From 0 to 1: Non-Consensus Insights and Future Outlook for Shenzhen Tech Companies (Part 1)
Preface
Shenzhen's history stretches back far beyond the reform era, but its modern identity is inseparable from a much more recent transformation. For a long time it was a peripheral coastal town centered on fishing and agriculture. Then came reform and opening up. In August 1980, Shenzhen was officially designated as a Special Economic Zone, and within a few decades it became one of the most important industrial and technological cities in China.
What makes Shenzhen special is not only the speed of development. It is the density of ambition. Towers rose from former salt flats. Migrants from across the country arrived with the hope of changing their fate. More importantly, the city became one of the strongest launchpads for Chinese technology companies that went from zero to one and, in some cases, from one to changing the world.
This essay draws on a research trip I joined through a Tsinghua innovation and entrepreneurship program. We visited companies and institutions across Shenzhen and spoke with both executives and frontline teams. What I care about here is not standard corporate PR, but the more interesting layer underneath: the non-consensus insights, strategic logic, and product instincts that help explain why some companies become category-defining.
Part 1 covers Tencent, DJI, and Bambu Lab.
Tencent
Company Overview
Tencent is one of the world's leading internet technology companies. Founded in 1998 and headquartered in Shenzhen, it built its early dominance through social platforms and now spans social products, gaming, fintech, cloud infrastructure, advertising, and AI.
Mission
Value for users, tech for good.
Non-Consensus Insights
Shared by Tencent's Hunyuan agent product team.
1. In the LLM era, “effectiveness” needs a new metric system
The first insight was that agent products should not be judged only by old internet metrics.
Traditional products optimize for DAU, MAU, GMV, and scale. The implicit ambition is to build one app used by one hundred million people. But agent products suggest a different north star: problem-solving rate. Does the system actually solve the user's problem in a meaningful way?
This shift matters because some forms of value do not fit neatly into conventional growth dashboards. Emotional companionship is one example. A capable agent can play the role of a patient listener, absorb emotional overflow, and offer communication strategies that users find genuinely helpful. Social value is another. Tencent mentioned cases where AI systems were used in educational simulations around sexual abuse prevention, helping adults teach children how to respond to dangerous situations. The value here is real, but it is not captured by a simple engagement graph.
One concrete example they gave was a baby-feeding assistant that tracks milk intake and compares it against broader infant nutrition patterns to offer better suggestions to new parents. That is not a mass entertainment product. But it could still be an extremely useful one.
2. Agents may unlock a long-tail revolution
The team's second argument was that the LLM era may overturn the winner-take-all logic of the mobile internet era.
The traditional model says that one giant app with one hundred million daily active users is inherently more valuable than a thousand niche tools. The agent model challenges that assumption. One million highly specialized agents, each serving a small but real need, could create more total value than a single super-app.
Their framing was simple: small can be beautiful. A niche agent serving a vulnerable but underserved group may have lower user count and higher real value. Instead of ten thousand shallow features covering sixty percent of everyone's needs, perhaps one million focused agents can solve ninety-five percent of people's needs in depth.
That does not just change product design. It changes the social distribution of technical value. A deeply vertical product can matter more than a mass-market one.
3. Founders may need to stop chasing giant user pools
The third insight was about entrepreneurship. In their view, the next generation of founders should stop dreaming about building the next WeChat or Douyin and instead focus on digging deeper demand wells.
In practice, that means looking for highly specific problems that a smaller number of users are genuinely willing to pay for. Rare disease medication guidance was one example. The logic is not scale-first, but pain-first.
This also changes the competition among large platforms. If the future belongs to many agents rather than a few super-apps, then the real battleground is not app count but infrastructure: developer tools, ecosystem richness, distribution, and deployment. Tencent's own strategic advantage, in their view, lies in linking agent distribution to existing traffic windows such as WeChat official accounts and leveraging its accumulated high-quality content.
My Take
The end state of the internet may not be a handful of giant apps controlling everything. It may be a particulate ecosystem in which countless agents disappear into the fabric of everyday life.
If that happens, the consequences are large.
For incumbents, the real business becomes infrastructure. They stop being only apps and start resembling utilities.
For founders, the game shifts from traffic anxiety to depth of demand. Solving a narrow problem extremely well may become a more rational strategy than chasing vanity scale.
For society, this creates a new path for inclusive technology, because small populations with intense needs become economically addressable.
But this future is still a hypothesis, not a fact. As of early 2025, the real value created by agents remains limited. Excessive optimism is not serious thinking.
Anthropic's piece on building effective agents makes a point the market often avoids: many of the capabilities currently celebrated as “agentic” can also be implemented with traditional workflows or RPA, often with better predictability and stability. Agents add flexibility, reasoning, planning, and natural language interaction. But the real question is whether that flexibility creates value that older systems truly could not deliver.
That is the standard that matters. Not novelty, but net new value.
When more complexity is warranted, workflows offer predictability and consistency for well-defined tasks, whereas agents are the better option when flexibility and model-driven decision-making are needed at scale.
DJI
Company Overview
DJI is the global leader in civilian drones. Founded in 2006 and headquartered in Shenzhen, it dominates the consumer drone market and has built products used in aerial imaging, agriculture, mapping, public safety, and industrial inspection.
Mission
To become a technology company that continuously advances human civilization.
Non-Consensus Insights
Shared by a Tsinghua senior working on drone solutions.
1. DJI's moat is not just technology. It is industrial control.
The most striking point was that DJI's dominance is not reducible to “better engineering” in the abstract. A major part of the moat lies in supply chain integration.
According to the discussion, DJI used industrial chain advantage to drive the cost of drone solutions down dramatically. A solution that once cost around RMB 2 million could be pushed toward RMB 300,000. Even if competitors could reproduce some of the technical know-how, they could not easily reproduce the manufacturing efficiency, supplier coordination, and cost structure.
That creates a brutal barrier: others may be able to build it, but they cannot build it at the same price-performance point.
2. Pricing itself was a weapon
DJI's pricing strategy was described as a form of market-clearing aggression. Before DJI's airport-style drone solutions entered the market, similar products could sell for around RMB 2 million. DJI reportedly entered at roughly RMB 300,000, with distributor pricing even lower.
This was not just discounting. It redefined the market. Traditional integrators that depended on high margins were squeezed out, and the industry was pushed into an era of cost transparency.
The hidden logic is that hardware advantage can subsidize software and ecosystem monetization. In some enterprise inspection cases, hardware is only the initial wedge. Software subscriptions, algorithmic planning tools, accessories, and upgrades contribute a much larger share of lifetime value.
3. Relentless focus beats broad expansion
The senior I spoke with emphasized DJI's internal product philosophy: radical focus.
Frank Wang reportedly admired Steve Jobs and maintained an unusually narrow strategic fixation. DJI did not try to become a general hardware conglomerate. It focused intensely on drones, especially flight stability, control algorithms, and safety. That concentration allowed the company to build overwhelming depth in one domain rather than shallow presence in many.
The same logic showed up in its closed innovation model. Core algorithms are treated as black boxes. Reverse engineering the hardware is not enough if you cannot reproduce the control stack. In a category where stability failures can literally make the product fall out of the sky, the algorithmic moat matters.
4. Some categories win because they stop needing salesmanship
One of the most interesting observations was that DJI's sales teams often do not need to “push” in the usual sense. The product gap itself pulls the market in. Customers come because the technical and cost advantages are too large to ignore.
That kind of position is rare. It usually appears only when a company has compressed multiple advantages into one sentence the customer can instantly understand.
5. Talent design matters as much as product design
DJI also appears to have built a strong funnel for a specific type of technical talent: young, highly capable engineers with deep interest in robotics and strong hands-on instincts. Competitions such as RoboMaster create a better filter than conventional interviews because they reveal both skill and temperament.
The discussion also highlighted a long-term incentive structure and the company's unusual independence from outside capital. If your cash flow is strong enough, you do not have to let investors dictate the roadmap.
6. Why electric control instead of fuel?
There was also a technical discussion on why DJI chose electric systems instead of fuel-based approaches.
Fuel systems are more vulnerable to environmental disturbance and slower in fine-grained response. Lithium battery systems enable faster control feedback, more precise adjustments, and lower mechanical complexity. That makes them better suited to dynamic flight conditions.
The remaining bottlenecks are familiar: battery performance in extreme cold, flight control challenges such as frozen blades, and communication limitations in mountainous or obstructed terrain.
My Take
Before this visit, I knew DJI was important. I did not fully understand why it was great.
What stayed with me most was not a single data point but an overall feeling: restraint. From the employees to the architecture of the headquarters to the industrial design of the products, everything conveyed the same message. The matte gray palette did not feel dull. It felt controlled, confident, and precise.
One design principle mentioned during the visit was subtraction. Avoid ornamental features. Focus on the core need. Simplify the product. Improve stability.
A good product, as one person put it, should have a core advantage that can be summarized in a single sentence.
For DJI drones, that sentence is straightforward: stable control and flight safety.
That is strong product thinking. It is also strong marketing thinking. The simpler the core proposition, the easier it is to own the customer's mind.
DJI went from founding to category leadership in under two decades. If you smash a drone into pieces and sell the components separately, they are not worth much. The value comes from integrating them through knowledge, systems, and algorithmic control. Economically, that is exactly what higher total factor productivity looks like.
$$ Y = A · Kα · Lβ $$
Bambu Lab
Company Overview
Bambu Lab is a Shenzhen-based 3D printing company founded in 2020. It focuses on making desktop 3D printing faster, smarter, and more accessible through a combination of hardware, software, and machine intelligence.
Mission
Bring 3D printing into every household.
Let 3D printing accelerate industry.
Non-Consensus Insights
Shared by the chairman's secretary.
1. The first challenge is balancing technology and cost
Professional-grade 3D printing always runs into a structural tension: pushing performance higher often pushes cost higher as well. Low-end products sacrifice stability and precision. High-end products may be excellent but too niche to scale.
Bambu's answer appears to be staged execution. First, solve machine stability. Then improve material compatibility. Then build an ecosystem around content, users, and creation.
This sequencing matters. A company that tries to solve everything at once usually solves nothing well.
They also emphasized supply chain improvement and in-house work on critical components. That is how you create room for lower prices without turning the product into junk.
2. The second challenge is global regulation and data governance
The international market is not flat. Regulatory regimes in Europe and the US create real friction around data rights, compliance, and national security concerns. That raises operating costs for any hardware company trying to scale globally.
The strategy described to us was pragmatic rather than ideological: comply early, limit exposure to sensitive data, and define product boundaries in ways that reduce regulatory conflict. In some cases, that means restricting certain scanning features. In others, it means balancing hardware openness with software control.
This is not glamorous work, but it is how global hardware companies survive.
3. Growth depends on moving from geeks to ordinary users
Bambu's user segmentation was especially revealing.
The first layer is the enthusiast market: people who love the technology for its own sake, print car parts for fun, and enjoy the act of making. They matter because they validate the product and generate word of mouth. But they are not enough to sustain mass-market growth.
The second layer is early consumers who use 3D printing for semi-practical scenarios such as education, home accessories, and customization.
The third and most important layer is the mainstream household market. That shift requires a different product philosophy. If the product is to become a normal tool, not a geek toy, it has to become safer, easier, and lower-friction. A child should be able to use it. The threshold for design has to fall.
4. The moat is a full-stack product philosophy
The company framed its moat in three layers.
On the hardware side, they use algorithmic compensation to overcome mechanical error and improve dynamic precision.
On the software side, they focus on slicing and compatibility instead of trying to control the entire design workflow. That lowers switching costs for users and makes the product easier to adopt.
On the product side, they follow a subtraction philosophy similar to the one I heard at DJI: do not pile on features just to look sophisticated. Put engineering effort behind the actual printing problem.
The long-term strategy is also clear: grow the installed base through more accessible hardware, then build stickiness through content platforms, model libraries, and community.
My Take
Bambu Lab is one of the most interesting companies I visited.
When I asked for the product's single-sentence core advantage, the answer I got was direct: with our product, you are always using the best product in the industry.
That level of confidence sounds dangerous unless it is backed by real execution. In this case, it seems to be.
Bambu's founder, Tao Ye, previously led DJI's consumer drone unit and spent years working closely with Frank Wang. Even though Bambu entered the market later than many incumbents, its first major product landed like a shockwave. It quickly became one of the most successful Chinese unicorns in the category.
What I found especially interesting is the continuity in product philosophy. At DJI, I heard again and again that great outcomes came from obsessive focus on product. Bambu seems to have inherited that same method. Less noise. Less ornament. More engineering. More product.
That is not only a cultural style. It is a competitive weapon.
Part 2 covers Narwal, Huawei, and a broader reflection on Shenzhen.