00 - The China Question in Robotics

Varad Mohod — Tue, 14 Apr 2026 02:00:58 GMT

Every time a robotics or hardware company comes up, the same question follows.

What happens when the Chinese companies enter?

Let’s start from 0.

Sometimes that question is lazy. Sometimes it is the whole investment case.

I spend a lot of time looking at robotics, industrial automation, and hardware-heavy systems. One pattern keeps repeating. A startup shows a strong demo, real technical progress, and early customer pull. Then someone says the line everyone knows by heart. China will build it cheaper.

Sometimes that person is exactly right.

But the right lens is more specific than that. China is a serious threat when the company is basically selling a box. China is much less fatal when the company is selling a system that becomes part of how work gets done.

That distinction matters more than almost anything else.

Start with first principles

In robotics and hardware, value usually accrues in five places.

Core component cost and manufacturing
System performance and reliability
Integration into the customer workflow
Regulatory, certification, and security clearance
Recurring economics after install

China is strongest in the first category and increasingly strong in the second. Great companies survive by owning the third, fourth, and fifth.

That is the whole game.

If you are underwriting a company in this space, this framework matters because it tells you where margin survives when the category gets harder. If your startup only owns the hardware, you are exposed. If it owns workflow, trust, and recurring economics, the picture changes fast.

Why China is such a serious force

The China threat is not just cheap labor. That framing is shallow and outdated.

China has scale, supplier density, faster iteration loops, manufacturing depth, and a massive domestic market that lets companies learn faster than most Western startups can. According to the International Federation of Robotics, China was by far the world’s largest industrial robot market in 2024. It represented 54 percent of global industrial robot deployments, with about 295,000 units installed. Chinese manufacturers’ domestic share rose to 57 percent in 2024, up from about 28 percent over the past decade. China’s operational robot stock reached 2,027,000 units, the largest in the world.

That is not a niche industrial statistic. That is a picture of scale, learning rate, and supplier depth.

The broader manufacturing base matters too. The World Bank puts China’s manufacturing value added at roughly $4.66 trillion in 2024. That means there is enormous density across tooling, components, subassemblies, logistics, and process know-how.

Policy is aligned with that industrial base. China’s official humanoid robot roadmap set goals to establish a preliminary humanoid innovation system by 2025 and a secure, reliable industrial and supply chain system by 2027. (SCIO) You do not need to romanticize state planning to see what this means. It means sustained pressure, sustained capital, and sustained industrial focus.

This is why the China question keeps coming back. The threat is real. It is just not uniform.

Where China actually wins

China is most dangerous in categories with five traits.

The first is that the product is mostly a bill of materials story. Motors, batteries, frames, controllers, cameras, sensors, actuators, and acceptable software wrapped around them.

The second is that the product sells into a broad horizontal market instead of a deeply embedded workflow.

The third is that deployment is simple enough that customers can compare vendors and switch without much pain.

The fourth is that there is little regulatory friction, little security scrutiny, and little need for trusted local sourcing.

The fifth is that the business depends on one-time hardware revenue rather than a durable revenue layer after install.

This is where Chinese competition gets brutal. Companies do not need to be the absolute best. In many categories, being 80 to 90 percent as good at 40 to 60 percent of the price is enough to crush margins and reset customer expectations.

That is why generic hardware businesses get punished first.

Drones are the cleanest example

The DJI story should be required reading for anyone investing in hardware.

Reuters reported that DJI still held 54 percent of the global commercial drone market in 2021 even after regulatory pressure and share loss. More recently, Reuters reported that DJI still sells more than half of all U.S. commercial drones. That is extraordinary staying power in a market that has had years of geopolitical pressure.

The historical lesson is even more useful than the market share number. Reuters wrote in 2016 that most startups chasing consumer drones had been stung by DJI, which kept slashing prices. 3D Robotics raised more than $125 million and still got wrecked when it tried to sell the Solo drone into a market where DJI had already reset the economics.

That is what commoditization looks like in real life. It is fast, painful, and often fatal for companies that confuse technical quality with market power.

Industrial robots are following a similar pattern

The IFR data is worth sitting with for a minute. China is no longer just the largest buyer of industrial robots. It is increasingly a serious producer as well. Domestic manufacturers now control the majority of the home market. That is a major shift from a decade ago, when imported systems dominated.

That changes the underwriting lens for generic robot makers. If a startup is selling a broad, comparable hardware product into a market with rising Chinese manufacturing strength and weak switching costs, you should assume long-term margin pressure.

EVs and batteries tell the same story

The IEA says that almost half of car sales in China were electric in 2024, and China accounted for almost two-thirds of all electric cars sold globally. Sales in China increased by almost 40 percent year over year in 2024.

This did not happen because one Chinese company built one brilliant product. It happened because supply chain density, scale, policy, and price competition all reinforced each other. Once a category becomes manufacturable at scale with enough standardization, Chinese players become very hard to beat on cost and iteration speed.

Humanoids show the price umbrella resetting early

Humanoids are early, noisy, and full of hype, but the signal is already visible. Unitree lists the G1 at $13,500 and the H2 at $29,900. Whether those exact platforms end up dominating is secondary. The strategic point is obvious. Chinese players are trying to shape price expectations before the category fully matures.

If your thesis depends on premium-priced generic humanoid hardware, you should be nervous.

Where the threat gets overstated

There is another side to this story.

Many robotics markets are not really product markets. They are workflow markets.

That sounds obvious, but investors routinely miss it. They look at the machine, not the operating context around the machine. The customer usually cares far less about the robot in isolation than the founder does. The customer cares about uptime, process fit, labor savings, training burden, service response, compliance, and whether the system works inside their operation.

That changes the buying decision.

A cheaper machine does not automatically win when replacing the vendor means retraining teams, changing software integrations, revalidating a workflow, or increasing downtime risk.

This is where the strongest robotics businesses stop behaving like hardware vendors and start becoming the system of record for a workflow.

That is where real moats get built.

The real moat in robotics rarely sits in the robot

This is the line I wish more people would say plainly.

A robot arm is not a moat. A drone is not a moat. A humanoid body is definitely not a moat.

At best, those are starting points.

The strongest moats in robotics usually sit in one or more of five places.

1. Workflow integration

If the product becomes part of how the customer operates, switching gets painful.

Amazon is a useful example, and the real lesson goes back to Kiva. Amazon did not just buy robots. It bought a wedge into fulfillment workflow and then built a much bigger operating system around it. Today, Amazon says it has deployed its 1 millionth robot across more than 300 facilities worldwide, and its DeepFleet model is designed to improve robot travel time by 10 percent across that network.

The moat is not a standalone robot. The moat is the operating layer around the robots.

Symbotic shows the same principle from another angle. In fiscal 2025, it reported $2.12 billion of systems revenue, $29.6 million of software maintenance and support revenue, and $98.5 million of operation services revenue. It had 50 systems in deployment and 48 operational systems under software maintenance and support contracts. That tells you exactly where the business lives. The robot matters. The business lives in deployment, software, operations, and long-tail service.

2. Installed base plus recurring revenue

This is where the best robotics businesses stop looking like hardware businesses.

Intuitive Surgical is the standard example because it makes the point so clearly. At year-end 2025, Intuitive had an installed base of more than 12,100 systems, including 11,106 da Vinci systems and 995 Ion systems. In the fourth quarter of 2025 alone, instruments and accessories revenue reached $1.66 billion, up 17 percent year over year. (Annual report)

That is the model. The system creates the installed base. The installed base creates recurring instruments, accessories, service, training, and procedure volume. That is why a cheaper copycat does not automatically win in surgical robotics.

3. Ecosystem control

Universal Robots is a good reminder that platforms matter. UR+ is not just a partner badge. It gives customers a prebuilt ecosystem of end effectors, software, sensors, and application modules. That shifts the buying decision away from who makes the cheapest arm and toward who gets the customer to a working application with the least friction.

Cloning the metal is easier than cloning the ecosystem.

4. Service and uptime

This is less glamorous and more important.

ABB markets connected services around up to 25 percent fewer incidents and 60 percent faster response time. The exact number is less important than the principle. Industrial buyers care deeply about uptime. In many environments, downtime costs more than the upfront hardware price difference.

Founders often underestimate this because they love the machine. Customers usually do not. Customers love reliability.

5. Regulation, certification, and trusted sourcing

Security and compliance can redraw the market.

The Defense Innovation Unit’s Blue UAS Cleared List exists because some buyers do not treat all drones as interchangeable. According to DIU, the program has now certified more than 39 systems and 165 components for government users. (DIU transition note) That is not just bureaucracy. It is market structure.

The FAA’s BVLOS rulemaking matters for the same reason. Regulation shapes who can operate, under what conditions, and at what scale. In regulated markets, trust, certification, and domestic sourcing can become part of the moat.

What investors get wrong

The first mistake is treating hardware IP like a complete moat.

Patents matter. Clever engineering matters. Novel designs matter. None of that guarantees durable economics if the category itself compresses into a price war. Patents help at the margin. They rarely solve the commercial problem.

The second mistake is assuming software automatically fixes hardware risk.

Software matters when it improves uptime, deployment speed, operational performance, fleet management, or workflow lock-in. Thin software wrapped around interchangeable hardware is not a moat. It is a presentation layer.

The third mistake is confusing technical brilliance with category quality.

A startup can build an excellent robot and still be building in a structurally ugly market.

The fourth mistake is hand-waving away China as if it is just a procurement talking point. It is not. It is a real competitive force with supplier density, manufacturing scale, domestic demand, and price aggression behind it.

The moat stack I actually care about

When I look at robotics and hardware businesses, this is the hierarchy I use.

First, workflow ownership. The product becomes part of how the customer gets work done.

Second, deployment data that improves outcomes. Not generic AI claims. Real operating data that improves uptime, routing, tuning, failure prediction, throughput, or labor efficiency.

Third, certification, security, and trusted procurement. Defense, healthcare, utilities, critical infrastructure, and regulated categories matter here.

Fourth, installed base plus service network. Field support, spares, maintenance, response times, training, and financing.

Fifth, ecosystem control. Developers, partners, accessories, APIs, and standards compliance.

Sixth, manufacturing advantage. This matters, but I discount it heavily for startups. Very few young companies will out-manufacture China on cost.

Break the market into four buckets

Talking about robotics as one giant category is sloppy. The market breaks down more usefully into four buckets.

1. Commodity-ish hardware

This includes generic drones, robot dogs, standard robot arms, AMRs without deep software, and broad humanoid platforms.

This is where I am most cautious.

These businesses often end up in spec-sheet competition. If the buyer can compare speed, payload, battery life, and price side by side, and if deployment is simple, gross margins usually get squeezed.

What saves you here: domestic procurement barriers, strong channel distribution, unusual application-specific performance, local service, or a niche where buyers care more about trust than lowest price.

2. Integrated enterprise systems

This includes warehouse automation, surgical robotics, industrial cells, grid inspection systems, and specialized construction robotics.

This is where things get much more interesting.

Here the robot is only one piece of the customer’s operating stack. The business can become durable because the company owns deployment, integration, data, and service.

What saves you here: deployment complexity, outcome-based ROI, long implementation cycles, sticky software, and customer pain around ripping out the system.

3. Regulated or sovereign systems

This includes defense drones, public safety systems, utility inspection, critical infrastructure, and some healthcare workflows.

This bucket can be very attractive if procurement timing is survivable. China can be constrained here by compliance, security, or sourcing requirements.

What saves you here: certification, auditability, domestic sourcing, security clearance, and local manufacturing credibility.

4. Picks and shovels

This includes simulation, robotics middleware, safety software, orchestration layers, perception modules, deployment tooling, and other cross-platform infrastructure.

This is often the best risk-adjusted bucket of all. You avoid being the full hardware stack while still capturing growth.

What saves you here: cross-platform adoption, developer trust, multi-vendor compatibility, and software-like economics.

The practical diligence test

When I look at a robotics or hardware startup, I keep coming back to one question.

If a Chinese company built a comparable physical product in 12 to 18 months and sold it for materially less, what would still belong to this startup?

That question forces honesty.

If the answer is not much, you are looking at a company that may live inside a margin trap.

If the answer is workflow ownership, integration, service relationships, installed-base data, recurring revenue, certification, and trusted procurement access, then there may be something durable there.

From there, I drill into a more specific checklist.

Is the product a SKU or a system?

SKU businesses get competed away faster.

What percent of gross margin survives if ASP drops 30 to 40 percent?

If the model breaks immediately, the company is structurally exposed.

What is the real switching cost?

Not the founder fantasy. The real customer pain.

Does switching require retraining?
Does it require revalidation or recertification?
Does it require software rewrites or integration changes?
Does it increase downtime risk?
Does it create spare parts or service risk?

Does the startup own the customer relationship after install?

If the integrator owns it, the moat may sit somewhere else.

Is there recurring revenue?

Software, maintenance, consumables, upgrades, financing, spare parts, monitoring, or workflow subscriptions.

Does performance improve with installed base?

If yes, how exactly?

This is where a lot of AI claims fall apart. I want specifics. Better routing, better failure prediction, faster deployment, lower incident rates, better task completion, fewer service calls.

Is the company selling into a procurement environment where China is constrained?

Defense, healthcare, utilities, and certain government markets matter here.

How localized is the supply chain?

A lot of anti-China positioning falls apart once you look under the hood. Some startups are still deeply dependent on Chinese components.

Does the founding team understand industrialization?

This sounds basic, but it matters. A lot of robotics founders are excellent technologists and weak operators. In China-exposed categories, weak industrialization gets punished fast.

What is actually a threat

The real threat is not just that Chinese companies can copy the hardware.

The real threat is that they can ship comparable hardware quickly, at lower price, with tighter supplier loops, while your company is still debugging field failures and defending a premium price that customers are not obligated to keep paying.

That combination kills startups.

What is not automatically a threat

Chinese competition is much less fatal when the startup is selling on uptime, workflow ROI, compliance, service quality, and operating fit rather than just sticker price.

It matters less when deployment takes months, not days.

It matters less when the system is integrated into ERP, MES, WMS, PLC, BIM, hospital workflow, or utility operations.

It matters less when recurring revenue dominates lifetime value.

It matters less when the buyer needs trusted sourcing, local support, or regulatory clearance.

In those markets, a cheaper machine alone is not enough to win.

My own investor lens

I like vertical systems more than horizontal gadgets.

I like deployment-heavy businesses more than catalog hardware.

I like recurring revenue more than one-time revenue.

I like markets where uptime and trust matter more than upfront price.

I like products that generate operational data that gets more valuable with scale.

I like ecosystems and service layers because they are hard to rip out.

I get nervous when the entire pitch rests on the machine itself.

I get more nervous when the machine sits in a category where customers can compare vendors on spec sheet, price, and delivery time.

And I get very nervous when the founder treats Chinese competition like a stereotype rather than a force that needs to be designed around.

The bottom line

The broad statement that Chinese companies will take over every hardware category is wrong.

The equally lazy statement that strong technology protects a startup is also wrong.

The truth is more useful.

China is a serious threat in markets where the product is standardized, hardware-led, easy to compare, and easy to swap. China is much less decisive in markets where the company owns the workflow around the machine.

That is the real lens.

If you are investing in a robot, be careful.

If you are investing in a system that becomes part of how a customer works, there is a real chance to build something enduring.

In this space, the moat rarely sits in the robot.

It sits in everything the customer would have to rip out with it.

Back to 0.

By Varad - Investor at Zacua Ventures.

start from 0.

rebuild from truth.

back to 0.

double-o

Subscribe now

00