What Swiss SMEs Can Learn About AI Resilience From China's Response to US Sanctions

In short
When NVIDIA CEO Jensen Huang declared US export controls a failure in 2025, he spoke an uncomfortable truth: sanctions accelerated China's AI self-sufficiency, and NVIDIA's market share fell from 95 to 50 per cent. Huawei's Chairman thanked the US government in 2026 — forced independence enabled SMIC's 7nm breakthrough. For Swiss SMEs, the strategic lesson is clear: vendor diversification and technology sovereignty are no longer optional but business-critical. Relying on a single provider today means accepting geopolitical risks you cannot control.
Why Jensen Huang's Admission Changes the Rules
In 2025, NVIDIA CEO Jensen Huang broke with the official US line: the export controls designed to restrict China's access to cutting-edge chips had failed. Instead of slowing China's AI industry, they accelerated its independence. NVIDIA's market share in China fell from 95 to 50 per cent within a few years — not because the technology became worse, but because planning certainty disappeared.
This development directly affects Swiss SMEs. When the world's largest chip manufacturer admits that geopolitical intervention halved its market position, the risk model for every technology purchase shifts. The question is no longer just: which solution is technically superior? But also: how dependent does this vendor make me on decisions I cannot influence?
The Market Share Collapse in Numbers
NVIDIA dominated the Chinese AI chip market in 2020 with 95 per cent. After tightened US export controls in 2022 and 2023, that share fell to 50 per cent by 2025. Huawei, SMIC and local vendors filled the gap — not through technological superiority, but through availability and political planning certainty.
Huawei's Thank-You: How Sanctions Force Innovation
In 2026, Huawei's Chairman publicly thanked the US government — for the forced transformation. Sanctions eliminated the option to continue relying on Western suppliers. The result: SMIC achieved a breakthrough in 7nm fabrication, China's semiconductor industry grew faster than planned, and Huawei positioned itself as a symbol of national technology sovereignty.
For Swiss companies, the parallel lies not in chip fabrication but in strategic posture: dependency only becomes visible when it becomes a problem. A 2025 CFR report predicts Huawei will reach only one to five per cent of NVIDIA's AI compute by 2027 — the technological gap remains enormous. But the business lesson is clear: planning certainty trumps technological excellence when access becomes uncertain.
What Vendor Lock-In Means for Swiss SMEs
Vendor lock-in emerges gradually. You start with a platform, build workflows, integrations and competencies on it — and discover that switching becomes prohibitively expensive. In AI, the risk intensifies: models are not portable, APIs are not standardised, and training data sits in proprietary ecosystems.
US sanctions against China show that vendor lock-in is not just a cost issue but a geopolitical risk. If your business model depends on a US platform and that platform suddenly restricts access for regulatory reasons, you lose not only a technology — you lose agency.
- Dependence on a single cloud provider makes you vulnerable to its regulatory risks
- Lack of portability increases switching costs exponentially
- Geopolitical escalation can disrupt supply chains even in neutral markets
- Multi-vendor strategies require more coordination but reduce concentration risk
Technology Sovereignty as a Swiss Principle
Technology sovereignty does not mean autarky. No Swiss SME will build its own foundries or train LLMs from scratch. But you can consciously shape dependencies: through vendor diversification, open-source preference where possible, and infrastructure decisions that enable portability.
The discussion around AI sovereignty between the USA and Europe shows that even neutral actors must think strategically. China's response to US sanctions is the extreme example — but the principle applies to every market: those who lose control over their technology supply chain lose strategic agency.
Practical Steps Toward Vendor Diversification
Do not place all AI workloads on one platform. Use multiple model providers in parallel — including from different jurisdictions. Prefer open standards and APIs that allow migration. Document where lock-in exists and plan exit scenarios for critical dependencies.
The Multi-Vendor Strategy as Risk Hedge
A multi-vendor strategy increases short-term complexity: you need more integration, more governance, more training. But long-term, you buy optionality. If one vendor fails, becomes blocked by regulation, or doubles its prices, you have alternatives.
In the context of Chinese AI models now capturing 46 per cent of the enterprise market, this strategy becomes even more relevant. You do not have to adopt Chinese vendors — but you should know that your competitors might, gaining cost advantages and redundancy.
The lesson from NVIDIA's market share collapse is simple: no vendor is too large to come under pressure from external factors. Your resilience does not come from choosing the best vendor, but from the ability to switch between several.
Geopolitical Risks as a Planning Factor
Geopolitical risks were long a footnote in IT strategies. That is changing. When the US restricts China's chip access, Europe demands its own cloud sovereignty, and India fights digital colonialism, technology becomes a theatre of strategic rivalry.
For Swiss SMEs, this means: neutrality does not protect against collateral damage. If your US cloud provider excludes Chinese customers or a European data protection regime introduces new requirements, you are affected — even if you are not the target of the measure.
50%
NVIDIA market share loss in China due to export controls (2020–2025)
Why Swiss Companies Should Act Now
The good news: you do not have to wait for a crisis to hit. Insights from Swiss AI investments, which exceed the European average, show that local companies are fundamentally willing to invest strategically. The question is whether these investments include resilience.
Vendor diversification costs short-term efficiency. But efficiency without resilience is fragility. China's response to US sanctions shows that forced independence works — but only if you start early and have the resources. As an SME, you have neither Huawei's budget nor China's state support. But you can make decisions today that give you options tomorrow.
Concrete Action Areas for SME Decision-Makers
Start with a dependency audit: where does your business depend on a single vendor? Then prioritise by criticality: which failures would hurt operationally, which would stop the business? Finally, define redundancy thresholds: where is the effort for alternatives worthwhile, where is risk acceptable?
- Identify single points of failure in your AI infrastructure and document their impact
- Test at least one alternative vendor per critical workload — even if you do not use it in production
- Ensure data and models are exportable and not locked in proprietary formats
- Consider geopolitical diversification: one US, one EU and one neutral vendor minimise regulatory concentration risks
The Underestimated Risk: Regulatory Fragmentation
Even without export bans, diverging data protection, AI and security regulations fragment global markets. What is legal on one platform today may be banned in your target market tomorrow. Multi-vendor strategies create flexibility beyond geopolitical crises.
Conclusion: Resilience Is the New Efficiency
Jensen Huang's admission and Huawei's thank-you mark a turning point: technology procurement is no longer just about performance and price, but strategic control. For Swiss SMEs, this does not mean panicking or isolating — but planning consciously.
Vendor diversification, technology sovereignty and resilience are no longer abstract concepts. They are the answer to a world in which geopolitical risks shape AI supply chains — and in which dependence on one vendor means dependence on its government. The lesson from China's response is simple: those who plan early have options. Those who wait can only react.
Frequently asked questions
- Why is vendor diversification suddenly so important for Swiss SMEs?
- Because geopolitical risks such as US export controls show that even dominant vendors can suddenly become inaccessible. NVIDIA's market share loss in China from 95 to 50 per cent is a warning signal: dependence on one vendor transfers its regulatory risks to your business.
- What does technology sovereignty mean concretely for an SME?
- Not autarky, but conscious control over dependencies. You do not need to manufacture your own chips, but you should ensure your data is portable, that you can switch between vendors, and that no single point of failure paralyses your business.
- Has Huawei really closed the technological gap to NVIDIA?
- No. According to a CFR report, Huawei will reach only one to five per cent of NVIDIA's AI compute by 2027. The strategic lesson, however, is not technological parity but that forced independence works and planning certainty is often more important than cutting-edge technology.
- Which vendors should a Swiss SME combine in a multi-vendor strategy?
- Ideally from different jurisdictions: a US vendor (e.g. OpenAI, Anthropic), a European one (e.g. Mistral, Aleph Alpha), and possibly a neutral or Asian vendor. This minimises regulatory concentration risks and increases bargaining power.
- Are the additional costs of vendor diversification even affordable for SMEs?
- Short-term, diversification increases integration and coordination effort. Long-term, it reduces the risk of vendor lock-in, price explosions and regulatory outages. The question is not whether, but how much redundancy you can afford — and how much downtime you cannot afford.
Sources
- NVIDIA CEO Jensen Huang on US export controls (CNBC, 2025)
- Huawei Chairman thanks US government (Bloomberg, 2023)
- SMIC 7nm breakthrough and Huawei Mate 60 Pro (TechRadar, 2026)
- Council on Foreign Relations: US-China Tech War (CFR, 2025)
- Jensen Huang: China will build its own chips (Reuters, 2025)
- Vendor lock-in risks in AI (DW, 2023)
Would you like to explore this topic for your company?
Check Availability