The annual Citigroup Robotics and Physical AI Leadership Summit concluded this Tuesday, bringing together founders, investors, operators, and industry executives to assess the current state of "Physical AI."

In a summary for clients on Wednesday morning, Citi analyst Heath Terry outlined the core takeaway: the industry is progressing from proof-of-concept to commercial deployment, but scaling robotics remains a significant hurdle.

Terry told clients that labor shortages, manufacturing reshoring, and a favorable regulatory environment are accelerating enterprise demand. However, data scarcity, talent bottlenecks, battery life limitations, and high deployment costs remain major friction points.

Data is the Primary Constraint: Current Accumulation is a "Drop in the Ocean"

A central limitation repeatedly highlighted at the summit was the scarcity of data.

Instawork pointed out that even if the entire industry collects tens of millions of hours of real-world data by 2026, this volume would represent mere "basis points" relative to the total data needed for high-level robotic performance, not "percentage points."

An analogy made this clear: if the ultimate required data volume is a swimming pool, the data collected so far wouldn't fill a bucket.

Unlike digital AI, where large language model foundation models carry most of the value and can be rapidly replicated, the core value of physical AI lies in proprietary, task-specific data collected in real environments, combined with specialized hardware and safety certifications.

This means that for nearly every new scenario or task, data accumulation must start almost from scratch.

Additionally, power supply, battery life, and chip architecture are emerging as critical bottlenecks. Attendees noted that existing semiconductor platforms are designed for data center workloads, not optimized for real-time edge inference on mobile platforms.

Who is Succeeding? The Starting Point is "Solving a Real Pain Point"

Companies showing the fastest commercial progress—whether in humanoid robots, warehouse Autonomous Mobile Robots (AMRs), autonomous trucks, or construction robots—exhibit a similar success path.

Terry believes that the near-term drivers of investment returns are specialized AMRs and systems from companies like Locus Robotics and Dexterity, rather than the highly publicized general-purpose humanoid robots.

While humanoid robots have attracted substantial investment enthusiasm, near-term commercial returns are still primarily coming from these "specialized machines."

$20 Billion Inflows, with Logistics, Warehousing, and Automotive as Key Battlegrounds

Over the past two years, the physical AI sector has attracted approximately $20 billion in cumulative investment, with applications spanning warehousing, logistics, trucking, construction, aviation, and defense.

Last week, BMW disclosed that upgraded humanoid robots are now operational on the production line at its Spartanburg, South Carolina plant.

On the demand side, several summit participants identified logistics, warehousing, and automotive manufacturing as the current core end-markets for automation adoption. These scenarios share common characteristics: high-frequency, highly repetitive tasks suitable for robotic replacement.

Persistent labor market tightness and accelerated domestic manufacturing reshoring are two key structural drivers pushing automation demand.

Automation can enhance production capacity, increase equipment uptime, and improve operational efficiency and precision, thereby supporting a healthy return on investment.

RaaS Model: A Key to Unlocking the SME Market

High upfront costs have long been the biggest barrier to robot adoption for small and medium-sized enterprises (SMEs). The emergence of the "Robotics-as-a-Service" (RaaS) model, which converts a one-time capital expenditure into a usage-based operational expense, significantly lowers the adoption barrier.

Terry specifically highlighted Symbotic's "Warehouse-as-a-Service" product (GreenBox/Exol), suggesting this model helps extend warehouse automation solutions to a broader customer base, including SMEs previously deterred by cost.

A Decade-Long Marathon

Terry's final assessment was clear and direct: physical AI is a decade-long buildout, not a rapid explosion like chatbots.

Advancements in AI and large language models, along with the growing availability of real-world and simulation data, are driving continuous technological iteration—with deeper hardware-software integration and systems becoming "smarter" through accumulated use.

However, this process is gradual, not a sudden leap.

Citi believes long-term value will accrue to companies that master the data flywheel, solve real-world deployment problems, and achieve the highest safety standards.