What Surprised Me Most on Our Deep Tech China Trip

Notes from inside an ecosystem defined by constraint, coordination, and speed

When I first began planning this trip for investors — five days across four cities, from Beijing and Shanghai to Hefei and Shenzhen — I expected to come home with company-level observations: what each startup was building, who seemed ahead, which technologies felt real. We met top venture investors, toured EV factories and an eVTOL (flying taxi) test base, walked the halls of the China International Industry Fair and the Shenzhen Logistics Expo, and visited companies working on everything from quantum computing to brain–computer interfaces to humanoid robotics.

Instead, what stood out most wasn’t any single company or technology, but the structure that tied them together. Across cities and sectors, the organizations we saw operated within an ecosystem that has grown more coherent, pragmatic, and infrastructure-driven than I remembered.

Below are the main things that surprised me — organized not by sequence of visits, but by significance.

1. Deep Infrastructure and the Government as Platform

Ten years ago, Chinese innovation policy was broad and aspirational. Local governments talked about “supporting startups,” but few knew what that meant in practice. That has changed. Today, many major municipalities and districts have identified a few strategic verticals and concentrated their resources accordingly—some people call this “deep infrastructure.”

We saw this shift firsthand during our AI Trip, when the World AI Conference (WAIC) in Shanghai—originally launched by Xuhui District—anchored all our meetings in the district’s Caohejing area. Xuhui, one of the wealthiest districts in both Shanghai and China, could easily afford such ambition. What surprised us was how this focused, coordinated approach had spread to other regions and into much deeper tech sectors.

Take Minhang District, for example. Once largely agricultural, it was a quiet suburb when I lived in Shanghai. Today, it’s working to define the future. One of its declared focus areas is brain-computer interfaces (BCI), an initiative it has been pursuing since 2018. The district has physically reorganized its ecosystem around the Shanghai Brain Science Center, clustering multiple BCI companies within the same complex. When we visited one leading firm, we found ourselves in a lobby shared with several others, complete with a large exhibit detailing how the district’s sprawling BCI industrial park would be built out phase by phase.

During that visit, we learned that the district had deliberately co-located a top national neurosurgery department just five minutes away, giving companies direct access to clinicians and patients. Insurance providers will also be involved early, ensuring that reimbursement pathways will exist once the technology reaches the market. The district built shared research platforms offering advanced instrumentation, data analysis, and multi-omics sequencing—collective infrastructure for neuroscience and BCI research that supports collaborative, large-scale innovation. Rather than a loose collection of startups, it’s a coordinated ecosystem engineered with purpose and limits.

The government’s most valuable involvement here is not financial but logistical and relational. District commerce officials act like category managers, introducing startups to large enterprises, coordinating pilot programs, and removing bureaucratic friction that might otherwise slow momentum. One hardware incubator even joked that its biggest competitor is now the local government—not because officials take equity or interfere, but because they’ve become adept at identifying what founders need and delivering it. Sometimes that means access to shared equipment; other times, introductions to manufacturers or customers—the kinds of portfolio-level services top venture investors might offer in Silicon Valley.

When public funding does appear, it usually comes later in a company’s life cycle, once technical and commercial risk has fallen. Contrary to common assumptions in the West, there is virtually no pure equity investment at the seed stage in China; most subsidies target experienced or highly skilled talent, and venture-style speculation is almost nonexistent. When someone in our group asked whether the government invests in early-stage startups, the deep-tech VC hosting us laughed: “The government doesn’t speculate.” Risk-taking isn’t part of its mandate. The system is designed to build infrastructure and enable opportunity, not pursue ownership. Instead of scattering funds across hundreds of companies and hoping one succeeds, districts pick a lane, construct the scaffolding, enact favorable policies, and then let commercial logic take over.

By the way, this topic alone could fill multiple research papers—and probably a book. I won’t go deeper here, but it’s worth clarifying that I’m not necessarily advocating for this kind of heavy, hands-on infrastructure. It isn’t the right approach for every industry. During our AI trip in July, for instance, we visited a government-funded incubator that had effectively outsourced daily management to a local private startup. The result was a more hands-off, market-sensitive model that better fit what AI software founders actually needed.

Other sectors, however, require the opposite. Take eVTOL—flying taxis. In that field, the government must play a central role, because no one else can approve test-flight bases or manage the airspace regulations we saw in cities such as Hefei. Hefei, incidentally, is also building what it claims will be the largest EV industrial park in China and actively courting leading suppliers to set up factories there. It’s a massive coordination exercise, and the government’s role is indispensable.

Weather kept us from taking the eVTOL test flight, so I settled for a photo instead.

So this is not a one-size-fits-all model. Some entrepreneurs shared horror stories about receiving too much attention—spending valuable time hosting delegations or filling out paperwork instead of building. But lately, I’ve heard far more positive accounts. The learning curve has been steep, but local governments are getting better. Districts now compete with one another to attract companies, improve processes, and share best practices. In that sense, it’s starting to look a little like Silicon Valley—VCs competing to build the best platform teams and offer the most effective portfolio support.

2. The Architect of Innovation is … Finance

I come from a finance background, yet only in recent years have I grasped how deeply finance underpins everything. China’s limited financialization—its relative aversion to speculative capital markets—is often praised as a source of stability. But stability has its tradeoffs. Without broad, steadily compounding assets like the S&P 500 that we take for granted in the U.S.—and a capital market that commands over 60% of global investable equities—longer-duration risk-taking becomes structurally harder. Domestic venture capital stays smaller, exits are more constrained, and valuations consistently lower. Those differences ripple through the entire innovation system.

For entrepreneurs, that reality means fewer funding rounds and shorter runways. They can’t assume another raise in twelve months at a sharply higher valuation, so they build differently—prioritizing early revenue, lean cost structures, and quick payback periods. For investors, portfolio construction starts from the exit backward. One venture capitalist told me he spends 20 to 30% of his time thinking about where and how each company can go public—both before investing and long after—to engineer the best possible outcome. He tries to invest in the shrinking overlap where his portfolio companies could tap either domestic or international capital markets. When I mentioned this to another VC friend, he laughed and said he spends closer to 90% of his time doing the same.

Unlike the United States’ highly liquid, disclosure-driven markets, China’s domestic exchanges in Shanghai, Shenzhen, and now even Hong Kong open and close depending on sectoral priorities. Policy determines not only timing but also which industries are in favor. And as geopolitical tensions restrict overseas options, Chinese VCs are relying more heavily on these domestic markets, since many companies that would once have listed in the U.S. or other Western exchanges can no longer do so for regulatory and political reasons on both sides.

Depending on where you sit in the ecosystem, this design can make life either very easy or very hard—and that’s the point: it’s a deliberate choice with real consequences. Entrepreneurs bear the cost of limited financialization through lower valuations and fewer liquidity paths, while investors must structure every deal with a distinct exit scenario already in mind. Few can afford to fund long-horizon research without a clear route to commercialization.

That constraint shapes strategy at every level. One quantum computing startup we met narrowed its focus to a single type of calculation—one that could be performed with photons at room temperature—because building a broader machine capable of tackling multiple problems would have required far more capital than the local market could bear. We saw the same strategy play out in other sectors as well, such as humanoids and BCI.

Yet constraint breeds discipline. When capital is scarce and exits are uncertain, speculative storytelling carries less weight. The system rewards what can be monetized, and that discipline shapes everything—from product design to R&D scope to hiring. It’s one reason Chinese deep-tech companies feel so grounded in near-term practicality. The constraint is financial, not cultural—and it explains far more than any theory of national temperament.

To be clear, this dynamic is separate from China’s state-run research institutes, which focus on ultra-long-term national priorities—space, defense, nuclear energy, and similar sectors. What I’m describing here concerns commercially oriented innovation, largely funded by the private sector, with the government acting as facilitator rather than operator.

As I wrote recently on Twitter, the United States excels at mid-horizon innovation—technologies five to fifteen years out—because its financial markets reward patient capital and tolerate uncertainty. Chinese entrepreneurs, by contrast, must prioritize nearer-term viability. The government fills the other end of the spectrum, underwriting multi-decade “moonshot” programs that fall outside private risk appetite but are strategically important.

This division of labor—private sector near-term, government ultra-long-term, and less support in between—helps explain why Chinese innovation can appear more “boring” while American innovation often seems more “inspiring.” But it’s a bias, not a rule. Not every startup or sector fits neatly into this framework, and it shouldn’t be treated as definitive. Think of it less as a theory and more as a lens—a way to understand the structural incentives shaping two very different innovation systems.

3. Unexpected Transparency

I didn’t expect the level of transparency we encountered. Almost every company had a zhanting (展厅)—an exhibition space. Yes, they carried a touch of vanity, but they also made each business instantly legible and opened space for frank Q&A. A typical zhanting displays working demos, case studies, and a clear timeline of milestones. They exist for a reason: many companies host a steady stream of delegations—customers, district officials, potential partners—and with that many visitors, communication must be systematic.

At one humanoid robotics company, transparency was literal. In one corner, the team had fully deconstructed a robot: every servo motor, actuator, and circuit board laid out on tables, labeled and visible. It was the opposite of secretive. They were also remarkably open about manufacturing costs. When I asked why, the presenter, a senior engineer, looked puzzled. “You could find out very quickly anyway,” he said. “Everyone uses the same supply chain, and pricing is transparent. That’s not our advantage. Our advantage is in what we’ve built ourselves.”

The humanoid robotics company had fully deconstructed its robot for display—this is the arm.

And really, there’s little to lose by being transparent. A battery-tech entrepreneur we met last year explained it best. His startup was at the cutting edge, but he estimated his technological lead at only eighteen months. When I asked if that was due to espionage, he laughed. “No,” he said. “People will just figure it out.” I’ve heard that sentiment often. In this environment, technology alone doesn’t confer a lasting advantage. The edge comes from continuously creating new technology and embedding it in a coherent business strategy, executed with timing and discipline. You don’t have to open-source everything—but the best defense isn’t secrecy. It’s momentum.

4. China’s Hardware Flywheel

The link between capital structure and technological direction is clearest in hardware. Chinese hardware firms benefit from something their software peers often lack: a vast local customer base. That domestic demand means they can reach scale even without global exposure.

At the Shenzhen Logistics Fair, two autonomous vehicle companies illustrated this dynamic. Each had sold over 10,000 driverless logistics vehicles used for warehouse and industrial transport. One, Neolix, said its deliveries were up tenfold from the previous year. Those numbers may not sound extraordinary in software terms, but for industrial hardware they’re significant—and far ahead of comparable U.S. figures. In August alone, the city of Shenzhen, which treats autonomous delivery as a key municipal initiative, recorded almost a million driverless deliveries.

Hardware thrives in China because the pieces line up: the capital to begin manufacturing, the supply chain to scale it, and the domestic market to absorb output. Even mid-tier hotels now routinely use service robots, and new commercial spaces are designed around smart devices. By contrast, Chinese software firms face a structural challenge—low willingness to pay. Both consumers and enterprises expect software to be cheap or free, and many SaaS products struggle to find viable domestic business models.

The implication is straightforward: China’s innovation gravity tilts toward hardware out of economic necessity. The ecosystem supports it—from factories to financing to customers. You’ll see it supported in our upcoming State of Chinese AI Applications report.

5. Borrow, Adapt, Improve

It’s easy to mistake expediency for dependency. At the humanoid robotics company we visited, a few parts carried European-sounding names, but that didn’t mean the firm relied on foreign IP. The reality was far more pragmatic: most Chinese companies start with whatever components work best—sometimes imported—until domestic suppliers spot a proven market and move in. When they do, they learn fast, often overtaking incumbents on cost and quality. It isn’t about catching up; it’s about building better on top of what works.

Few founders worry about where their parts come from, although sure, that might change with geopolitical uncertainties. But the domestic supply chain is dynamic, deep, and readily available. Trying to self-develop every component would slow them down while competitors surge ahead. What matters is speed. Using European hardware or open-source code isn’t a weakness; it’s efficiency. There are genuine gaps in capability, such as in the semiconductor supply chain, but oftentimes it is a choice to move faster and focus resources where they matter most.

Outsiders often miss this nuance. They see imported components and infer dependence. What we saw was the opposite—a system optimized for practicality, using whatever tools deliver the best results now. It’s not about where technology begins, but how quickly it’s absorbed, adapted, and advanced.

6. About Those Chips …

At a company running large-scale AI data centers, we heard something that echoed across many firms: nearly everyone operates with both a foreign and a domestic chip solution. The reason isn’t ideological—it’s economic. A senior AI executive put it plainly: many startups are too cash-strapped to afford foreign compute. “They can’t pay for NVIDIA even if they wanted to,” he said. Sanctions have tightened supply, but pricing alone makes domestic chips a practical necessity.

One autonomous-driving startup, for example, used domestic chips for clients that didn’t need the most advanced functionality but valued lower costs. The AI data-center team described spending months extending the life of older foreign-made chips while also developing heterogeneous clusters of domestic ones—different processors combined to approximate the efficiency of a single, more powerful foreign system.

The week didn’t change my views so much as give them depth. What stood out most was how China’s deep-tech ecosystem is defined by its balance of constraint and acceleration. Financial limits keep ambition realistic, while government infrastructure lowers friction—laying down the tracks so founders can move faster. It’s not a picture of perfect efficiency, but of momentum—a landscape where structure is designed to channel speed instead of slowing it down.

I left with a clearer sense of how this machinery fits together, and great interest in what will come next. We’ll be writing more about these dynamics in future briefings, and for those who prefer to experience it firsthand, we’re already planning 2026 trips focused on EVs, deep tech, and AI (and not just for investors but for young people as well). The best way to understand this ecosystem is still to walk through it—to tour an EV factory, stand beside a humanoid and shake its hand, or take a ride in a flying taxi—and see one version at least, of how the future is being built.