Is China Stealing the AI Competition?

The global race for artificial intelligence (AI) has been framed as a US-dominated competition for years. The straightforward assumption was that the United States was far ahead and China was playing catch-up, given the billions of dollars invested by companies such as OpenAI, Google, Meta, and Microsoft. However, that narrative is rapidly evolving. In 2025, a shift is taking place that is quieter but could be more disruptive. Not only are Chinese AI companies competing, but they are also reshaping how AI is developed, shared, and used worldwide. Additionally, they are accomplishing this through “highly capable, open-source, cost-effective models” that numerous US businesses are currently actively utilizing. A Surprising Reality: US Companies Using Chinese AI

The widespread adoption of China’s AI models, even within American tech companies, is one of the clearest indicators of China’s growing influence. The company Pinterest, which is based in San Francisco and has hundreds of millions of users, has made it clear that it has experimented with Chinese AI models to enhance its recommendation and shopping systems. According to CEO Bill Ready, the launch of DeepSeek R‑1 in January 2025 marked a turning point. Unlike most US models, DeepSeek was released as open source, allowing companies to download, customize, and run it internally.

Matt Madrigal, CTO of Pinterest, explained the significance of this: Open-source Chinese models can be custom-trained to meet particular business requirements. They can be 30% more accurate when fine‑tuned in‑house

Their prices are up to 90% lower than those of proprietary AI services. For businesses operating at scale, these numbers are impossible to ignore.

“Fast, inexpensive, and excellent” Pinterest is not alone.

Brian Chesky, CEO of Airbnb, confirmed that his company’s AI-based customer support is heavily powered by Alibaba’s Qwen model. His justification was straightforward: the model is “very good,” “fast,” and “cheap.” Other Chinese models that are gaining popularity worldwide include: Qwen (Alibaba)

Kimi (an AI for Moonshot) DeepSeek is Models created by ByteDance, TikTok’s parent company These models are especially popular among startups and enterprise developers who want flexibility without massive infrastructure costs.

The Advantage of Open Source Take a look at Hugging Face, the largest AI model sharing platform in the world, to gain a better understanding of China’s momentum. It houses models from Alibaba, Meta, Google, and independent labs around the world. According to Hugging Face product lead Jeff Boudier:

Chinese artificial intelligence models frequently occupy four of the top five trending spots. Qwen from Alibaba surpassed Llama from Meta as the model family with the most downloads. Categories pertaining to training effectiveness and customization are dominated by Chinese models. This matters because developers increasingly prefer open ecosystems over closed platforms. Open-source models let you do: Full control over data (critical for privacy and compliance)

Lower long‑term costs

Faster experimentation and innovation

In contrast, many US models remain “proprietary,” requiring businesses to send data through external APIs and pay ongoing usage fees. Meta, OpenAI, and the US Strategy Problem

US AI leaders are not falling behind technically — but they are making different strategic choices.

Meta’s Llama models were once the gold standard for open‑source AI. However, it is said that developers were dissatisfied with the release of “Llama 4,” which led Meta to work with external models, including those from Alibaba, to train its next generation. OpenAI, on the other hand, has primarily focused on: Scaling proprietary models

Protecting huge computing infrastructures

Monetization through subscriptions and advertising

In spite of the fact that OpenAI released two open-source models in 2024, the majority of its innovation is still restricted to paid services. This strategy makes financial sense because investors put a lot of pressure on US businesses to make money, but it can’t be used in markets where costs are important. What the Statistics Say A recent Stanford University AI Index report found that Chinese AI models have:

Matched or exceeded US models in multiple benchmark tasks

Wider acceptance as a result of lower deployment costs * Benefited from strong government and institutional support

According to the report, China’s AI expansion is not just technological but also “systemic,” involving alignment of research, industry, and policy. A Strategic Irony

Former UK deputy prime minister Sir Nick Clegg recently highlighted a striking contradiction.

While the US — the world’s leading democracy — is building closed, profit‑driven AI systems, China — often labeled an autocracy — is releasing models that democratize access to advanced AI.

At the same time, many US firms are chasing the vague goal of “superintelligence,” investing billions into future breakthroughs that may take years to materialize. Chinese labs, by contrast, are solving practical, present‑day problems: affordability, speed, and usability.

So, Is China Winning?

Not outright — at least not yet.

In foundational research, cutting-edge chips, and frontier models, the United States still leads. But China is winning in areas that matter deeply to businesses today:

Open‑source leadership

Cost‑efficient deployment

Rapid real‑world adoption

The AI race is no longer just about who has the smartest model. It is about who puts usable AI into the most hands.

And by that measure, China is not just competing — it is quietly redefining what winning looks like.

—

Bottom line: The future of AI may not be owned by a single country or company. But if current trends continue, open‑source, affordable AI — much of it coming from China — will play a central role in shaping how the world builds and uses intelligence at scale.d