China's AI Ecosystem Strategy: What Swiss SMEs Should Learn Now

While Western discourse celebrates every new model release with benchmark tables, China has changed the question entirely. At the World Artificial Intelligence Conference (WAIC) 2026 in Shanghai, held from 17 to 20 July, the contrast was hard to miss: the conversation barely touched on which language model scored highest. Instead, it centred on how AI actually reaches government, healthcare, manufacturing, finance, education and transport—and delivers measurable effect.

The core insight: competition at system level, not model level

The Tech Buzz China newsletter summarised the shift precisely after WAIC 2026: China increasingly measures AI progress at the level of the system rather than the model. Its competitive advantage is not expected to come from winning every benchmark or releasing the highest-scoring foundation model. This shift is deliberate, not incidental—and the conference's own numbers reflect it.

Of 1,024 exhibitors at WAIC 2026, 242 presented robots and hardware solutions, while only 130 focused on generative AI—of which 83 were AI agents. This distribution is not accidental. It shows where industrial energy is flowing: toward embodied, deployable systems rather than pure language-model demonstrations.

Five strengths, one ecosystem

China ties its AI approach to five concrete strengths that together form an ecosystem rather than an isolated technology. First, embodied intelligence—AI embedded directly into manufacturing ecosystems. Second, massive investment in energy and infrastructure: Shanghai alone runs AI infrastructure projects worth RMB 40.9 billion, with China Unicom adding a further RMB 25 billion. Third, domestic chip development, such as Huawei's Atlas 950 SuperPoD in 1,024- to 8,192-chip configurations, or Sugon's Dawning 8000. Fourth, open models with low-cost deployment. Fifth, state-supported markets that accelerate adoption rather than leaving it to chance.

1. Embodied intelligence—AI integrated directly into physical manufacturing processes
2. Multi-billion-scale energy and infrastructure build-out as the foundation for scaling
3. Domestic chips—proprietary semiconductor architectures for independence and cost efficiency
4. Open models with low deployment costs enabling broad adoption
5. State-supported markets that deliberately drive demand and integration

The 2030 goal is stated clearly: nationwide AI integration across manufacturing and healthcare, as reported by the South China Morning Post in August 2025. Not an arms race for the best model, but a systematic rollout at scale.

What industry voices at WAIC 2026 are saying

The shift away from a pure model race is also articulated by Chinese industry figures themselves. Yang Zaifei of Haizhi Technology put it directly to Yicai Global: the focus is moving away from pursuing ever-larger parameter scales toward the strategic capability to master and steer large language models—and to bridge the gap between basic research and industrial application.

Move away from pursuing larger parameter scales—the focus is the strategic capability to master and steer LLMs, and to bridge the gap between basic research and industrial application.

Wang Yishan of Senad Technology describes a similar logic for the logistics sector: deep integration of large AI models with the physical logistics economy—not as an add-on, but as structural fusion. International voices echo this impression of an open, low-friction innovation environment. Bill Reichert of Pegasus Tech Ventures, discussing AR glasses with real-time translation, said AI is going to flatten the world, describing Shanghai as an environment that builds a space where global talent can connect without friction. Gary Dvorchak of Blueshirt Group put it even more bluntly.

There's no fear of AI here.

The flip side: integration is the real challenge

An ecosystem approach does not mean everything runs smoothly. A survey by the Shenzhen Robotics Association among 119 industry participants paints a more nuanced picture: 88 percent believe embodied AI has genuinely improved. At the same time, 57 percent say current demos are still fairly ordinary. And 75 percent name hardware-software integration—not hardware itself—as the biggest challenge.

This is the real takeaway for Swiss decision-makers: even in an environment with enormous state backing, massive infrastructure budgets and proprietary chip architectures, integration remains the limiting factor—not the technology itself. Anyone who believes a more capable model alone solves operational problems is underestimating exactly this bottleneck.

What Swiss SMEs should take away, concretely

For Swiss SMEs, the lesson is not abstract. It translates into three principles derived from the Chinese approach, even though scale and resources naturally differ.

• Iteration over perfection: a well-integrated, average model embedded in a working process outperforms a perfect model with no operational connection.
• Integration over acquisition: the question is not which language model leads the market, but how it is embedded into existing systems, data flows and accountability structures.
• Adoption over ambition: broad, everyday use across several departments creates more value than a spectacular pilot with no path to scale.

Why this is ultimately a make-or-buy question

China's ecosystem logic also shows why integration is rarely something an organisation manages alone. It requires capacity for data connectivity, process design, change management and ongoing support—resources many Swiss SMEs simply cannot or do not want to build in-house to the necessary depth. This is exactly where the make-or-buy question begins: organisations that cannot perform the integration work themselves should outsource it deliberately, rather than attempting it incompletely in-house.

For Swiss companies seeking to position themselves in global competition without repeating the typical mistakes of Western AI strategies, it is worth looking beyond the domestic market. A Swiss AI podcast episode on strategy in the global AI race examines which strategic mistakes SMEs make most often in international comparison—and how to avoid them.

Conclusion: system beats model

WAIC 2026 did not invent this trend—it confirmed it. The decisive competitive advantage in the AI era does not come from the biggest or smartest model, but from the disciplined embedding of AI into real processes, organisations and value chains. For Swiss SMEs, this is both a relief and a clarification: what matters is not the most expensive tool, but the right integration—step by step, process by process.