Having intelligent robots assisting or working alongside humans is no longer limited to fiction.
Citi predicts that the global AI robot population will hit 1.3 billion by 2035, and increase to 4 billion by 2050. Integrated with artificial intelligence (AI), those robots analyse data from their sensors to perceive and understand their surroundings before responding in real time.
Investors are optimistic that AI robots will soon become mainstream. According to a Citi GPS report, venture capital investment in robotics reached US$10 billion ($13.46 billion) in 2023, of which 38% goes to Asia.
Almost half of all global deal activity for robotics in recent years has also been in Asia, and China has accounted for 78% of all robotics patents over the past two decades. What’s driving this optimism and where are the growth opportunities for AI robots?
Humanoids for industrial work
See also: Serial Achieva wins US$15.62 million deal to supply Nvidia-powered servers
AI advancements have enabled the development of humanoids or robots that resemble humans in shape. For instance, multi-modal AI allows humanoids to see, move and communicate with humans to learn, understand tasks and action them. Some humanoids have also shown their dexterity, such as Boston Dynamic’s Atlas being able to do back flips and parkour.
As the capability of humanoids continues to advance with improvements in intelligence and dexterity, it presents a large new market opportunity. Humanoids can not only help companies reduce labour costs by taking on manual and repetitive tasks but also improve safety in certain situations.
Citi estimates that the total addressable market for humanoids will reach US$209 billion by 2035 and there will be 13.3 million humanoids serving seven use cases globally.
See also: Chinese AI firm SenseTime cuts back on Singapore office space
Industrial use cases are expected to be a major driver of AI robot adoption. For instance, manufacturers can address the labour shortage using automated mobile robots (AMRs). These robots use advanced cameras and AI processors to continuously map their environment while keeping track of their own location via a real-time information feed so that they can navigate the factory floor independently.
Beyond improving operational efficiency, the Citi GPS report foresees AI robots being used in product design simulations and anomaly detection such as identifying methane leaks in pipelines.
Given its various applications, the global unit number of AMRs and automated guided vehicles running on factory floors or in warehouses is expected to increase from 2.4 million in 2024 to 28.3 million by 2035.
Serving humans
The convergence of an ageing population, caregiver shortages, and escalating healthcare costs presents a compelling opportunity for humanoid robots.
Given that approximately 70% of elderly individuals are open to robotic assistance with daily tasks, Citi estimates that 18 million units of caring robots will be deployed by 2035, and the figure will hit 71 million by 2050.
Remote monitoring is a key application of moving AI robots in healthcare. Telepresence robots, like the Giraff robot, enable the elderly to remotely interact with family members and doctors through a screen and camera mounted on a mobile robotic platform. The Giraff robot can also perform tasks such as medication dispensing and vital sign monitoring.
To stay ahead of the latest tech trends, click here for DigitalEdge Section
For patients with limited mobility, physical assistance robots can provide support with activities of daily living, including lifting, transferring, and dressing. Robots can also aid caregivers such as delivering medicine or laboratory materials safely.
Apart from healthcare, the labour-intensive hospitality sector will increasingly embrace humanoids. Case in point: 9.6 million humanoids will work in the hospitality sector by 2035, taking on tasks such as serving food in restaurants or delivering luggage to hotel rooms.
This is exemplified by Haidilao in Beijing and Shanghai, which operate without human intervention. Pudu Robotics’ Bella robots lead customers to their seats, serve food based on natural language instructions, and use 3D vision to move around without bumping into customers.
“By leveraging advanced AI technologies — such as large language models (LLMs), computer vision, machine learning, sensor fusion, deep learning, and topological navigation — we make our robots intuitive, precise, and efficient. These innovations have accelerated the commercialisation and widespread deployment of our robots, significantly boosting productivity and reliability,” says Pudu Robotics’ founder and CEO Felix Zhang.
Delivery robots are another growth area for AI robots, with the global online delivery market expected to grow to US$3,818.5 billion by 2050. Robots can help with last-mile delivery service by transporting parcels or food and groceries, ensuring timely deliveries through route optimisation, finding addresses and communicating with customers.
Taking on dangerous tasks
Some of the most impactful AI robot applications will likely be in areas linked to safety, security and military, according to Citi’s report.
A prime example of a safety AI robot is Boston Dynamics’s Spot. The quadruped robot is designed to navigate diverse terrains, making it well-suited for inspections in hazardous environments such as nuclear facilities and chemical plants. Spot can inspect reactors, detect leaks, and assess other critical components, minimising human exposure to radiation and toxic substances.
Meanwhile, military robots are capable of performing a diverse range of tasks, including transportation, logistics, reconnaissance, search and rescue, and combat operations.
Enhanced cleaning robots, AVs
Advancements in AI will enhance the capabilities of existing intelligent robots too. For example, equipping robotic vacuum cleaners with LLMs will enable them to understand instructions in human language and work in a more customised way.
Citi predicts a surge in the adoption of AI cleaning robots in commercial buildings. The global unit number is projected to grow at a 23% CAGR, reaching 14.2 million by 2035. This is driven by cost-effectiveness compared to human labour, as exemplified by Skyline Robotics’ Ozmo.
The skyscraper window cleaning robot with six-axis robotic arms uses AI to autonomously assess building facades, calculate optimal cleaning paths, and adjust its cleaning technique in real-time. Ozmo is estimated to clean windows three times faster than traditional methods, while also reducing on-site labour costs by up to 75%.
Autonomous vehicles (AVs) are making significant strides. Many are currently at Level 3, capable of autonomous driving with human oversight.
Some are already operating at Level 4, driving without any human intervention. Among them are Waymo’s AVs in San Francisco, which currently deliver 100,000 rides per week.
Countries like Singapore and China are actively promoting the development and testing of AI-powered vehicles.
Citi foresees more advanced autonomous driving (where vehicles can perform all driving tasks and are responsible for monitoring the driving environment) to take off in 2027. As such, 50% of cars in 2035 are expected to have advanced driver assistance systems or achieve L3 autonomous driving.
Drones are another growth opportunity for AI robots. Levitate Capital forecasts the consumer drone market to reach US$5.3 billion by 2030, while the enterprise drone market is estimated to hit US$29 billion. Some enterprise applications for drones include environment inspection, disaster relief operation, automatic surveying and mapping, personnel safety monitoring, and irrigation and crop spraying.
Hurdles on the path to ubiquity
Although AI robots can help revolutionise industries, their path to becoming mainstream is riddled with hurdles spanning technology, regulation and public perception. Citi’s report reveals the major factors hindering mass adoption, including:
- High manufacturing cost of robots — The increased sophistication and capabilities of robots often come with a higher price tag. To ensure durability and reliability, commercial AI robots require robust construction, further driving up costs.
- Battery power — The energy density of current batteries, especially lithium-ion, limits the operational duration of robots. AI-powered robots’ high computational demands further strain battery life too. While these challenges persist, promising advancements in battery technology, such as graphene and solid-state batteries, offer potential solutions to enhance energy density and efficiency.
- Energy consumption— To mitigate the escalating energy demands of AI, there is a pressing need to develop more energy-efficient algorithms. This involves optimising existing models, developing new low-power AI models, and leveraging techniques like pruning and quantisation. Data centres (which form the backbone of AI) must also be optimised for energy efficiency, which can be achieved by improving cooling systems and implementing advanced energy management software.
- Electronic waste (e-waste) — Many robots are designed with specialised parts unique to a particular manufacturer or model, making it difficult and costly to repair or upgrade them. The short lifespan of robots contributes to the growing volume of e-waste, leading to environmental pollution and the loss of valuable materials. To combat this issue, we must support “right to repair” initiatives that advocate consumers’ ability to repair their devices instead of being forced to replace them.
- Privacy concerns — Home robots can collect vast amounts of personal data, including visual, auditory, and behavioural information. While this data can enhance user experience, it also poses privacy risks if not handled securely.
- Regulatory grey zones — The increasing autonomy of robots is pushing the boundaries of traditional legal frameworks. A critical question arises: Who is liable when a robot causes harm? As robots make decisions with minimal human intervention, there is a need to develop new legal standards to ensure accountability and address the ethical implications of AI.
- Public scepticism — Public trust in AI is hindered by the low level of explainability, transparency, and causality in AI systems. Many AI models, particularly those based on deep learning, operate as “black boxes”, making it difficult to understand their decision-making processes. This lack of transparency hinders trust, especially in high-stakes scenarios.
- Workforce resistance — Historically, the introduction of automation has led to job displacement and wage suppression for human workers. This trend is likely to continue with the deployment of AI-powered robots. However, new job opportunities will also emerge such as robot maintenance technicians, AI system trainers, and data analysts. Reskilling programmes are crucial to help workers adapt to the changing job landscape.
Addressing these challenges will require concerted efforts from governments, industries, and academia. The pace at which we overcome them will determine how far and fast the AI robot revolution will unfold and deliver transformational benefits across industries.
