Vol.169 Industry Watch 30 | Challenges and Opportunities in the “Globalization” of Tech Consumer Brands
Summary
The next wave of globalization for tech consumer brands will no longer be driven by low prices or value for money, but by products that are “better, newer, more tech-enabled, and more differentiated,” created jointly by China’s supply chain and product talent. 李丰 used Insta360 as the example: after 7-8 years, Insta360 surpassed GoPro to become the world’s highest-share action-camera brand, and may already have received CSRC approval for a STAR Market listing. 牛浩田 put it more directly: “It’s not that we set out to globalize”; once an innovative product works, it naturally sells worldwide.
Tariffs will create short-term turbulence, but are unlikely to replace China’s manufacturing system wholesale with US manufacturing; Southeast Asia is more likely to become a “tier-two supply chain” for Chinese manufacturing. 牛浩田 argues that manufacturing competition is not about final assembly, but system capabilities spanning precision metalworking, powder metallurgy, and materials R&D. Industries such as footwear and apparel had already moved into Southeast Asia before the trade war; shifting production abroad has long been a way to hedge trade risk. Because COROS has businesses across the US, Europe, China, and other parts of Asia, it believes the impact is manageable over the next 12 months. Companies with 90% of revenue in the US would face much greater pressure.
A startup wins not by having more resources, but by avoiding the battlegrounds occupied by large companies and finding an ultra-narrow, irreplaceable use case. COROS once walked away from ODM work with annual revenue approaching RMB600M, and later exited the wristband market, where it was competing head-on with Xiaomi and Huawei. Its first sports watch was not best-in-class across the board, but its GPS tracked continuously for more than 28 hours, versus less than 20 hours for most rivals at the time, winning over Ultra Run and long-distance trail users. “Are you fighting as a special-forces unit or as an army group?” determines whether the same market size is viable for different organizations.
A company’s true boundary is organizational capability, not the resources on its books; in a crisis, what it can draw on is the trust accumulated in ordinary times. From late 2018 through 2019, COROS was nearly out of cash. It cut headcount by 30%, asked employees to accept voluntary pay cuts compensated with stock, and promised suppliers they would not lose a penny even if the company failed. The pay cuts were later repaid, while the stock was still given to employees. 牛浩田 describes a healthy state as having 10/10 capability while consistently doing 9/10 work with distinctive value; the greatest danger is using newly acquired resources to attempt “12/10 or even 15/10 work.”
The long-term opportunity in sports technology is to progressively democratize the scientific training once affordable only to professional athletes. Sensors can replace some laboratory equipment; cloud computing and AI may replace part of the expert-analysis layer; research findings can then be distilled into algorithms, creating a closed loop of data collection, decision-making, training, and feedback. Consumer demand is also shifting from material satisfaction toward more durable emotional fulfillment through sports and the outdoors. Opportunities remain highly local: Japan’s excess golf courses built during its bubble years pushed practice costs down to roughly RMB50 an hour, creating a mass-market model that China cannot easily replicate.
AI hardware is still in a phase where promises exceed delivery; the more credible long-term themes are persistent memory, proactive service, tool-like utility, and non-humanoid embodied intelligence. 陈芃 is “relatively disappointed” by most products. Bottlenecks include SoCs, power consumption, and model stability; recording products are among the few directions that have clearly improved the user experience. He is looking for a “bridge” between persistent memory and proactive capabilities. 贾凡 believes humanoids may not be the best short-term form factor, but embodied intelligence could enter concrete settings in 3-5 years through other forms. Emotional companionship should first target clearly defined groups such as elderly people who have lost their only child or children with autism, rather than users in general.
“Every consumer product can be rebuilt with AI” is currently closer to a false proposition, and AI democratization will not automatically create a moat. The guests agree that AI must serve product strength: understanding the use case, controlling the cost of continued use, and calibrating expectations to what can actually be delivered matter more than whether a company trains its own model. The result of “deifying it today” could be easier sales, faster returns, and brand damage. AI will lower the barriers to translation, imaging, operations, and other work, but it will also bring in more competitors—“the bar gets higher, not lower.”
Deep dive
1. Tech consumer brands’ overseas expansion is shifting from price arbitrage to an innovation premium
李丰 divides the overseas expansion of Chinese consumer goods into several successive stages: first cheap products, then apparel, small commodities, and value-for-money products driven by platforms. The current stage is about exporting mid- to high-value-added consumer products with technological innovation. “Better, newer, more tech-enabled, more differentiated” is where this round of brand value comes from.
Insta360 is the clearest example. 李丰 said that after 7-8 years of development, Insta360 had overtaken incumbent leader GoPro to become the world’s highest-share action-camera brand; at the time of recording, it may already have received CSRC approval and planned to list on the STAR Market.
Over the past 2-3 years, Frees Fund has invested in roughly 5-6 tech consumer companies across different categories. The event brought together COROS and an AI consumer roundtable around one central question: not whether brands should go overseas, but how innovative brands already operating globally can continue building moats as the trade regime is reshaped.
2. Tariffs can redirect production, but cannot easily recreate a US manufacturing system
牛浩田 starts with the long-term end state: a clear loser between China and the US would not be “good for everyone,” because the loser would drag the other side down as well. Industry will ultimately return to a structure that allows all sides to benefit sustainably.
He argues that the US cannot simply replace Chinese manufacturing. The issue is not worker capability, but willingness to perform assembly-line work, social structure, and the completeness of the industrial chain. Phones can be assembled in Vietnam, but precision metalworking, powder metallurgy, and materials R&D have become system capabilities in China that are difficult to replicate.
Vietnam, Malaysia, and eventually India are more likely to become “tier-two supply chains for Chinese manufacturing.” Labor-intensive industries such as footwear and apparel had already relocated before the trade war; better energy and other infrastructure have reinforced their advantages. Dispersing production across countries has long been a way for global brands to hedge trade risk.
3. Revenue diversification and organizational restraint determine whether a company can ride out the shock
COROS made some operational preparations last year. Because it has meaningful business across the US, Europe, China, Southeast Asia, and other parts of Asia, damage in any single region would not immediately threaten its survival. 牛浩田 therefore sees tariff turbulence over the next 12 months as a short-term problem the company can get through.
He acknowledges that conditions would be much harder for companies with 90% of their business concentrated in the US. But as long as a product is genuinely recognized by the market and the team continues solving problems, “what cannot be replaced is the company”; everything else is more a fluctuation over time. Even a short-term cash crunch may be understood by investors who believe in the company’s long-term value.
A crisis can even force a company to refocus. 牛浩田’s standard is to have 10/10 capability while consistently doing 9/10 work that is achievable and distinctive; if abundant financing leads a company to attempt “12/10 or even 15/10 work,” resources can push the organization beyond its real boundary.
4. The central mission of sports technology is to democratize professional training layer by layer
Traditional sports science first served professional athletes and Olympic teams: people and specialized equipment collected the data, experts built models and training decisions, and athletes executed them before entering the next feedback cycle. The loop works, but its personnel, equipment, and usage costs are high.
Sensors mean basic behavioral and physiological data no longer depend on specialized medical equipment. Cloud computing and AI may replace part of the expert-analysis layer, while research institutions continue producing new models that can be converted into reusable algorithms. “It gradually develops from high-end professional users toward the broader population.”
牛浩田’s mission for COROS remains tiered service: high-need professional users who can bear higher costs will always exist, but the company must also convert professional content into products that ordinary people can use at broadly affordable prices, rather than merely selling technical specifications.
5. Globalization is not a strategic slogan; it is the natural result of an innovation product working
牛浩田 uses Sony’s Walkman to explain globalization: “It’s not that we set out to globalize.” Shrinking a recorder enough to carry around and run on 2 batteries was itself a major innovation. Once a product solved a common global need, international markets followed.
China can now produce new products such as pool-cleaning robots because upstream product talent and downstream industrial capacity have matured together. The ability to create new products has reached, and in some cases surpassed, the global norm. Global sales are therefore not an additional choice, but the inevitable outward expression of product capability.
COROS followed a different path because of the industry’s particular context. China’s outdoor-sports market was then a follower: top trail runners competed overseas, while marathon culture was learning from Europe and the US. Entering the US and Europe in 2016-2017 was not about succeeding abroad before succeeding at home, but about believing that advanced markets produce advanced brands.
6. A company’s taste matters more than copying benchmarks, and slower iteration can improve efficiency
COROS did not treat Garmin and other incumbents as answers to copy directly. 牛浩田 believes imitating the surface of someone else’s success easily becomes “Dongshi imitates Xishi”; the company must know who it wants to become. Distinctive taste is itself part of its value in the market.
The team studies Apple when discussing products and Nintendo when discussing business, especially Nintendo’s “lateral thinking with withered technology” principle. As a result, COROS hardware can take more than 2 years to iterate, giving the company more time to validate reliability and think through the product while avoiding forced launches dictated by market tempo.
Between hardware updates, software and application innovation continue to add value. “We launch new products for the sake of the product.” The test is not whether the technology is new, but whether users have a better solution to a real problem.
牛浩田 also cites Murata Manufacturing’s barometers: an excellent B2B supplier does not merely deliver components, but proactively studies applications the customer has not yet considered. That support is what makes Apple unable to do without it. The lesson is not a single technology, but the operating mechanism behind a good company.
7. The hard part of “listening” is resisting the lure of a bigger market
牛浩田 admits that COROS products “actually have a lot of flaws.” The company is not made up of geniuses; the one thing it may be able to do well is maintain a student mindset and remain willing to listen. But the more user voices it hears, the harder the trade-offs become. The real capability is knowing whom to serve.
Focus means not switching direction simply because another market is more profitable, has more users, or makes more noise. It means recognizing that the company can initially do only the smallest thing in front of it well. The team needs to use the product and face its problems alongside core users, finding new solutions through interaction rather than rushing to extract standard answers from interviews.
COROS has reflected that it once overemphasized innovation while neglecting details, leading the company to repeatedly emphasize that “haste makes waste.” There is a huge gap between a good idea and a good product: the real difficulty is thinking through every step after different users take the product home, not coming up with a clever concept.
8. Macro trends set direction; local events determine how quickly a category takes off
牛浩田 describes consumption upgrading as moving first from material to emotional needs. When incomes rise quickly, luxury goods and high-end dining deliver the first wave of direct happiness; later, sports, the outdoors, and other hobbies grow because they provide more durable emotional anchors. The pattern can be observed in Europe and the US, Japan, and parts of Southeast Asia.
Macro trends cannot replace regional analysis. The rise of marathon culture in China was closely tied to the pandemic experience, and China’s pandemic path differed from that of other countries. Its sports population, participation patterns, and category demand will therefore follow different trajectories.
Japan’s mass-market golf is a more specific counterexample. Too many courses were built during the bubble economy; construction costs were high, but operating costs were not. Oversupply eventually pushed practice prices down to roughly RMB50 an hour, making participation possible even for ordinary wage earners and cooks. “This is something that would not happen in China.”
9. Walking away from nearly RMB600M in annual ODM revenue was COROS’s first real strategic pivot
Starting in 2012, the team provided ODM services for early-stage overseas smart-wearable companies, with capabilities spanning hardware, embedded development, and mass production. By 2014, white-label business was becoming increasingly difficult. 牛浩田 concluded that ODM no longer fit China’s stage of development and that the company would be eliminated if it did not transform.
The company’s annual revenue was then close to RMB600M, yet COROS stopped accepting all new orders and shifted its entire effort to the brand business. “Once you have thought it through, stop doing it.” Its first own brand, Weloop, continued providing online service for years after being discontinued, until it was finally shut down in what the program calls “last year.”
The company chose wearables because behavioral and physiological data could, for the first time, be collected by electronic products. Healthcare, insurance, and corporate health checks could all lower costs and improve outcomes as a result. 牛浩田 favors “long and big” trends: the cycle must be long enough for an ecosystem to generate the small units that large companies cannot complete all at once.
10. Reaching No.3 in China’s e-commerce market proved that mass-market wearables had no room for a startup
After launching a product during the 2016 “618” shopping festival, Weloop quickly reached No.3 in China’s e-commerce market at the time, behind only Huawei and Xiaomi. Many wearable companies created during the 2012-2013 capital boom had already collapsed, and COROS briefly believed it was close to profitability.
“Double 11” delivered the opposite answer: the cost of keyword advertising alone tripled, while product margins were razor-thin. A higher ranking did not create a healthy business; the problem lay in the chosen battleground and commercial structure.
Xiaomi’s band could serve as a traffic funnel, while Huawei’s wearables supported a defensive strategy to protect its handset leadership. Neither company required wearables to make money independently at the time. A wearable-only startup with no fundamental product differentiation could not win under the same rules.
The transformation was therefore not simply a shift from bands to watches, but a move away from “things every large company does” toward a small market that large-company organizations cannot serve. 牛浩田’s analogy is “special forces” versus “an army group”: a bestselling Garmin model might sell so little inside a phone company that the project would be killed outright.
11. The 2018-2019 cash crisis tested years of credit, not last-minute rhetoric
牛浩田 emphasizes that a startup must first choose the right direction, but even with the right direction, success is “essentially a matter of luck.” From late 2018 through 2019, feedback on a new product was acceptable while cash flow was nearly exhausted, and there could still be a long distance between positive word of mouth and sufficient sales and profitability.
The company cut headcount by 30%, leaving those who stayed to shoulder more work. It also invited employees to accept voluntary pay cuts, recorded the difference, and compensated them with company stock. After the crisis passed, COROS repaid the money and did not take back the stock: “The stock was simply given to everyone.”
COROS promised suppliers that even if it went bankrupt, they would not lose a penny, and it set aside cash for payables. Longtime shareholders from the ODM business continued to support the company firmly. The company had no institutional investors at the time, which 牛浩田 believes may have made things easier than for many other founders.
牛浩田 believes solidarity in a crisis cannot be manufactured at the last minute. Not making unreliable promises to employees, not taking advantage of suppliers, honoring commitments to shareholders, and saying the same thing to outsiders, insiders, senior leaders, and frontline staff all increase the “chance of having luck” when it matters.
12. More than 28 hours of GPS battery life turned an imperfect watch into an irreplaceable product
COROS’s first sports watch was designed to meet the running, cycling, swimming, and other needs associated with triathlon. Compared feature by feature with the best products on the market, it was “slightly behind” across the board and still had obvious shortcomings. Its one exceptional strength was low power consumption: under the same system-performance requirements, it achieved the lowest power draw in the world at the time.
The watch could track GPS continuously for more than 28 hours. Twenty-four hours was “an absolute guarantee,” while 28 hours was “highly likely to be achievable”; most competing products lasted less than 20 hours. The difference was not critical for ordinary users, but it solved exactly the problem faced by Ultra Run and long-distance trail runners who had to carry power banks.
After customer service repeatedly received questions about whether the watch could be used while charging, the team identified the people who needed the product most and concentrated all marketing resources on them. Efficiency improved immediately. 牛浩田 later required every new product to identify this hook during definition, rather than deciding the selling point only at the marketing stage.
Company growth should obey both capability and the market. Build capability when it is lacking, but once capability matures, do not force the company to move faster than the market. Entrepreneurship is not about making enough money quickly and stopping work; it is about “finding yourself a happier job.” Even if the work remains difficult over the long term, the team should gain growth and an experience it will not regret.
13. The reality of AI consumer hardware is early and constrained, but recording products have shown value
陈芃, an investor at Ant Group, is “relatively disappointed” by most AI hardware. Many products have little to do with the current wave of foundation models; SoCs are not yet sufficient to process tasks reliably, and power consumption limits continuous use. Recording products are one of the clearer positive examples: combined with large language models, they have rapidly improved user convenience.
贾凡, co-founder of Yuanluobo, divides the opportunity into 2 categories: adding AI to mature products such as story machines, and using AI and robotic arms to create new categories such as chess-playing robots. But AI still has clear boundaries. In many settings it merely produces a predetermined output; if the product definition crosses that boundary, the final experience will differ completely from expectations.
芦胜波 sees 3 layers of opportunity in pet products: personalized and scientific feeding; helping owners better understand their pets and upgrade emotional interaction; and using AI to improve user research, product development, organizational management, and operations. The third layer may be the first to produce measurable enterprise efficiency.
肖培庆, an architect at Amazon Web Services, believes Transformer and generative AI make Her-style free interaction and Jarvis-style assistants directions worth exploring. Models will absorb part of the startup space around Prompt Engineering, RAG, and similar tools, but 2 things are certain: language models’ reasoning capabilities will continue to improve, and costs will continue to fall.
14. Persistent memory and proactive service are the main lines; humanoids and universal companionship may not be the answer
陈芃 is focused on 2 unresolved capabilities. Persistent memory could lead to personalized AI, while proactive service will determine whether wearables can develop into long-term functionality. Many proactive reminders today can be handled with traditional rule-based AI; the real opportunity is to find the “bridge” between memory, proactivity, and personalization.
贾凡 uses the Beijing humanoid-robot marathon to illustrate the mismatch in expectations. In the short term, humanoids may not be the best vehicle for embodied intelligence, and capital may previously have moved too fast. But the public may also undervalue the long-term potential after seeing one event full of problems. He expects many applications to land over the next 3-5 years, though the form may not be human.
Emotional companionship does not emerge automatically from large models. 贾凡 recalls that chatbots from more than 10 years ago were often abandoned after a week; the smart speakers that survived were the ones that could reliably play music. Companionship is “different for every person,” but products aimed at specific groups such as elderly people who have lost their only child or children with autism, combined with multimodal understanding, still have substantial potential.
肖培庆 has tried AI Pin and Rabbit. Rabbit is well designed but “optional”: it has created neither a sense of companionship nor a must-have need, making it look more like a transitional form. What is being underestimated instead is internal productivity: a recruiting company may not sell AI software, but use it to lower delivery costs; AI comics and videos can also be sold directly as outputs.
15. AI adoption should be designed backward from frequent pain points, hardware constraints, and verifiable data
芦胜波 breaks pet ownership problems into “smelly, dirty, tiring, annoying, and expensive.” He starts with “smelly,” something that happens every day and almost everyone hates, then asks whether AI can solve it more intelligently and thoroughly. He also warns that expectations must be managed: “Selling becomes easier, but returns become simpler too,” while the cost and brand damage remain with the company.
陈芃’s approach to AI glasses is to first map use cases in the phone and physical world, determine where large models can outperform existing services, and then run model experiments. He brings hardware constraints such as SoC, power consumption, and heat dissipation back into the decision, ultimately selecting only 1 or 2 deliverable functions rather than continuing to pile on features.
Even though sales of the second-generation Meta Ray-Ban are strong, 陈芃 continues to ask whether the current form factor can truly carry Meta’s future models and services. Sales data cannot replace product definition. Teams must keep using data to understand users rather than relying on imagination to confirm demand in an industry’s early stage.
贾凡 recommends prioritizing tool utility while paying attention to category whitespace in consumer robots. Yuanluobo’s chess robot previously had no comparable product, giving the team time to understand users and refine the product at a slower pace. Global trade shows and media also want to cover novel categories, leaving room for early adopters to pay for novelty and for companies to keep iterating.
16. AI will not rewrite commercial fundamentals; it will widen the gap in product and organizational capability
On the claim that every consumer product can be rebuilt with AI, 陈芃 and 贾凡 both lean toward calling it a false concept for now. The first questions should be whether the product has a foundation for intelligence and what intelligence would actually improve. Every solution deserves reconsideration as technology advances, but “AI is not necessarily required”; users will not pay indefinitely for delays, intermittent delivery, or gimmicks.
肖培庆 remains optimistic, citing how people before trains could imagine only faster carriages and how early smartphones later replaced many computer functions. Humanoid robots completing impressive movements in specific environments show that hardware already has some capabilities. Algorithms’ inability to adapt to complex environments does not mean the underlying route has failed.
芦胜波 is also relatively optimistic, while emphasizing that not everything needs AI. He believes AI will create many new categories, such as AI voice recorders. The industry is still exploring which products are easiest to launch, which users will pay for, and which can generate attractive margins.
So-called technology democratization is still far from complete. 陈芃 points out that models are not stable enough even during free trials, so paid products face a higher bar. Continued use creates continued costs, while subscription prices may not cover a sufficiently broad market. “Integrating AI on its own often produces very poor results right now.”
The guests ultimately return the moat to traditional factors: use-case understanding, product experience, channels, brand, organization, talent, and efficiency. Brands need to understand AI’s boundaries and future evolution while focusing on specific users. Investors should use these tools more themselves. Hardware founders can start with low-cost cloud APIs to test and iterate. AI lowers the barrier to entry, but also makes competition “more intense.”