First, the conclusion
As of 2026, the fusion of robotics and AI is a little dangerous if we talk about it only in terms of "dream humanoid robots."
What really drives business is the more humble front lines. Transportation within warehouses, piece picking, visual inspections, sorting at food factories, cleaning of commercial facilities, security at stations and airports, inspections of bridges and plants. We don't have enough people, we can't harvest food, we can't turn it around at night, and the quality varies. At such sites, the reason for introducing AI robots has become clear.
What has changed with the advent of generative AI is that robots have moved from being mere "moving machines" to systems that can see, make decisions, and move according to the situation. Trends such as NVIDIA's Isaac GR00T, Jetson Thor, and Google DeepMind's Gemini Robotics and RT-2 are efforts to generalize robot brains. In addition, companies like Physical Intelligence, Figure AI, Covariant, and Skild AI are also competing for foundational robot models and general-purpose operational capabilities. It looks a little narrow as it is an article about NVIDIA only. In reality, we are talking about the industrial layer, which includes AI models, simulation, edge inference, actuators, SI, and even maintenance.
However, the market remains skeptical. PoC will increase. More demo videos will be added. The question is whether this will translate into monthly sales, recurring contracts, maintenance income, and gross profit margins.
The essence of what we are aiming for in 2030 is not the robot itself. In order to incorporate robots, we need to reconfigure on-site processes, collect data, and create a system that can be operated within the company. There is room for investment here.
What will change with robotics AI?
Traditional industrial robots are quite strong. However, what he was good at was repeating predetermined movements with high precision in a suitable environment. This is still effective in areas such as welding in automobile factories, surrounding semiconductor manufacturing equipment, and transporting electronic components.
What AI-fused robots are aiming for is outside of that.
| point of view | conventional robot | AI fused robot |
|---|---|---|
| operation | Repeat taught trajectory | Adjust by checking the situation with cameras and sensors |
| object | Parts with the same position and shape | Irregular luggage, bags, food, and daily necessities |
| Deployment environment | Dedicated lines, fences, fixed equipment | Warehouses, stores, existing factories, public spaces |
| value | Replacement of manual labor | Labor shortage, quality, nighttime operation, data digitization |
An obvious example is piece picking. The camera recognizes different-shaped products in a box, changes the way they are grabbed, and puts them into the next box. This can be done with conventional robots, but the more objects there are, the more difficult it is to create. AI robotics is trying to expand its range of applications by combining image recognition, grasp planning, and retrying after failure.
The other thing is language instructions. The robot breaks down instructions such as `put the red box back on the shelf'' and `pick up the parts that fell on the floor'' into tasks. Although this is still largely a matter of research and demonstration, it is significant that the basic model will be used in robots. If a software AI agent operates the work screen, physical AI operates the site itself.
The important point here is that we are not talking about AI robots immediately working like humans. Rather, in reality, it starts with a limited amount of work, and the scope of application expands as more data accumulates. Investors and management should not get this order wrong.
2030 market forecast
Forecasts for 2030 vary considerably depending on the definition. This is because the scope of inclusion is different for AI robots, AI in Robotics, Embodied AI, service robots, industrial robots, and humanoids.
| market segment | Main predictions | How to read |
|---|---|---|
| AI Robots | MarketsandMarkets forecasts $6.11 billion in 2025 to $33.39 billion in 2030, CAGR 40.4% | High growth market focused on AI-equipped robots |
| AI in Robotics | Grand View Research predicts $20.4 billion in 2025 to $182.7 billion in 2033, CAGR 32.0% | A broader look at the overall AI technology incorporated into robots |
| Overall Robotics | ABI Research predicts $50 billion in 2025 to $111 billion in 2030, CAGR 14% | Wide market including industrial, service, and mobile robots |
| Service Robotics | Grand View Research predicts $68.1 billion in 2026 to $107.8 billion in 2030 | Implementation areas such as stores, cleaning, delivery, medical surroundings, etc. |
| Humanoid Robot | ABI Research predicts $6.5 billion in 2030, Goldman Sachs predicts $38 billion in 2035. | Expectations are high, but as of 2030 there will be a wide range of definitions and speed of spread. |
What we want to read from this table is not the size of the numbers themselves.
Rather, the AI robotics market has a structure of `if narrowly defined, it is small, but with high growth,'' and `if broadly defined, it is already large, but the growth rate is slowing down.'' Looking only at humanoid robots, the dream is big, but it is easier to predict the actual demand for factory, logistics, and service robots until 2030.
This is a practical point for management. Just because the market will grow in 2030, there is no need to suddenly introduce humanoid robots across all companies. Companies that start with tasks that are easy to measure effectiveness and accumulate field data and operational know-how will have an advantage.
The stock price reaction is similar. Humanoids are highly popular, so they tend to be bought in anticipation. AMR, machine vision, reducers, actuators, edge AI, SI, and RaaS are more likely to show up in financial results sooner.
Points to see in the Japanese market
Market forecasts from overseas research companies are useful, but if you are looking at Japanese stocks, you should also look at domestic constraints.
In Japan, there has been a prolonged labor shortage in logistics, manufacturing, nursing care, facility management, cleaning, security, and infrastructure maintenance. According to the Ministry of Health, Labor and Welfare's general job placement system, the effective job openings-to-applicants ratio continues to be over 1:1, and the Labor Economic Trends Survey shows that there remains a feeling of a labor shortage. The introduction of robots is no longer a ``future technology'' but is becoming a realistic option for workplaces where robots cannot be used.
On the policy front, the Ministry of Economy, Trade and Industry has launched the AI Robotics Strategy Review Council, and the RING Project, which aims to solve local labor shortages through robot implementation, has also started moving forward. It is also important to think about robot-friendliness, which means creating facilities, elevators, entrance/exit, communications, and operational rules that are easy for robots to use. Even if the performance of the robot itself improves, if the facility does not respond accordingly, cleaning robots and delivery robots will stop midway.
This is an issue typical of the Japanese market. We need to look at not only companies that make hardware, but also reduction gears, servos, FA equipment, sensors, controls, machine vision, AMR, robot SI, maintenance, facility management, and BPO. Rather than deciding on individual stocks, I would like to see which companies have the ability to "enter the field and keep moving without stopping."
Business areas that are easy to implement
The areas that are most likely to be introduced initially are areas where the damage in the event of failure is limited, the scope of work is narrow, and the effects can be measured numerically.
Manufacturing and logistics will focus on AMR (autonomous mobile robots), AGV, piece picking, palletizing, and visual inspection. The reason for the investment is easy to explain because warehouse transportation is affected by labor shortages and fluctuations in e-commerce demand. Piece picking is difficult, but there is room to introduce it if you narrow down the objects and process.
For services and stores, serving, cleaning, security, and guidance are practical. Rather than replacing all human customer service, the robots will be used for simple movement, patrolling, floor cleaning, and routine guidance. From the store's point of view, the value is not only in countermeasures against labor shortages, but also in the ability to collect operation logs and congestion data.
Drones, quadrupedal robots, and AI image recognition will be combined for infrastructure inspection and disaster response. The more difficult it is for people to enter, such as bridges, tunnels, power transmission equipment, plants, and disaster sites, the stronger the reason for introducing it. The main objectives are to improve safety and frequency rather than reduce labor costs.
Demand is high in the medical and nursing care fields, but the order in which they are introduced should be carefully considered. It is easier to start with peripheral tasks such as transporting, monitoring, recording, cleaning, and medication support, rather than suddenly entrusting physical assistance. This is because of safety, demarcation of responsibility, and psychological resistance from users.
Three business trends towards 2030
The first is the generalization of RaaS (Robot as a Service).
Introducing robots requires initial investment, maintenance, repairs, software updates, and on-site adjustments. It is difficult for small and medium-sized businesses to take the plunge through outright purchases. RaaS reduces the implementation burden with monthly charges and performance-linked fees. For companies, it will be easier to try it out as an operating expense rather than a capital investment. Profitability is difficult to gauge unless you look at not only the number of units installed, but also ARR, MRR, churn rate, renewal rate, operating rate, maintenance costs, and gross profit per unit.
The second is the expansion from cobots to general-purpose humanoids.
Collaborative robots have expanded the idea of working safely alongside humans. The next issue is a robot that can operate in a human environment without significantly changing the layout of existing factories and warehouses. It is in this context that Tesla, Unitree, AgiBot, Boston Dynamics, etc. come into focus. However, the main battlefield until 2030 will not be full-fledged household robots, but limited tasks in factories, warehouses, research, and hazardous work.
The third is collaboration with AI infrastructure.
It would be difficult to put all the intelligence into a single robot. Learning is performed in the cloud or at an AI data center, and inference is performed using edge AI in the field. Edge computing platforms for robots like NVIDIA's Jetson Thor, simulation and development platforms like Isaac, 5G/6G, and digital twins come together as a set.
Here's the structure:
| layer | Main role | Points for investors to look at |
|---|---|---|
| cloud gpu | Large-scale learning, simulation, and digital twins | Training demand, GPU investment, model update frequency |
| Edge AI | Real-time reasoning in the field, low-latency control | Embedded AI such as Jetson series, power consumption, thermal design |
| robot body | Sensors, actuators, reducers, controls | Mass production, parts procurement, maintainability |
| On-site operation | SI, RaaS, maintenance, SLA, data collection | ARR, occupancy rate, churn rate, gross profit rate |
If you read robotics AI only based on GPU demand, you will be missing out. Learning is supported by cloud GPUs, but low-latency, power-saving, and highly reliable edge AI inference is effective in the field. When the robot stops, the site stops. The reliability required is much heavier than chat AI.
This is also connected to the semiconductor and data center market prices. If demand for AI infrastructure expands beyond generated AI chat to include robots, vehicles, factories, and medical equipment, the quality of computing demand will change. However, buying a GPU does not mean you will make a profit. Only when it is confirmed that the robots can continue to operate and increase on-site productivity will there be a strong recovery story for AI infrastructure investment.
Companies that will remain profitable in 2030 will not just be companies that make robots. In fact, companies that can undertake everything from installation, operation, maintenance, software updates, and on-site data utilization are more likely to earn profits. Robot manufacturers appear to be the main players, but what investors really need to look at is the depth of recurring earnings surrounding them.
Common features of failing companies
Companies that tend to fail when introducing AI robots think that if they put robots in place of people, that's it.
In fact, it's the opposite. If robots are to be introduced, it will be necessary to reconfigure the flow lines of the site, shelf heights, floor differences, product masters, work instructions, exception handling, maintenance communications, and demarcation of responsibilities. If you don't have the idea to slightly change the site to match the robot, it's easy to get stuck in PoC.
Another is not to create in-house human resources.
Initial implementation is possible even if you leave it to the vendor. However, if the system cannot be determined within the company when the system stops on-site, when the object changes, or when the work rules change, operations cannot continue. We don't just need robot developers. He is an "operations manager" who knows field operations, looks at data, and can talk with vendors.
The last step is to introduce it without KPIs. Because of the lack of manpower, because of AI, because of subsidies. If it is included solely for these reasons, the effect will be ambiguous. If you do not measure work hours, number of vacancies, number of accidents, quality defects, night operations, maintenance costs, and training hours before introduction, it will not be possible to evaluate the system after introduction.
Implementation roadmap
A realistic roadmap is not to suddenly achieve full automation.
| step | things to do | Success conditions |
|---|---|---|
| try small | Use RaaS and rental for limited tasks such as transportation, cleaning, and inspection | Effectiveness can be measured without stopping the worksite |
| save data | Accumulate operation logs, failure patterns, work hours, and maintenance history | Improvement points can be seen in numbers |
| change the process | Redefining shelves, flow lines, work instructions, and exception handling to be based on robots. | The division of roles between humans and robots becomes clearer |
| expand horizontally | Expand to multiple locations and multiple processes | Lower implementation costs and standardize operations |
Companies that follow this order will be strong. Rather than making a big profit with the first introduction of a robot, we will turn the site into data and lower the unit cost for the next introduction. This is where RaaS, SI, maintenance, and software update businesses are born.
Merely saying ``Introduced'' in the presentation materials is not enough. Unless additional installations, repeat rates, operating rates, maintenance sales, ARR, and gross profit per unit are shown, the stock price will run out of steam at some point. The stronger the theme of a robot stock, the more likely it is to be anticipated.
Numbers that connect the workplace and financial results
The KPIs that management follows overlap considerably with the KPIs that are evaluated in financial results.
| KPI | Reason to watch |
|---|---|
| occupancy rate | Is it actually used in the field? |
| Task success rate | Can it withstand commercial operation rather than demo? |
| Number of human interventions | Measure autonomy and operational load |
| Payback period | Profitability including personnel costs, quality, and night operation |
| Maintenance costs/downtime | Affects the profits of the hardware business |
| Conversion rate from PoC to full implementation | Doesn't it end with proof? |
| Multiple location deployment rate | Has the operation been standardized? |
| MTBF・MTTR | Read failure frequency and recovery time |
| SLA achievement rate | Measuring the reliability of RaaS and maintenance contracts |
| ARR・MRR・Churn rate | See the quality of subscription revenue |
The stock market has already started pricing in robotics a little ahead of its time. In the areas related to AI semiconductors, FA, machinery, sensors, precision reducers, SI, and warehouse automation, there will be times when stock prices will not rise simply with good news.
The numbers are good. The problem is the inclusion.
The story that the market will be large in 2030 is a strong long-term theme. However, in the short term, stock prices respond to orders, gross profit margins, inventories, R&D expenses, exchange rates, capital investment in China, and interest rates. Robotics AI is a dream, but it is also a very realistic theme in financial results.
This may be a bit of a side note, but maintenance sales are often overlooked in robot IR. The number of PoCs and the number of devices introduced look impressive. However, the profitability of RaaS and SI becomes visible only when it continues to work in the field, maintenance contracts are renewed, and it is deployed horizontally across multiple locations.
Personally, I would like to prioritize the introduction rate and maintenance sales over the number of PoCs. Demo videos create expectations, but it is continued operation that creates profits. If we ignore this, robotics AI will become a ``great technology, but an unprofitable theme.''
risk scenario
The biggest risk is that implementation stops at PoC.
The robot moves well in demonstration experiments. However, in commercial settings, the objects change. People cross. The floor gets dirty. Communication is cut off. Stops at night. If these exceptions cannot be met, we will not be able to proceed with the actual implementation. Companies that only track the number of PoCs will be viewed with suspicion by the market.
Next, there is the risk of AI washing. Does the demo video show autonomous operation or remote control? How many attempts were made and how many times were they successful? Companies that do not disclose this information are likely to lose momentum when commercialization is confirmed, even if their products are bought in advance of expectations.
Geopolitics is also important. A robot is a mobile object that collects spatial data. As data from factories, warehouses, hospitals, and public facilities is handled, it is likely to be subject to U.S.-China tensions, export regulations, data crossing borders, and security screening. Even if China's low-cost aircraft are strong, there are restrictions on their introduction depending on the country and industry.
Finally, there is the capital-intensive nature of the hardware. Funds are needed for mass production lines, parts inventory, maintenance networks, quality assurance, and return handling. Unlike software companies, increasing sales does not necessarily mean increasing gross profit. Even if shipments increase, if gross profit margins decline and working capital increases, the stock market will cool down.
summary
Robotics AI is becoming a theme that will determine the competitiveness of companies towards 2030. In Japan, there is a combination of labor shortages, logistics constraints, an aging manufacturing workforce, and aging infrastructure, and the pressure to introduce it is structural.
However, this is not a problem that can be solved by buying a robot. Rather, companies that are able to change their workplace to incorporate robots are strong. Try it out on a small scale with RaaS, collect data, modify processes, and expand to multiple locations. It may seem simple, but this order will make a difference in 2030.
If you're just chasing flashy humanoid robots, you'll get a little detoured. AMR, machine vision, reducers, actuators, edge AI, SI, maintenance, and RaaS are faster and closer to profitability.
In the context of ESG, robotics AI is not just about saving labor. Contribute to CO2 reduction by reducing industrial accidents, moving dangerous work remotely, reducing the burden on aging workplaces, increasing the frequency of infrastructure maintenance, and reducing wasteful movement in logistics and factories. Companies that can see these effects as KPIs are more likely to be evaluated not just as theme stocks but also in the context of human capital and safe investment.
Physical AI is not the next headline in the generative AI market, but an implementation theme that promotes labor saving and data conversion in the field. It is too late to wait for full-scale adoption in 2030. In the end, which sites, which numbers, and how much have they improved? I'll be back there.
Related pages
- Unitree Robotics IPO分析｜フィジカルAI株として見る強みとリスク
- Meta Compute報道はGPU供給過剰の始まりか？ AIクラウドと半導体株急落の深層
- AIエージェントは理系採用を変えるのか：AI採用・ES無力化と採用DX銘柄の読み方
source
External information was confirmed on July 7, 2026. Since each company has a different definition of market forecast, this text treats it as a range rather than a single forecast.
- MarketsandMarkets, Artificial Intelligence (AI) Robots Industry worth $33.39 billion by 2030
- Grand View Research, Artificial Intelligence In Robotics Market Report, 2026-2033
- Grand View Research, Industrial Robotics Market Size, Share | Industry Report, 2030
- Grand View Research, Service Robotics Market Size And Share Report, 2024-2030
- ABI Research, The Global Robotics Market Outlook
- Goldman Sachs, The global market for humanoid robots could reach $38 billion by 2035
- NVIDIA, Isaac GR00T - Generalist Robot 00 Technology
- NVIDIA, Jetson Thor | Advanced AI for Physical Robotics
- Google DeepMind, Gemini Robotics
- Google DeepMind, RT-2: New model translates vision and language into action
- Physical Intelligence, π0
- Figure AI, Helix
- Covariant, The Covariant RFM-1
- Skild AI, Skild Brain
- Tesla Investor Relations, Tesla Releases First Quarter 2026 Financial Results
- Unitree Robotics, Humanoid Robot - Unitree Shop
- AGIBOT, AGIBOT WORLD
- Ministry of Health, Labor and Welfare, 一般職業紹介状況(令和8年4月分)について
- Ministry of Health, Labor and Welfare, 労働経済動向調査：調査の結果
- Ministry of Economy, Trade and Industry, AIロボティクス戦略検討会議
- Ministry of Economy, Trade and Industry, Robotics & Regional Initiative Networking Group (RING Project) Established to Solve Regional Labor Shortages
- METI Journal, 霞ヶ関のコンビニでロボットが活躍中