Key facts
- Physical AI is shifting from concept to commercial deployment, facing scaling challenges.
- Data scarcity, battery limitations, and chip architecture are identified as key bottlenecks.
- Robotics-as-a-Service models are crucial for lowering customer adoption barriers.
- Citi's top industrial automation picks include Rockwell Automation, Emerson Electric, and Honeywell.
- Approximately $20 billion has been invested in physical AI applications over the past two years.
Citi's annual Robotics & Physical AI Leadership Conference concluded Tuesday, with analysts identifying key trends and investment opportunities in the burgeoning sector. Analyst Heath Terry noted that the industry is progressing from proof-of-concept stages to commercial deployment, though significant operational challenges remain in scaling robotic solutions.
Key drivers accelerating enterprise demand include labor shortages, reshoring initiatives, and favorable regulatory environments. However, data scarcity, talent constraints, battery limitations, and high deployment costs present persistent friction points. Citi suggests that companies possessing proprietary real-world data, addressing specific labor bottlenecks, and utilizing Robotics-as-a-Service (RaaS) models to reduce upfront customer expenses are best positioned for success.
Citi's preferred industrial automation stock picks include Rockwell Automation, Emerson Electric, and Honeywell, alongside Symbotic, Ralliant, and Belden. The firm views these companies as well-positioned to benefit from increased investments in automation, supported by a constrained labor market and expanding domestic manufacturing.
Humanoid robots are attracting considerable investor interest, with approximately $20 billion invested in physical AI applications across various sectors like warehousing, logistics, and defense over the past two years. BMW recently showcased upgraded humanoid robots operating on its factory floors in South Carolina, highlighting the growing integration of such technologies.
Conference participants consistently identified data scarcity as a critical constraint, with estimates suggesting that even extensive data collection will represent a small fraction of what is ultimately needed for high-level robotic performance. Panelists also pointed to power, battery longevity, and chip architecture as emerging bottlenecks, noting that current semiconductor platforms are not optimized for real-time edge inference on mobile devices.
Commercially advanced companies in areas such as humanoids and autonomous trucking typically address specific, high-pain labor problems, employ RaaS models, and prioritize safety and reliability. While humanoids are generating significant enthusiasm, near-term returns are currently driven by purpose-built autonomous mobile robots (AMRs) and specialized systems from companies like Locus Robotics and Dexterity. Citi anticipates physical AI to be a decade-long buildout, favoring companies with strong data capabilities, practical deployment solutions, and high safety standards.