Western firms test dual AI strategies with Chinese and US models

IMD Business School professor Amit Joshi argues that global companies should consider integrating Chinese AI tools alongside Western models to stay competitive, despite security and regulatory concerns. His research suggests that Chinese AI platforms like DeepSeek have achieved remarkable cost efficiency and industry-specific customization that could benefit Western enterprises willing to navigate the associated risks.

What you should know: China rapidly closed the AI gap after initially lagging behind when ChatGPT launched in 2022.

• DeepSeek emerged in January 2025 as an open-source model that matched or exceeded OpenAI’s performance while using significantly less expensive infrastructure.
• Chinese companies pivoted quickly from traditional machine learning to generative AI, creating vertically integrated systems designed specifically for AI workloads.
• The approach contrasts with Western models that built on existing general-purpose infrastructure originally designed for other applications.

How Chinese AI differs: Chinese companies focused on industry-specific applications rather than building the most powerful general-purpose models.

• Ant Group, which operates China’s Alipay payment platform, created an AI medical app integrated into Alipay, using healthcare-specific data and infrastructure optimized for quick medical inferences.
• This represents a fundamental difference from Western approaches that prioritize building the best general-purpose AI using the highest-quality chips and broadest datasets.
• Chinese models are trained on the Chinese internet in addition to global data, giving them different capabilities and specializations.

Companies already experimenting: Several major Western corporations are testing dual AI ecosystems across different business functions.

• BMW is integrating DeepSeek into vehicles for Asian markets, while Bosch is developing AI-enabled automotive computers based on Chinese AI.
• Procter & Gamble uses Chinese AI alongside Western models for hyper-personalized marketing, partnering with Douyin for interest-based e-commerce.
• Alibaba’s AI helps hundreds of millions of B2B vendors create marketing materials and product descriptions they couldn’t afford to produce manually.

Why this isn’t winner-take-all: Unlike social media platforms that depend on network effects, AI models can coexist because their value doesn’t depend on user count.

• “This is less like social media where we have one dominant social media platform and one chat platform and one search, etc. But this is more akin to mass production where we’ve got about a dozen or more car companies in the world, all of them more or less successful, all of them differentiated in some way,” Joshi explained.
• Companies can use different models for different tasks—Western AI for cutting-edge research and drug discovery, Chinese models for customer service and supply chain optimization.

The security challenge: Western companies must carefully evaluate data privacy, regulatory, and ethical risks when engaging with Chinese AI systems.

• DeepSeek’s open-source nature allows companies to run it on their own servers with reduced cybersecurity risks.
• Future non-open-source Chinese models will require more careful risk assessment regarding data handling and infrastructure dependencies.
• Companies should test Chinese models in sandboxed environments without exposing private or confidential information.

Key limitations to consider: Chinese AI models exhibit clear bias and censorship reflecting their training data.

• When asked about Tiananmen Square, DeepSeek essentially responds that “nothing happened” or refuses to answer controversial political questions.
• Western models have their own biases and content filtering, but the specific limitations differ significantly.
• Companies dealing with politically sensitive topics or requiring unbiased historical information need to account for these constraints.

What executives should do now: Joshi recommends hands-on experimentation to understand the practical differences between AI ecosystems.

• “Go play with DeepSeek right now,” he advises, suggesting executives test it safely in isolated environments to understand its strengths and weaknesses.
• Companies should monitor Chinese business models built around AI, not just the technical capabilities of the models themselves.
• The real competitive advantage may come from innovative applications and business models rather than just superior technology.

The three Cs advantage: Chinese AI development focuses on customization, cost-effectiveness, and real-world calibration.

• Systems are built from the ground up for generative AI rather than adapted from existing infrastructure.
• Models are specifically calibrated for practical applications like healthcare, supply chain, and consumer goods rather than winning benchmarks.
• This approach has produced tools that may be more suitable for specific business functions than general-purpose Western alternatives.