From 0 to 1: Non-Consensus Insights and Future Outlook for Shenzhen Tech Companies (Part 2)
Preface
Shenzhen is one of the clearest examples of what concentrated development looks like. Within a few decades, a former peripheral town became a city that now functions as both an industrial engine and a technological frontier.
This essay continues my notes from a Tsinghua-led research trip across Shenzhen. We visited companies, research institutes, and entrepreneurship platforms, and spoke with people working at different layers of the system. I am less interested in polished narratives than in the strategic logic beneath them.
Part 2 covers Narwal, Huawei, and several closing reflections on startups, cloud infrastructure, and manufacturing.
Narwal
Company Overview
Narwal is a Shenzhen-based robotics company founded in 2016. It focuses on intelligent home cleaning products such as robot vacuums and wet-dry cleaning systems, and has become one of the most important players in the smart cleaning category.
Mission
Lead technological innovation and experiential breakthroughs so every household can enjoy a freer and better intelligent life.
Non-Consensus Insights
Shared by Narwal's software lead and HR team.
1. The anti-tangle brush is more important than it looks
One of Narwal's most interesting product moves is its anti-tangle hair-cleaning solution.
Many competitors approached the same problem by adding more complexity: dual rollers, cutting components, and increasingly elaborate mechanical systems. Narwal's approach was structurally simpler. A one-sided floating conical brush combined with pressure differentials addressed a painfully common user scenario in a cleaner way.
That matters because real category leadership often begins with one concrete user pain point, not an abstract AI story. This feature is not only functional. It also becomes brand shorthand. In the same way that Dyson became associated with cyclone separation, Narwal can become mentally associated with hair handling.
2. AI vision may matter more for future home data than for current cleaning performance
Narwal talks a lot about AI, large models, and binocular vision that can identify hundreds of object types. But the team itself seemed quite sober about where the true value may lie.
For floor cleaning today, the core requirements remain obstacle avoidance, route planning, and cleaning coverage. Distinguishing a chair from a table matters less than navigating space effectively.
The deeper strategic value of advanced visual perception may therefore be elsewhere: long-term spatial data accumulation. If a system can build increasingly rich 3D understanding of the household environment, that data may become part of a broader future home ecosystem rather than merely an immediate improvement in vacuuming efficiency.
That is a more credible story than simply saying “AI makes cleaning better.”
3. Hardware pricing is an asymmetric game
Chinese hardware cost advantage does not automatically translate into easy global dominance. Tariffs and freight eat into the advantage quickly. Narwal's response, as described to us, is modularization.
If the core unit is standardized globally and only certain modules are localized, the company can manage tariff exposure more intelligently. At the same time, consumables such as dust bags and filters may carry higher margins than the hardware itself, producing a familiar razor-and-blades logic.
In other words, the initial machine is only part of the economic model.
4. Premium pricing requires different psychological framing in different regions
The team's regional view on pricing was pragmatic.
In North America, a product can be positioned between established premium competitors and sold through an “AI cleaning butler” narrative that appeals to technology-forward users.
In Southeast Asia, a lighter product configuration at a much lower price point may be a better fit for first-time buyer psychology.
This is a reminder that global hardware strategy is not just logistics plus translation. It is a map of local willingness to pay, symbolism, and household adoption patterns.
5. Celebrity selection is a brand-positioning problem, not just a traffic problem
Narwal's explanation for selecting Yu Shi as spokesperson was more thoughtful than I expected.
The bet was not merely on traffic. It was on symbolic fit. His image can carry a certain “Eastern minimalist technology” brand identity, which helps differentiate Narwal from the harder-edged style of some foreign competitors. It also overlaps with the emotional preferences of a significant portion of the company's consumer base.
This is a useful reminder that consumer hardware branding is not just functional. It is aesthetic, emotional, and cultural.
6. The category is moving from tools to service systems
Their forward-looking view of the robot vacuum industry was also worth noting.
First, these devices may gradually evolve from standalone tools into nodes within a distributed home cleaning system, linked to other devices and services.
Second, the real competition will shift away from simple hardware parameters such as suction power or battery life and toward scenario coverage: edge cleaning, multi-floor handling, and fewer dead zones.
Third, the business model may move from one-time product purchase toward service logic, where consumables, subscriptions, and data services matter more.
Fourth, competition will intensify because robot vacuums are no longer treated as novelty gadgets. They are increasingly perceived as legitimate household necessities.
My Take
Narwal was the most thoughtful host on the entire Shenzhen trip.
That judgment comes from details: the logistics were carefully arranged, the product presentation was polished, and even the small touches in the meeting room reflected an unusual level of care. These things sound trivial until you realize they are not. They reveal how a company thinks about users, presentation, and experience.
One distinction made by Narwal's technical team stayed with me. In 2B markets, growth is often led by channels and relationships. In 2C markets, growth is led by product. That sounds obvious, but it explains a lot.
Many 2B robots, such as hotel delivery machines or lobby greeters, feel functionally narrow and aesthetically rough. Consumer robots, by contrast, are forced to solve real household scenarios while also satisfying higher expectations around design and user experience.
Narwal's two biggest strengths, in my view, are deep user insight and a real willingness to spend effort on specific pain points. When I asked what their biggest advantage was, the answer came immediately: Narwal is the company most willing to pay the cost required to solve real user problems.
That is a serious answer.
Huawei
Company Overview
Huawei is a global ICT company founded in 1987 and headquartered in Shenzhen. Its business spans telecom infrastructure, enterprise networking, consumer devices, cloud, and AI.
Mission
Bring the digital world to every person, home, and organization, and build a fully connected intelligent world.
Non-Consensus Insights
Shared by a director from Huawei Cloud's startup ecosystem team.
1. Cloud infrastructure does not make money by itself. Scenarios do.
One of the clearest points from the Huawei session was that infrastructure alone is not where the real economic value sits. Cloud is foundational, but foundation is not the same as monetization.
Cloud becomes economically meaningful when it is attached to actual business scenarios: industrial workflows, healthcare systems, mining, logistics, overseas deployment, compliance, AI workloads, and so on. In that sense, Huawei's ecosystem strategy is not simply about selling compute. It is about radiating technical capability through vertical use cases that can actually support profit pools.
This is why ecosystem building matters. If you connect enough companies, partners, and use cases, the platform becomes more than a hosting provider. It becomes a coordination layer.
2. Startup support is a low-cost experimentation machine
Huawei Cloud's startup program was framed as an accelerator for trial and error.
Vouchers, free compute resources, global deployment nodes, brand endorsement, and ecosystem connections all reduce the cost of testing an MVP. That makes cloud support functionally similar to a form of non-cash early-stage financing.
Global nodes matter especially for startups going abroad. Infrastructure is not just about speed and uptime. It is also about localization, regulatory confidence, and customer trust in foreign markets.
The brand endorsement piece matters too. For a startup, borrowing credibility from Huawei can materially improve how customers, partners, and investors perceive the company.
3. China and the US are different cloud markets
The Huawei team stressed that Chinese and American cloud markets should not be treated as if they are interchangeable.
US customers tend to be more willing to pay for standardized SaaS products, and the small business environment is often more mature. Chinese enterprise customers care more about visible results, customized solutions, and short-term ROI.
This changes the product strategy. A cloud platform operating in China cannot just copy an American playbook. It has to think more carefully about vertical depth, implementation support, and industry-specific know-how.
4. Big Tech cloud may become the next era's venture capital
The boldest claim from the session was that “big company cloud” may become a new form of venture investing.
Huawei Cloud does not buy startup equity in the traditional VC sense. Instead, it reduces startup cost with compute, tooling, technical adaptation, ecosystem access, and brand support. In weak venture markets, that support can become a decisive variable in whether a company survives.
There is also a filtering mechanism at work. Startups that successfully integrate with the platform, meet technical conditions, and demonstrate product clarity have already passed a certain threshold of seriousness. That makes the cloud ecosystem an upstream screening layer for financial investors.
5. Technical stack lock-in is strategically powerful
Once a company builds important workloads on a cloud platform's differentiated stack, switching becomes expensive. If you train on one chip architecture, rely on one framework, and wire core services into one ecosystem, migration is not a simple hosting move. It can mean rebuilding part of the company's technical foundation.
That switching cost is economically meaningful. Early integration can turn into long-term platform dependence.
My Take
This visit gave me a more concrete understanding of why every large technology company seems to be building cloud.
Before this trip, I had spoken several times with people in startup ecosystem teams at Huawei Cloud, AWS, and Google. One question always bothered me: why are big tech companies all so determined to do cloud?
The answer became much clearer in Shenzhen.
First, cloud is part of a broader role transition. In the digital economy, major internet companies are moving from ad platforms, transaction intermediaries, or value-added service providers toward something closer to infrastructure utilities.
Take a simple modern AI product stack. A web app may be deployed on Vercel, store data in Supabase, and use an orchestration layer such as Dify for agent experimentation. Even a relatively small product quickly becomes dependent on multiple infrastructure vendors. A large cloud platform's ambition is to absorb more and more of that stack into a single environment.
That changes the nature of power. In the future, companies may relate to cloud platforms the way fish relate to water. The platform becomes less like optional software and more like an operating substrate.
Second, cloud is also a capital markets story. Traditional internet businesses such as advertising and e-commerce eventually face growth ceilings. Cloud offers sticky enterprise revenue, more predictable cash flow, and a stronger case for a second growth curve. In downturns, enterprise infrastructure spending is often more resilient than consumer internet monetization.
That makes cloud not just a product category, but a valuation strategy.
Closing Reflections
This was my third trip to Shenzhen and, outside Beijing, the city I have visited most during my years studying on the mainland. But this time I felt something more clearly than before: Shenzhen is not merely prosperous. It is structurally important to China's technological modernization.
Beyond the companies discussed above, we also visited Tsinghua's Shenzhen research institute, Shenzhen Innovation Academy, Shokz, BYD, and others. I cannot cover everything in one piece, but I do want to end with two broader reflections.
How should we judge an early-stage startup?
For a startup before Series A, standardized quantitative metrics are often too weak or too noisy to be decisive. Yet investors, employees, and partners all need to make judgments before the data is mature.
My own framework has two dimensions: people, and problem-solution fit.
1. People
When judging the founding team, I look for three things:
- Passion, passion, and more passion.
- Whether they truly understand the problem they are trying to solve.
- Whether they can execute and actually ship.
In short: strong conviction, real understanding, and the ability to execute.
2. Demand and solution
Here I ask three questions:
- How painful is the pain point?
- How many people share that pain?
- How effectively does the solution relieve it?
In short: solve a problem that a meaningful number of users are genuinely willing to pay to fix.
Why the auto industry still matters
People often say that automobile manufacturing is the jewel in the crown of modern industry. That phrase can sound abstract until you get close to a company like BYD.
The reason is not only the size of the market. It is the density of capability. Car manufacturing requires extremely complex technical integration, supply chain coordination, precision processes, and scale discipline. Those capabilities spill outward.
At BYD, manufacturing capability is transferable. Large-scale die-casting, display systems, precision materials, and coating processes do not stay confined to vehicles. They can extend into consumer electronics, drones, and contract manufacturing for other technology companies.
That is why the automotive industry has such large multiplier effects. A single vehicle contains tens of thousands of components and activates an enormous chain of upstream and downstream industries. In the EV era, batteries, semiconductors, materials, control systems, and intelligent in-car systems all reinforce that effect.
The more I looked at Shenzhen, the clearer this became: technological strength is not just about apps, models, or interfaces. It also depends on industrial systems that can absorb complexity, scale precision, and continuously manufacture the future.
