Robotics says it has already deployed more than 3,000 robots, including about 2,000 outside China, and is considering an IPO. Robotics companies are beginning to pursue international scale much earlier in their lifecycle.

Robotics used to be a machine-selection problem.

Then it became a software problem.

Now it is becoming something larger:

a technology, cybersecurity, procurement, and geopolitical supply-chain problem at the same time.

The robot on your floor may have a payload rating.

It also has a passport.

Start managing both.