The key constraint is not investor enthusiasm. It’s real-world data, battery life, edge chips, safety certifications, and the cost of deploying machines in a messy industrial environment.
While humanoids are gaining traction, short-term ROI still lies in dedicated automation, warehouse AMR, and specialized robotic systems.
Enduring winners are likely to be those with unique deployment data, clear labor bottleneck solutions, and robotics-as-a-service models that reduce upfront costs for customers.
Robots face the real world
Citi’s Robotics & Physical AI Leadership Conference left a clear message that physical AI is no longer just lab talk or venture capital slide deck. We are moving from proof of concept to commercial deployment. But the catch is just as important. This is not a scaling curve like the chatbot boom. Robots don’t live in a clean digital sandbox. They work in warehouses, factories, ports, trucks, construction sites, and defense environments, creating friction in every corner of the real world.
It is becoming easier for the demand side to understand. Labor shortages, reshoring, tighter supply chain controls, and stronger regulatory trends are driving companies to automate their physical economies. The problem is that deploying a robot is not the same as releasing a new software model. This means hardware, safety certifications, batteries, sensors, chips, maintenance, customer training, and data loops that actually need to be built rather than collected from the internet.
That’s the real difference between digital AI and physical AI. In large language models, the base model can carry much of the value. In physical AI, the values are much closer to the ground. Unique real-world data, task-specific deployment history, safety performance, and the ability to solve one expensive labor bottleneck at a time outweigh the blanket promise that robots will do everything.
At the conference, they also gave concrete numbers on the size of the stakes. Roughly $20 billion has been spent on physical AI over the past two years, with applications spanning warehousing, logistics, trucking, construction, aviation, and defense. Humanoids are in the spotlight, especially as companies like BMW test upgraded humanoid robots on factory floors. However, short-term returns on investment are still likely to come from purpose-built systems, warehouse autonomous mobile robots, and specialized automation platforms rather than a general-purpose humanoid revolution that will arrive overnight.
Data remains a challenge. Even the tens of millions of hours of robot data expected to be collected in 2026 may still be just the foundational point of what will ultimately be needed for high-level robot performance. That line is important. This shows that the market isn’t just a trade where you throw money at a theme and wait for magic. Physical AI requires iterations, scars, edge cases, and real operating history. Robots need to learn not only what the tasks look like in the demo, but also how they fail on bad weather days, crowded walkways, poor lighting, low battery power, and situations where human workers are too close.
Hardware bottlenecks are equally important. Power, battery life, and chip architecture are all constraints. Most current semiconductor platforms are built for data center workloads rather than real-time edge inference on mobile platforms. This means that the physical AI stack still requires its own industrial neural system. Robots need to think locally, move safely, conserve power, and respond instantly. This is a completely different issue than serving tokens from a server rack.
The takeaway from the industry for investors is that companies exposed to automation should continue to benefit over long cycles. Rockwell Automation, Emerson Electric, Honeywell, Symbotic, Ralliant, and Belden were prioritized for pure automation, warehouse automation, sensors, test and measurement, and industrial networking. The logic is simple. As companies further automate their physical economy, robots won’t be the only picks and shovels. These are the controllers, sensors, software, networking, safety systems, and industrial architecture that allows machines to operate with less human intervention.
Robotics-as-a-Service may be one of the more significant business model changes. This is important because the list price of automation can scare away small and medium-sized customers. RaaS changes the conversation from heavy upfront investment decisions to more manageable operational cost decisions. This should help increase adoption, especially in warehouse and logistics environments where labor is obvious, tasks are repetitive, and ROI can be measured in throughput, uptime, and accuracy.
In my view, this is where the trade becomes more interesting and less obvious. Physical AI is not just a meme version of robotics with a new label slapped on it. That is, AI is slow to connect to the physical economy. So this theme could be bigger than the anthropomorphic headlines, but it would also be slower, messier, and more industrial than the market would like to admit. Money doesn’t just chase robots on the factory floor. Track who owns deployment data, manage safety envelopes, reduce deployment costs, and transform automation from a scientific project to an operational tool.
The danger is extrapolating the digital AI curve to a physical world that refuses to scale cleanly. Chatbots can improve by responding to billions of prompts. Robots improve by overcoming real tasks. This is difficult to create a flywheel, but once it starts spinning, it has great protection.
