Sona Comstar's DENSO partnership and its entry into Robotics and Physical AI were the two biggest strategic announcements from its latest conference call. The announcements themselves are easy to understand. What interests me more as an investor is what sits underneath them.
Are these genuinely large new markets? Does Sona have a reason to win in them? Are these businesses structurally better than the components Sona already sells? And, equally importantly, where is the market opportunity being overstated?
My current view is that the high-voltage business with DENSO has a relatively straightforward investment case. Electrification is increasing the value of motors and power electronics inside every vehicle, irrespective of whether India eventually prefers battery EVs, strong hybrids or plug-in hybrids.
Robotics is more speculative, but potentially much larger. Here, however, I am more interested in Sona supplying motors, precision gearboxes, actuators and perception systems than in its ambition to manufacture complete robots. The distinction matters.
The DENSO opportunity is not really a bet on hybrids
It is tempting to look at the DENSO JV and immediately start forecasting strong-hybrid sales. That may actually be the wrong way to look at it.
The products relevant to Sona's high-voltage JV, mainly traction motors, motor-generators and liquid-cooled traction inverters, sit inside several different powertrain architectures.
- A battery EV needs a high-voltage traction motor and inverter.
- A strong hybrid may need a traction motor, another motor-generator and power electronics.
- A plug-in hybrid requires an even more capable electrical drivetrain because it must move the vehicle for meaningful distances without starting the engine.
The important underlying variable is therefore not hybrid penetration. It is high-voltage electrification penetration. This distinction matters because India's current data is giving us conflicting signals about which powertrain will win.
Electric passenger vehicles had a strong 2025. EV share increased to around 4.6% of passenger-vehicle sales, compared with roughly 2.6% a year earlier in one wholesale-based dataset. Strong hybrids reached around 2.6%.
The divergence became more pronounced in H1 2026.
India sold 139,065 BEVs during H1 2026, up 65% year on year, taking their share to about 5.4%. Strong-hybrid volumes were 57,885, almost exactly unchanged from H1 2025, giving them approximately 2.2% share.
This is interesting because 2025 had initially looked like the beginning of a large hybrid wave. Strong-hybrid sales had grown to around 1.07 lakh units, up more than 80% from 2024. But H1 2026 shows why extrapolating one year's growth can be dangerous.
The strong-hybrid market is still dependent on a handful of vehicles. The Hycross and Hyryder alone accounted for around 80% of H1 2026 volumes, and the entire segment was essentially flat as declines in models such as the Grand Vitara and Hycross offset growth elsewhere. That said, if more automakers introduce strong hybrids and plug-in hybrids, hybrid market share could rise sharply.
I have considerably more conviction that the electrical content inside Indian vehicles is rising.
If BEVs win, Sona-DENSO has a market. If hybrids gain share, it has a market. If PHEVs establish themselves as a middle ground, it still has a market. That makes the DENSO JV less dependent on predicting the winning powertrain technology.
The ₹24,000 crore TAM is large, but it needs to be pulled apart
Management cited S&P Global Mobility and said hybrid and electric passenger cars and commercial vehicles in India could exceed 2.3 million units in 2030. Sona translates that into an addressable opportunity of around ₹24,000 crore.
The first thing I wanted to know was whether that number makes sense at all. ₹24,000 crore divided by 23 lakh vehicles gives an implied average content of approximately ₹1.04 lakh per vehicle. That initially looked high to me. But a bottom-up check actually supports it.
The US Department of Energy estimates the manufacturing cost of a commercial 100 kW electric traction drive, consisting of one traction motor and one inverter, at around $1,140. Of this, approximately $756 is the motor and $384 the inverter. At current exchange rates, that puts us remarkably close to the ₹1 lakh range.
The comparison is not exact. Indian manufacturing economics differ, motor power differs between vehicles, and supplier selling price is not the same thing as manufacturing cost. But it gives us the right order of magnitude.
So I am reasonably comfortable with Sona's content-per-vehicle assumption. My skepticism lies somewhere else.
TAM is not the same as an open market
A ₹24,000 crore market does not mean ₹24,000 crore is available for Sona-DENSO to compete for.
Toyota already operates inside an established Toyota-Aisin-DENSO ecosystem. Its current Indian strong-hybrid drivetrain is already highly integrated with that supply chain.
Hyundai and Kia have Hyundai Mobis, which manufactures motors, power-control units and complete electric-drive systems.
Honda has its own e:HEV architecture and its relationship with Astemo.
SAIC and MG have access to a large Chinese component ecosystem.
These OEMs can still source components externally and suppliers change over time, but this is very different from entering a completely open market.
That is why I would not take the ₹24,000 crore number and casually assume Sona-DENSO can capture 20% or 30%. The more useful question is:
How much of this market will actually come up for competitive sourcing?
This is where Tata, Mahindra, Indian commercial-vehicle manufacturers and future localization programs become particularly important. These companies do not necessarily have a Hyundai-Mobis-style captive supplier that can provide an entire electrified powertrain stack internally.
The other opportunity is second sourcing. As EV volumes rise, OEMs may not want to depend on one supplier for a mission-critical component. Localisation requirements and cost reduction can also reopen sourcing decisions even when an OEM already has a global supplier. So my conviction is high on the existence of the market.
It is lower on the assumption that the entire reported TAM is contestable.
The more interesting number may be ₹11 to ₹12
One comment during the conference call stood out more to me than the ₹24,000 crore TAM.
Management was asked about the capex required for the high-voltage business.
For its existing two-wheeler motor business, Sona said ₹1 of capex could support ₹8 to ₹9 of revenue.
For high-voltage systems, management indicated that ₹1 of capex could potentially support ₹11 to ₹12 of revenue.
The explanation was:
"It is know-how, the number of iterations you've done... It is more engineering intensive."
This is potentially important. A business growing primarily through engineering rather than large amounts of incremental fixed capital can have attractive economics if it scales. The difficult part of an inverter is not simply buying a machine that manufactures inverters.
The value sits in semiconductor selection, power-module packaging, thermal management, cooling, control algorithms, reliability, electromagnetic compatibility and the accumulated iterations required to make all of these work together.
Management also explicitly said:
"High-voltage technology is not easy to crack."
and argued that competition is consequently lower than in the two and three-wheeler motor market.
I would still be cautious with the word "moat".
Bosch, Valeo, Aisin, Hyundai Mobis and several Chinese Tier-1 companies are formidable competitors.
But the barrier is clearly higher than simply manufacturing another mechanical automotive component.
If Sona can combine DENSO's high-voltage technology with its own ability to engineer products to Indian cost targets, the resulting business could have an attractive combination of high content per vehicle and relatively low capital intensity. That is the part of the DENSO opportunity where I have the strongest conviction.
The 2030 number itself requires aggressive growth
There is another reason I would treat the ₹24,000 crore TAM as a scenario rather than a guaranteed destination.
H1 2026 passenger-vehicle sales consisted of roughly 1.39 lakh BEVs and 58,000 strong hybrids.
Annualising that gives less than 4 lakh electrified passenger vehicles.
Sona's 2030 estimate talks about more than 23 lakh hybrid and EV passenger and commercial vehicles.
Commercial vehicles make the comparison imperfect, but the basic implication remains the same.
India needs several years of very rapid electrification for the 2030 estimate to be reached. That is possible.
EV passenger-vehicle volumes are currently growing 65% to 80% depending on whether wholesale or retail data is used. But maintaining anything close to those growth rates becomes progressively harder as the base increases.
So I would not build my investment case around exactly 23 lakh vehicles in 2030.
The more durable thesis is simply that high-voltage content per vehicle is large, penetration remains low today, and the penetration curve still has a long runway.
That is enough.
Robotics is a very different kind of opportunity
The robotics vertical interests me for almost the opposite reason.
High-voltage automotive systems have an identifiable market today. We can count vehicles, estimate motor power and approximate component content. Robotics is far more uncertain.
Forecasts for humanoids in particular are so different that they are almost unusable for valuation.
Sona's annual report cites Bank of America research forecasting 10 million humanoid robots in annual sales by 2035 and says Sona could potentially address more than 50% of a humanoid robot's bill of materials through its motors, gears, controllers and other technologies.
Goldman Sachs, meanwhile, has estimated around 1.4 million humanoid shipments by 2035 and a $38 billion market. 10 million versus 1.4 million is not a small forecasting difference.
It tells us that nobody really knows what humanoid adoption will look like ten years from now.
So I would not value Sona by taking a humanoid TAM, assuming a market share and multiplying it by Sona's potential content per robot. That would create precision where none exists.
Fortunately, Sona does not need humanoids to work for the robotics strategy to become meaningful.
The broader robotics market is already real
Industrial robotics does not depend on humanoids becoming mainstream. Globally, 542,000 industrial robots were installed in 2024. That was more than double the level ten years earlier. There were about 4.66 million industrial robots already operating globally by the end of 2024. Interact Analysis estimates annual industrial-robot shipments increased to roughly 550,000 units in 2025 and could exceed 760,000 by 2030, implying around 6.7% annual growth.
India is even more interesting. India installed approximately 9,100 industrial robots in 2024, a record, making it the sixth-largest market globally. Ten years earlier, annual installations were barely above 2,000.That works out to roughly 16% annualised growth over the decade.
India is now expected to record around 12.5% annual growth in industrial robot shipments between 2025 and 2030, the highest growth among major markets in Interact Analysis' forecast. This is an important distinction. I am skeptical about forecasting millions of humanoids.
I am not skeptical about industrial automation continuing to grow.
AMRs may grow even faster
Sona's decision to work on Autonomous Mobile Robots also sits inside a market that is already expanding quickly.
Interact Analysis estimates the global mobile-robot market was just below $5 billion in 2024 and could reach around $14 billion by 2030, growing approximately 19% annually. There is also a structural shift occurring inside this market.
Traditional Automated Guided Vehicles generally follow predetermined paths. AMRs use sensors, mapping and software to navigate more dynamically around people and obstacles.
Interact expects AGVs' share of mobile-robot revenue to fall from about 33% in 2024 to around 20% in 2030 as AMRs take a larger share. So Sona is entering the segment at a time when the underlying technology mix is itself changing. That does not mean Sona will win the AMR market. But it means the market it is targeting is not hypothetical.
The ₹800 crore order book deserves both attention and restraint
Sona already has an ₹800 crore Robotics and Physical AI order book. ₹600 crore of that came from three recently announced programs covering an advanced robotic subsystem, perception engineering for an AMR and a radar perception solution. Management also said production starts for these programs range from the current quarter to within approximately 15 months.
This gives me more confidence than a prototype announcement would.Customers are willing to award programs.
But there are two reasons I would not get carried away by the ₹800 crore number.
First, it represents only around 3% of Sona's total ₹24,000 crore net order book.
Second, Sona's order book represents expected revenue over the lives of awarded programs, potentially extending across several years. It is not annual revenue. So ₹800 crore is meaningful evidence of customer traction. It is not yet evidence that robotics will materially change Sona's earnings. The composition of the next ₹800 crore will tell us much more.
The product I am most interested in is not the AMR
If I had to pick one area of Sona's robotics roadmap with the best combination of adjacency and barrier to entry, it would be precision robotic gearboxes. Robots need an unusual type of gearbox.
The motor inside an industrial robot or humanoid joint can rotate extremely quickly. The robot joint needs slow, controlled movement while producing very high torque. A gearbox converts that speed into torque. But robotics adds another requirement: precision.
Even a tiny amount of backlash or mechanical play can affect the position of a robotic arm. For industrial robots performing welding, machining or assembly, repeatability can be critical. That makes precision reduction gears considerably more difficult than ordinary gearing. We can see the barrier in the competitive structure.
Nabtesco estimates it has approximately 60% of the global market for precision reduction gears used in medium and large industrial robot joints. It has developed this technology over roughly three decades.
Normally, a market with one company holding 60% share would make me cautious about a new entrant. In this case, it also tells us why the product could be valuable if Sona succeeds. Precision gear manufacturing is already one of Sona's foundational capabilities. The question is whether automotive precision can be taken to robotics-level tolerances, durability and backlash requirements.
I do not yet have enough evidence to say the answer is yes. But if Sona receives a meaningful production order for an industrial robotic gearbox, I would consider that a much more important milestone than launching another AMR prototype. It would demonstrate that one of Sona's strongest existing manufacturing capabilities has successfully crossed into a new industry with high technical barriers.
Actuators could be even more valuable
The next logical step after a precision gearbox is an integrated robotic actuator. This is where Sona's roadmap starts becoming genuinely interesting.
A robotic actuator can combine frameless motor + precision gearbox + encoder + bearings + sensing + controller into one compact joint.
Instead of selling one motor to a robot manufacturer, Sona could supply much of the complete joint. That increases content per robot significantly. It also combines capabilities Sona has accumulated from different places. Its traditional drivetrain business contributes gears. Its motor business contributes electric machines and controllers. NOVELIC contributes sensing, radar and embedded software. Its engineering organisation brings system integration.
This is the point where Sona's claim that it could theoretically address a large percentage of the humanoid robot BOM begins to make technical sense. Not because Sona will manufacture the robot's AI brain. But because a robot can contain many actuated joints. A humanoid may need actuators across shoulders, elbows, wrists, hips, knees, ankles and hands. If Sona supplies one component, its content is limited. If it supplies the complete motion module across several joints, the opportunity changes materially.
This is where I have the strongest conviction inside the robotics vertical. Not on humanoid volumes.
On Sona's ability to increase its content per machine by moving from component to subsystem.
Frameless motors alone are less exciting
Humanoids and cobots need compact, high-torque motors, making frameless motors an attractive market. But there are already established companies such as Maxon, Kollmorgen and several Chinese motor manufacturers. I therefore do not see a large moat merely in Sona manufacturing a frameless motor.
The product becomes more interesting when bundled with the gearbox, sensing and control system. The same reasoning applies to radar. Radar hardware itself will become increasingly competitive.
NOVELIC becomes more valuable if Sona can provide sensor + signal processing + perception software + integration rather than simply selling another radar module. The value moves upward as Sona takes responsibility for a larger part of the customer's problem.
Where I am most skeptical: building the complete robot
Sona intends eventually to manufacture selected complete robotic platforms, initially AMRs and cognitive cobots.
This is where my conviction falls. The AMR market already has established players. The cobot market includes Universal Robots, ABB, FANUC, Doosan, Chinese manufacturers and several specialist companies.
The difficult part of building a successful robot company is not just assembling motors, gearboxes and sensors. A complete robot needs navigation software, motion control, fleet management, safety systems, application software, integration with factories and warehouses, customer support, and an ecosystem of compatible tools. These are quite different capabilities from precision manufacturing.
Sona may build them over time, but it still has to prove this. More importantly, I am not sure Sona needs to become a successful global robot brand for this vertical to create substantial value. There may be a better position in the value chain.
The picks-and-shovels strategy may be more attractive
Nobody knows which humanoid manufacturer will dominate ten years from now. Nobody knows which AMR platforms will survive consolidation. But almost all of them will need motors, gearboxes, actuators and sensors. That creates an attractive supplier position. Sona does not necessarily have to pick the winning robot. It can supply the companies competing to build the winning robot. This is very similar to the position automotive Tier-1 suppliers have historically occupied. The manufacturer takes enormous platform risk. The component supplier tries to qualify its product across several platforms.
For Sona, this approach also uses capabilities it already has rather than requiring it to recreate an entire robotics software ecosystem. That is why I would much rather see ₹2,000 crore of future robotics orders coming from precision gearboxes + actuators + perception systems than ₹2,000 crore dependent entirely on Sona's own AMR platform succeeding. The first would validate the technological adjacency. The second introduces an entirely new business model.
Robotics also has one attractive characteristic that is easy to overlook
Automotive components are ultimately constrained by automotive production. Global light-vehicle volumes do not grow particularly quickly. An auto-component company therefore grows mainly by taking market share or increasing the value it supplies per vehicle. Robotics opens another route. Sona can take technologies developed for automotive use and sell them into an entirely different installed base. Its gearbox factory is no longer addressing only cars. Its motors do not have to address only EVs. Its sensors do not have to address only ADAS. This increases the addressable market without requiring Sona to abandon its existing technology base. That is a much more attractive form of diversification than buying an unrelated high-growth company.
The two verticals actually have something important in common
At first, DENSO high-voltage powertrains and robotics look unrelated. From an investor's perspective, however, the common theme is becoming clearer. Sona is trying to increase the amount of engineering value it sells per machine.
- A differential gear is one component.
- A differential assembly has more content.
- A motor adds more.
- A motor-controller system adds still more.
- A high-voltage electric drive combines multiple technologies.
- A robotic actuator combines the motor, gearbox, electronics and sensing again.
- The progression is essentially component → assembly → subsystem → system.
Sona itself describes its technology roadmap in these terms.
This is the part of the strategy I find most compelling. The company does not need automobiles to disappear.
It does not need humanoids to arrive next year. It needs to keep increasing the value of the products it can engineer and manufacture for every machine it serves.
Where my conviction stands today
I have strong conviction that high-voltage electrification is a meaningful new market for Sona.
The ₹1 lakh-plus content-per-vehicle assumption can be independently justified. Electrification penetration remains low enough in India to allow years of growth. The product is engineering-intensive rather than heavily capital-intensive. DENSO materially reduces the technology-development risk.
I have moderate conviction on the ₹24,000 crore 2030 TAM.
The content assumption looks reasonable. The volume assumption requires very rapid electrification, and a meaningful part of the market belongs to OEMs with existing captive supplier ecosystems.
I have strong conviction that industrial automation and mobile robotics are structurally growing markets.
Global robot installations have doubled over a decade. India has risen to sixth globally and is forecast to grow faster than other major markets. AMRs are growing much faster than traditional industrial automation.
I have high interest but not yet high conviction in Sona's precision robotic gearboxes and actuators.
The technological fit is unusually strong, and the concentrated incumbent market suggests genuine barriers to entry. But Sona still needs customer qualification to prove it can transfer automotive manufacturing expertise into robotics.
I have low conviction today in using humanoid TAM estimates to value Sona.
When respected forecasts differ between roughly 1.4 million and 10 million annual units by 2035, the forecast itself is not dependable enough to build a valuation around.
And I remain skeptical about complete AMRs and cobots until Sona demonstrates that it can compete not merely in hardware but in software, integration and the surrounding ecosystem.
What I would watch next
For the DENSO business, the most important announcement will not be another TAM estimate. It will be the first major customer nomination.
I would particularly watch whether the JV wins programs outside the traditional Japanese OEM ecosystem. A Tata, Mahindra, Indian commercial-vehicle or export nomination would tell us something important about whether DENSO technology combined with Sona's cost engineering can open markets that DENSO did not previously dominate.
For robotics, I would watch the type of orders rather than simply the size of the order book.
An ₹800 crore order book is a useful starting point. But a production award for a precision robotic gearbox or integrated actuator would change my assessment much more than another ₹500 crore engineering-services order.
That would prove that Sona is moving from being capable of participating in robotics to owning one of the mission-critical pieces inside the robot. And that may ultimately be the more interesting opportunity.
The market tends to focus on the exciting end product. The EV. The humanoid. The autonomous robot.
But Sona Comstar has historically made its money one layer below the visible product.
If it can do the same in the next generation of electric powertrains and intelligent machines, it may not need to build the winning EV or the winning robot. It only needs to make components that the winners cannot easily build without.