Hardware Investing, Evolved: Hot Sectors and Contrarian Bets — 张涵 x 李立远 x 张凯 x 陈锋
Summary
- This intelligent-hardware revival is not a single AI narrative, but the simultaneous maturation of the supply chain, engineer density, spillover capabilities from autos and robotics, and global consumer demand in 2024–2025. Investors saw talent and resources moving into the sector at CES 2024, valuations rising in the second half of 2024, and consensus forming in 2025; they believe China’s supply-chain and engineering dividends will remain intact for the next 5 years, and that world-class brands are more likely to emerge from consumer electronics.
- High prices are not evidence that new hardware lacks PMF; investors have systematically underestimated consumers’ willingness to pay a premium to solve real pain points. An RTK mower surpassed $2,000 before costs had fallen; the team had modeled a 3-year useful life and labor substitution, while consumers were really buying relief from “not wanting to stand under the sun themselves.” Another product, priced above $2,000 despite limited sensor iteration and little change to the product itself, also sold out, reinforcing the view that “the market always rewards the pioneers.”
- The biggest opportunity for makers is not the existing geek TAM, but AI, electrification and automation putting creative capabilities in the hands of the mass consumer. 张涵 describes 3D printing as a “spatial teleportation toy”: find a model online, press one button, and hours later a virtual object enters the physical world. Automatic-transmission guitars, fishing rods, surfboards, and easier-to-use CNC and laser engravers are not selling feature lists; they are selling the idea that “good products feel like magic,” giving ordinary people skills they did not previously possess.
- AI hardware should work backward from demand and context, rather than treating AI as a mandatory product definition. 陈锋 distinguishes machine learning, which has long been embedded in the operating logic of robot vacuums and drones, from AI-native hardware built around companionship, recording and new forms of interaction; edge compute, battery life and model quality remain limited, and cloud collaboration may improve the experience, but “the need may have always existed”—AI is only one of the technologies that can satisfy it.
- The real prize in AI Phones is not an assistant entry point, but the highest level of system access, traffic allocation and the business model. The Doubao phone and Nubia’s efforts made the 2025 debate explicit: some internet companies want AI to obtain OS-level access, while apps and super-platforms fear that it would “move the foundation of my traffic business”; if AI OS becomes new infrastructure, the fight is closer to “who flips whose table,” and may force platforms to rewrite the logic of advertising, paid traffic and monetization.
- Smart glasses have a high theoretical ceiling, but the startup opportunity begins with trade-offs, not cramming display, imaging, payments and every service into one pair of glasses. Wearables were described as “10x harder” than ordinary battery-powered hardware: fashion, weight, battery life, Bluetooth interruptions, fit yield and software interaction can all become failure points. The conclusion is that “good smart glasses must first be good glasses”; incumbents tend to add features, while startups can win by subtracting around a single point such as display or first-person imaging.
- Early-stage hardware investing ultimately remains a bet on the founder, with the core profile condensed into a person of destiny, an ambitious operator, a hardball opponent and a sufficiently detail-oriented deliverer. A person of destiny has spent years preparing through their experience and major choices to build this company today, and would have done it even without the 2024–2025 boom; 李立远 adds imagination, strategic foresight and being “difficult,” while a full-stack hardware-and-software CEO must also be detailed, because any detail in the first product can break the entire delivery chain.
- Global brands should enter through high-end new categories, then convert creative leadership into supply-chain strength, brand equity and cash within an approximately 18-month window. The line that “Shenzhen entrepreneurs are making electronic luxury goods for high-end consumers worldwide” struck the speakers as too aggressive, but accurately captured the route: define a new, uncontested context, set your own price, then extend downward into mid- and low-end tiers. Ideas cannot remain secret forever, and crowdfunding may expose them 6 months early; what ultimately saves a company is not the idea, but a hedgehog organization that makes the strongest incumbent judge attacking it to be poor value.
Deep dive
1. The 2025 Hardware Boom Actually Started in Late 2023
张涵 looks back to CES 2024: once the pandemic-era disruption to trade shows ended, talent, attention and resources were visibly moving into consumer electronics; valuations began rising in the second half of 2024, and only became broad consensus in 2025. By his estimate, this cycle “has already been burning for about 2 years.”
A hard-tech discussion in late 2023 captured the deeper variables: manufacturing growth was pushing technology downstream, while cost curves, supply chains and mass-production capabilities improved together. Consumer demand did not suddenly appear; demand that could not previously be delivered finally became feasible.
Sequoia began making hardware bets in 2024, expanding from Xiaoguangzi to Even, Jike and HiFuture among others. Around the same time, capabilities accumulated in electric vehicles and other sectors began spilling into new consumer electronics, with the rapid growth of companies like EcoFlow seen as an early signal.
2. China’s Hardware Edge Comes from the Combination of Industrial Depth and Engineer Density
The four speakers share the view that Chinese consumer electronics benefit from two unusually fertile conditions: a complete industrial chain and dense engineering talent spanning R&D, production and manufacturing. Eastern Europe may have some of the talent, but China has a decisive advantage in both scale and industrial depth.
陈锋 recalls that in 2015–2016, it was still difficult to find engineers in China who understood motor control for automotive electronics; customers sometimes needed suppliers to explain power semiconductors and motor drives. Today, those specialists are abundant in the market—“this is no longer an obstacle.”
“Made in China” has also shifted in meaning: once associated mainly with contract manufacturing and low-end production, by 2025 it increasingly stood for quality systems, product definition and high-end demand. DJI, Anker and Insta360 provide brand precedents and proof points for Chinese companies going global.
That is why intelligent hardware generates more startup and investment enthusiasm than traditional white goods. It requires software, hardware, supply chain and brand to work simultaneously, and China currently has the clearest global advantage in that combination.
3. Founders’ Specializations Vary; Full-Chain Experience Is Scarcer
The teams they track come from a wide range of undergraduate backgrounds, including mechanical engineering, electrical and electronic engineering, microelectronics, Computer Science and design. In one small sample, roughly half studied mechanical engineering, but the discussion did not treat that as a fixed founder profile.
Truly strong hardware-and-software founders have often “managed enough different things in the past”: they may have started in software algorithms, moved into product management, then taken responsibility for mass production and manufacturing. Consumer hardware spans supply chain, embedded systems, software, hardware and delivery; a single-point specialty is usually not enough.
陈锋 also points to overseas users willing to try new products and pay a steep premium, allowing startups to validate PMF early. One American grandmother living on government assistance wrote to ask a company to honor a product discount until her benefits arrived; the story made him realize that discretionary consumption does not always follow a simple savings ledger.
4. The $2,000 Mower Shows That Pain-Point Intensity Can Override Price Sensitivity
The early team believed mowing was a large market, but assumed an RTK solution priced above $2,000, with a possible 3-year useful life, had not achieved PMF. They modeled the product against the cost of replacing labor, only to realize that users were primarily buying “not having to stand under the sun themselves.”
The first players to adopt RTK and launch at a high price captured the market upside. The lesson was not merely that the price was accepted, but that demand was much stronger than investors’ original pricing concern: “whoever charges into the market first eats the first crab, and may end up leading the industry.”
Another product sold explosively at more than $2,000 despite limited sensor iteration and no major change to the product itself. Marketing was part of the explanation, but the bigger correction was that consumers’ willingness to pay a premium for their real needs had repeatedly been underestimated.
5. Yard-Care Opportunities Are Shifting from Mowing to Non-Standardized Sensors
The three standard yard categories are usually mowing, pools and snow removal, but they are not at the same stage of competition. Big companies have already entered mowing, lidar costs continue to fall, upstream sensors are relatively standardized, and robot-vacuum players, capital and after-sales systems will all enter the fight.
Even if pioneers make mistakes, several generations of iteration can bring them close to PMF. Once a leading company gets ahead, smaller players face not just product competition but a combined battle across supply chain, capital, channels and after-sales service; the innovation window closes quickly.
Pools may still offer an opening: if underwater sensors can enable genuinely SLAM-capable path planning, they could reopen the category. But if the core solution is OEMed directly by an upstream supplier, a startup can easily be swallowed by a large company that sees the same opportunity.
6. Geeks Are Hard to Serve—and the Best Seed Users for Startups
张凯 retains the “difficult” side of geeks: they have low brand loyalty and will buy whatever is better, cheaper or more efficient. 陈锋 adds that this level of scrutiny exposes the product’s most genuine flaws and supplies use cases and extensions the company itself may never have considered.
“What smart people play with on weekends may become a mass-market consumer product 10 years from now.” A Day One market is usually not yet formed; otherwise there would be no room for a startup. Geeks represent a tiny starting point with a long iteration curve and potentially enormous upside.
Geeks can also tolerate imperfect delivery and see themselves as backers and co-creators, proactively proposing improvement ideas. For high-leverage consumer products, their value goes beyond first-batch sales: they help the product become general-purpose.
Maker businesses also have a rare compound structure: 3D printers, CNC machines, laser engravers and UV printers can sell complex electromechanical hardware, recurring consumables, community distribution and user-network effects. But the market is becoming increasingly obvious, so category choice and go-to-market must price competition in from the start.
7. 3D Printing Is Becoming a “Spatial Teleportation Toy”
张涵 believes leading 3D-printing products have moved beyond the pure-geek market: a mass consumer can choose something they like online, press one button, let the machine work, and open the chamber hours later to find that “the toy has been teleported from the virtual world into reality.”
李立远 believes additive manufacturing is still only in its “first 30%.” Many problems remain unsolved, and new players can still leapfrog with a decisive feature. If this is truly a $100B-scale category, there is no reason it cannot support 3 or 4 listed companies like robot vacuums.
AI is also lowering the front-end design barrier. Users no longer need complex modeling software or theory; they can call on community models or ask generative models to create new ones, then turn them into physical objects. Mass adoption depends not only on cheaper machines, but on shortening the complete chain from idea to object.
The same logic applies to subtractive tools, power tools and DIY workshops. Creation and small-batch production once requiring industrial equipment or purely manual work can now happen quickly; operators of front-shop, back-factory businesses can also expand their product ranges more efficiently.
8. Truly Mass-Market Hardware Sells Capability, Not Features
One investor summarized the principle: “good products feel like magic.” They give ordinary people skills they did not previously possess, or access to services that were once expensive, professional and difficult to learn. Hardware value thus shifts from selling features to expanding the boundaries of human capability.
LiblibAI lets people who cannot play piano get started with strumming and singing in 5 minutes; Xiaoguangzi lets ordinary users without expensive equipment or professional training capture deep-space images. That moment of satisfaction explains why interest-driven products can command a premium.
The speakers especially liked the “automatic transmission” definition: automatic-transmission guitars, fishing rods and surfboards all lower the learning barrier. Ordinary people get to the fun quickly, and some stay on as loyal users; electrification and intelligence expand the audience for the category.
Around 2024, 李立远 summarized AI’s consumer value as companionship, creation and efficiency. In hardware, that means emotional value, turning imagination into reality, and gaining access to capabilities that were previously difficult to experience. “The sci-fi world becoming real” is naturally consumer-friendly.
9. AI Has Already Changed Hardware’s Foundations, but AI-Native Products Still Need More Mature Experiences
陈锋 stresses that without machine learning, many modern hardware categories could not exist: robot vacuums and DJI’s flight and imaging features have long depended on AI to rewrite how the devices operate. AI’s impact on hardware therefore predates the current “AI hardware” label by years.
A second category is AI-native hardware in the narrower sense: products with changed interaction models that provide companionship, computational photography, factual recording or more intelligent responses. 陈锋 currently prefers products with clear pain points and demand, where AI drives iteration in the background, over products that rely on novel interaction alone to create a narrative.
His reservation is explicit: “edge AI may still be incapable of doing many things consumers want today.” He sees this as a matter of time, but different categories will follow different paths: some pioneers can capture the early dividend, while others must wait for the capabilities to mature.
The investment conclusion is therefore not to bet on a static product, but on people who can “stay alive long enough and become strong enough” in imaginative categories. They may change direction, oscillate or hit setbacks along the way, but the founder must be able to carry the load.
10. Demand Comes Before Technology; Context May Rewrite a Sensor’s Mission
Portable and wearable devices are first constrained by battery life: generative AI raises compute requirements, while edge models may not yet be good enough. The speakers see promise in smaller, specialized models and in cloud-edge collaboration that happens without the user noticing.
Imaging is the clearest stress test: information-processing volumes are enormous, and editing, enhancement and model inference are difficult to complete in real time entirely on-device. Edge compute, models and cloud services are all improving, and the long-term target is clear, but a genuinely good user experience will take time.
张凯’s principle is that demand “may not be created” by technology; it may have always existed, with technological progress finally making it satisfiable. AI is an important part of the current technology shift, but not the whole of it. If a context does not need AI, forcing it in may be intuitive but not necessarily logical.
AI’s context chain also changes the role of sensors. Data was previously collected around the human; in the future, multimodal information may first be organized around AI and then directed to the human. Product definition, sensor logic and the sensor’s “mission” could all change as a result.
11. The Core Conflict in AI Phones Is Highest-Level Access and Traffic Redistribution
At the start of the episode, the host mentioned the Doubao phone and called Nubia the most striking hardware story of 2025. At minimum, this suggests the phone market will not be quiet over the next year: incumbents will respond, and phone teams that have not yet found a new home may become scarce talent again.
Few companies possess large models, phone supply chains, hardware-definition capabilities, software experience and OS capabilities at the same time, so alliances are likely. AI OS could bind deeply to existing OS infrastructure, partially decouple from it, or operate phones, tablets and other hardware through a new layer.
The real dispute is how much access AI should receive. Some internet companies want OS-level capabilities, while application platforms will see that as “moving the foundation of my traffic business.” Each side will form alliances according to its position in the stack and try to constrain the power of the other layers.
Existing apps still have deep moats, and it remains unclear whether AI OS calling them will create net positive or negative value. If AI changes the experience deeply enough and moves fast enough, the transition may stop being controlled by any single company and become systemic.
12. Whether New Infrastructure Emerges Will Determine If Incumbents Are Forced to “Flip the Table”
陈锋 breaks the mobile-internet leap into “mobile plus internet”: the terminal infrastructure moved from desktop to phone, and incumbents were initially unable to react, leaving a window for new companies. Most AI tools still run in browsers, on computers and on phones, so the companies with the most resources remain firmly ahead.
The question for AI Phones is therefore whether an AI-OS phone can become new infrastructure, and who is willing to “pick up a rock and drop it on their own foot.” The first company to disrupt itself could be rewarded by the market—or could merely suffer an expensive failure. The speakers say it is impossible to conclude today.
Phone manufacturers may not have the ability to define an OS; internet giants may instead move first. MCP and new hardware could also shorten service chains, allowing users to call ride-hailing, food delivery and other internet services directly without opening apps or clicking through screens.
But super-apps were designed around human operating habits and sit atop deep service ecosystems. Without platform cooperation, the AI-rebuilt experience may not be good. OS players such as Apple and Google have a natural positional advantage, but still face a multi-sided contest involving apps, internet platforms, phone manufacturers and wearables.
13. System-Level AI May Break Through First on Weaker Hardware and a Second Phone
The group is more cautious on short-term forecasts than on the long term. System-level permissions involve too many interests; over the next year, companies may be “eager to try and take small swings,” but a full-scale push is unlikely. Over a 10-year horizon, the probability is considered high.
Weaker smart hardware may connect to broader service ecosystems first. Its own compute is insufficient, so it must rely on phones and internet services; phone makers may not yet see it as an existential threat and therefore may not completely block it.
张涵 is willing to imagine using 2 phones during a transition period: keeping WeChat and other sensitive relationships on the main phone, while giving AI the highest level of access on the other in exchange for efficiency. The product challenge is clear—users want agentic capability, but worry that AI will “take my WeChat and do something weird with it.”
The business model must also be rewritten. If AI means users stop looking at ads and paid-traffic pages, platforms will face a new question: who do they charge? A free AI Phone might buy penetration, but rising compute-chip costs, the manufacturing threshold for phones and the gap between software and hardware organizations all make the free route harder to sustain.
14. Glasses’ Potential Comes from Their Proximity to the Senses; Their Difficulty Comes from Having to Be Worn
张凯 has already seen software ecosystems such as maps move toward glasses, and some ride-hailing and food-delivery use cases could migrate as well. Glasses sit close to the eyes, ears and mouth, allowing them to carry vision, audio, microphones and voice commands at once; combined with rings or EMG bands, they cover key human-machine interfaces.
张涵 gives a brutal difficulty multiplier: battery-powered devices are 10x harder than plugged-in hardware, while wearables are “another 10x harder” than ordinary battery-powered hardware. Mobility, connectivity, charging, weight and human variation all magnify the problems.
Whether women will accept the product cannot be modeled on function alone. Glasses change one’s appearance; users may first assess personal image, brand feel and outfit compatibility, then consider smart capabilities. A product may be daily, weekly or monthly active, and each frequency implies a completely different product standard.
Smart glasses therefore sit between fashion brands and standardized 3C products. Teams trying to serve everyone and cram in every feature are plentiful; what is scarce is a founder willing to make trade-offs and clear on what to do first and next.
15. The Startup Opportunity in Glasses Is to “Make a Good Pair of Glasses First”
陈锋’s summary is simple: “Good smart glasses must first be good glasses.” Once weight, wearability, appearance and basic experience work, the company can penetrate a vertical need for a specific user group; all-in-one may come later, but is unlikely to arrive in one leap.
Even chose to emphasize display, while another company focused on first-person imaging. The latter offers a new information-acquisition perspective compared with conventional cameras; the former deliberately cuts away large amounts of functionality. Both use subtraction to trade for a deliverable experience at an early stage of the technology.
Glasses may be light, but their integration density is extremely high. Display solutions remain expensive, while hardware, software, OS-like interaction, manufacturing refinement and fit yield all need to be solved. A weakness in any one area can stop users from wearing them over the long term.
Incumbents have extensive services and naturally want to add payments and more features. The result may be a surge in software-engineering workload while hardware experience suffers. Startups’ opening lies precisely in restraint and subtraction.
16. Beyond “People of Destiny,” Hardware CEOs Need Ambition, Strategy and Detail
What 张涵 values most is the “person of destiny”: looking back at a founder’s background and major life choices, it feels as if most of their time was spent preparing for the company they are building today. The consistency is extremely strong. Whether or not there had been a boom in 2024 or 2025, regardless of market conditions or competition, they would still do it.
The team has no fixed preference for background: Xiaotiancai veterans, big-tech executives, serial entrepreneurs and fresh graduates can all work; success or failure in the past is not decisive. The first question is fit—whether the founder’s accumulated advantages match the company being built, and whether the investor and founder are directionally aligned.
李立远 prefers “ambitious operators” with strategic vision. Cognitive boundaries are also imagination boundaries; someone who does not want to build something large will struggle to build a large company. Founders must anticipate variables across incumbent competition, startup games and cycles, staying calm in good markets and steady in downturns.
He translates tough guy as “difficult,” and invokes hardball and not easy-going: only a founder who is genuinely hard to deal with can turn a company into a “hard bone to chew” in competitors’ eyes. Full-stack hardware-and-software projects must also be detailed, because any lapse in the delivery of the first product can break the entire chain.
17. 王骁逸 Proved in 1 Hour That Judgment Matters More Than a Perfect First Version
张涵 recalls that 王骁逸 was decisive in their first meeting, skipping pleasantries and going straight to the point like a high-speed machine. He quickly made the core proposition clear: turn Micro LED and waveguides directly into a To C product, while first ensuring that it was a good pair of everyday glasses.
The product was only a demo at the time, and the first version might not have met the ideal standard. But 王骁逸 knew what each stage needed to solve: first build deliverable hardware, then ensure the experience differed from products already on the market. After seeing the prototype, the investor felt that “supporting him was probably the right call.”
He also mapped the competitive field year by year: who the company would fight next year, the year after and the year after that, while sorting through the business-model differences among lens companies, frame companies, brand owners and large incumbents. Not every view was necessarily correct given the information available then, but so far the pacing has been “pretty accurate.”
Overseas-first strategy, channel building and the business model had also been thought through in advance. The most persuasive detail was that he carried his own product throughout the conversation. The investor’s judgment was direct: “If a CEO does not even love his own product,” that is itself a major error.
18. The Most Anticipated Products All Point to Consumer Access to Sci-Fi Capabilities
张涵 has never taken a driver’s license test, but looks forward to one day getting a pilot’s license. Flying motorcycles and the experience of going from point A to point B in 《Dragon Ball》 represent the direct impulse of “turning sci-fi dreams into reality.”
陈锋 likes the vertical-takeoff fixed-wing product formerly known as BlackFly and Opener, and relays that in the United States it can be operated under a relatively low airworthiness threshold below 200 kg. Online footage circling the Golden Gate Bridge and flying with the clouds has led many hardware founders to say they would build it if given a second startup.
Spatial imaging is another frontier of anticipation. HoloLens raised market expectations but underdelivered, while Vision Pro at least demonstrated the possibility of seeing a high-definition spatial world with the naked eye. Retention remains constrained by interaction, content and the long-term usability of near-eye displays.
The longer-term product is a spatial camera that can automatically reconstruct 3D scenes, allowing people to return to a moment in their lives and recover the sounds and visual experience of that world. Beyond that, “we need a Jarvis,” though its ultimate hardware form remains unknown.
19. The Most Convincing Post-Investment Moment Is When a Product Changes a Specific Person
An imaging startup had limited capital but insisted on shooting all of its own marketing materials because it “would not allow its materials to look bad.” The team also spent RMB700K–1M on a design vendor, even if it meant borrowing money, rather than compromise its fundamental standard for beauty.
张涵 connects technical accumulation with aesthetic obsession: an imaging founder who refuses to lower presentation standards at the hardest stage shows that this is not a financing narrative, but something he has “always wanted to make happen.”
In the Hypershell Disabled post-investment group, older people with mobility limitations and veterans recorded daily progress of 0.5, 0.6 and 0.7 miles. One person climbed again, years after hip surgery, the mountain they had once walked; other users liked the post and shared their own new experiences. 李立远 says that in moments like that, “it feels like we helped some people.”
20. Going Global Should Not Be a Special Move; Hardware Companies Can Be Global on Day One
China’s e-commerce infrastructure has long been stronger than overseas markets. Early outbound companies such as Zibao and Aoji first replicated paid-traffic, audience-operations and platform playbooks, then gradually moved from contract manufacturing and products toward brands; content and creator capabilities may follow the same path abroad.
The practical constraints are still tariffs, freight and overseas after-sales service, so the best early export products are those with sufficiently high premiums per unit of volume and weight. Overseas consumers’ willingness to pay for products such as mowers has repeatedly exceeded teams’ expectations.
But overseas markets are not the only answer. China remains enormous, and a clear demand point and value curve can allow a niche audience to support a substantial business. Companies should choose markets according to users, competition and delivery conditions, rather than treating “going global” as a single universal route.
张涵 wants to revise the term itself: it assumes domestic is the default and overseas is the exception. In the future, companies are more likely to be global on Day One, emphasizing whichever markets hold the user value. Chinese teams already have the capability and confidence to produce products that are good, useful, fun, interesting and high quality.
21. Global Brands Should Take the High End First, Then Move Downmarket
张涵 originally wrote: “These Shenzhen entrepreneurs are making electronic luxury goods for high-end consumers worldwide.” He later softened it because it felt too aggressive, but the core judgment remained unchanged: for a brand to take hold, the first move should be high-end.
It is extremely difficult to start with a decade of copycats and low-end supply chains, then suddenly become a premium brand. The current window is different: Chinese companies can define new categories where no competitor exists, set their own prices, and use supply-chain strength and iteration speed to launch differentiated mid- and low-end products later.
There are 2 high-end routes: make an expensive capability cheaper while preserving a premium experience, or define an innovative field first and take pricing power directly. Higher gross margins buy time to build the brand, serve users, absorb competition and fund R&D, while making it easier to bring a high-value brand back to China later.
李立远 searches for categories at the intersections of clothing, food, housing, mobility, sports and imaging. Exoskeletons combine mobility, sports and wearables; electric towable RVs combine housing and mobility. Intelligence and electrification are most likely to create new premium markets in these intersections.
22. A Brand Moat Is Built from Audience, Channels, Taste and Pickiness
张凯 believes small, fast-growing categories are actually well suited to brand building: a company can first capture the pioneers and establish connection and resonance with them. A good brand delivers more than function; it lets users “become the person they want to be” and creates a new persona.
Health products depend especially on trust and early resonance, an area Chinese teams still need to improve. One repeatable path is to put samples in the hands of leading overseas technology or sports KOLs, let pioneer users speak first, then broaden to the mass market. Beyond proving efficacy, the key is making more people aware that the product works.
Technology alone does not define the high end; taste and aesthetics matter just as much. Chinese companies historically excelled at features but were not always strong on beauty and premium feel. Today, startups with only 30 or 50 people are still hiring industrial designers, and both talent density and product finish have improved sharply.
张凯 has also found that people who get brand tone right are often “extremely picky”: they scrutinize every detail from the logo and product exposure to emails and websites, and hold themselves and others to high standards. That difficult, uncompromising standard can be exactly what produces a distinctive brand.
23. Imaging Is a Boundless Category, but Innovation Must Be Sharp as a Needle
The imaging market will not be fixed by today’s TAM; it keeps expanding with each technology cycle. From home video cameras, DV and cameras to drones, action cameras and 360 cameras, every new perspective, focal length and shooting paradigm creates new users.
Imaging’s complexity is both opportunity and barrier. Product definition, form factor, optics, aesthetics, automation, underlying technology and shooting perspective can all be innovated, while aesthetics can never be completely standardized. Startups can therefore continue to carve out new markets.
But “the needle has to be sharp enough.” If an innovation cannot puncture the old experience, an imaging company can quickly run into trouble. DJI remains one of the biggest variables: it moved from drones into handheld products and then toward 360 with Osmo 360, showing that strong companies can cross categories.
The advice is not to attack the mature camera market directly. Traditional camera makers still have deep ecosystems, making existing-market competition difficult and low-ROI. A more rational path is to start with a new context where “no one has what you have,” prove that the company’s other capabilities are also solid, then consider moving into the traditional market.
24. Low Usage Frequency Does Not Mean Low Category Value
The group uses drones as a counterexample: a drone may be used only once a month, yet DJI can still grow rapidly. The key to a hardware purchase is sometimes not daily usage, but the user’s feeling that when a trip or specific moment arrives, they “must have” the capability.
A high-end camera may come out only once a year for travel, but if the user will definitely bring it then, the first purchase is justified. Higher frequency mainly affects wear, replacement and ecosystem development; it does not automatically determine whether the category can work.
Drone replacement demand comes from better lenses, image quality and zoom. Even with infrequent use, consumers may want to buy when a new product arrives. Consumables, accessories, services and collectability can also produce recurring revenue.
There is therefore no universal pattern: some products must be always-on, while others are infrequent but offer powerful emotional value. High frequency does not necessarily create an ecosystem, and an ecosystem is only one type of moat. The real variables are price, replacement desire, reliability and the potential for follow-on purchases such as consumables.
25. Creative Leadership Lasts Only 18 Months; What Remains Is Supply Chain, Brand and Cash
One founder estimates the hardware lead at 18 months: roughly 3 months to validate the concept, another 3 months for the demo, another 3 months to validate mass production, and then time to refine manufacturing and build channels. If secrecy is maintained, competitors also need time to move from disbelief to conviction.
Crowdfunding exposes a product for roughly 6 additional months before shipment. Some companies therefore wait until the product is ready before going live, or avoid crowdfunding altogether; others decline to file patents that would reveal the solution and instead sign 2-year exclusive-supply agreements with key suppliers.
Competition cannot be avoided forever. While the technology remains capable of further iteration, the company can keep widening the gap through R&D and platformization. Once the technology stalls, “the last things that can save you are the supply chain, the brand and the cash in your hands.”
张凯 summarizes incumbent logic as “they do not release the hawk until they see the rabbit.” Startups should grow quickly while the market is still small, but cannot merely become a fat rabbit; they must fill their gaps and grow into a hedgehog “covered in spikes.” The best defense is making the strongest opponent conclude that the return on attacking is too low.