This article first appeared in Digital Edge, The Edge Malaysia Weekly on February 10, 2025 - February 16, 2025

When anything challenges the American narrative, the immediate response is one of resistance and scrutiny. The latest flashpoint? DeepSeek — a 20-month-old Chinese artificial intelligence (AI) start-up that trained its R1 model on less than US$6 million (RM27 million) — which has supposedly managed to develop an AI powerful enough to disrupt the global tech industry and send shockwaves through the US stock market, and to a certain extent Malaysia’s AI-linked and data centre construction counters such as YTL Power International Bhd (KL:YTLPOWR), Southern Cable Group Bhd (KL:SCGBHD) and Gamuda Bhd (KL:GAMUDA).

Since OpenAI’s launch of ChatGPT in November 2022 took the internet by storm, AI development has become a high-stakes game requiring staggering investment. US tech giants have poured billions into building and refining their models, making it seem nearly impossible for a competitor to achieve similar breakthroughs at a fraction of the cost. So how has a Chinese firm managed to do just that? The answer is as old as innovation itself: necessity breeds invention.

As Financial Times columnist Angela Zhang put it, China’s efficiency in AI is not an accident. It is a direct response to US-led export restrictions. By limiting China’s access to advanced AI chips, the US has inadvertently spurred its competitor’s innovation. In a way, Washington has itself to blame for pushing China to find a way to work without specialist US tech.

It all started in December when DeepSeek unveiled its “V3” AI model, stating that it had been trained using just 2,048 Nvidia chips. Then, just days before the recent market brouhaha, the company introduced R1— a reasoning model designed to tackle complex problems — positioning it as a direct competitor to OpenAI’s o1 model.

Within days, this supposed state-of-the-art free AI assistant skyrocketed to the top of Apple’s App Store downloads, triggering alarm bells in Silicon Valley. The shockwave was debilitating — investor anxiety wiped nearly US$2 trillion off US tech stocks, the biggest being Nvidia, whose share price plunged 17% on Jan 27, resulting in a market cap loss of nearly US$600 billion.

But here is the thing: if DeepSeek truly managed to build an AI engine at a fraction of the cost, isn’t that precisely the kind of technological leap the industry celebrates — except this time, the wrong country did it? Instead, the immediate response has been one of suspicion, and much of it comes from sources with their own vested interests.

Take OpenAI’s recent accusations that DeepSeek used its proprietary models to train its own open-source large language model (LLM), R1. OpenAI believes that DeepSeek engaged in “distillation” — a method where smaller models learn from larger ones to replicate performance at a lower cost. OpenAI’s terms explicitly forbid using its outputs to train competing models, and White House AI adviser David Sacks has gone so far as to call the evidence of distillation “substantial”. If proven, this could be a serious breach of intellectual property.

Then there is the security angle. DeepSeek has been dogged by concerns over data privacy, user tracking and potential backdoors. The US Department of Commerce is investigating whether the company illegally accessed American semiconductors despite restrictions. Meanwhile, reports of exposed internal databases containing sensitive user data have emerged, fuelling further distrust.

The DeepSeek controversy closely mirrors the 5G battle between the US and China, where Huawei’s dominance in next-generation telecom infrastructure triggered global security fears and economic retaliation. Huawei’s 5G technology was effectively blacklisted by the US and its allies on security grounds. If history is any guide, DeepSeek’s rise could lead to even stricter controls on AI collaboration, semiconductor exports and digital infrastructure — potentially splitting the world into competing AI ecosystems, much like 5G did for telecom networks.

This means DeepSeek’s challenge is not just technical or legal, it is reputational. It comes from the US’ biggest economic competitor, and the Chinese state’s tight control over its corporations only amplifies the fear of government interference. Whether justified or not, companies such as DeepSeek will have to find a way to navigate their reputation for “espionage” if they want to win global trust. As at Feb 5, Australia has banned DeepSeek from all government devices over security concerns, ordering the removal of its products and services from government systems. Home Affairs Minister Tony Burke cited an “unacceptable risk” to national security, though the ban does not extend to private citizens. This follows Australia’s earlier ban on TikTok over security risks.

The move follows similar actions in Italy and Taiwan, as governments worldwide scrutinise DeepSeek.

All of this feeds into a broader reality: the US tech industry — built on its dominance in semiconductors, cloud computing and AI — now finds itself grappling with a competitor that is not playing by the old rules.

Whether DeepSeek turns out to be a breakthrough or a mirage, one thing is certain. AI is no longer just about technology, it is a geopolitical race and the balance of power is shifting faster than anyone expected.

Save by subscribing to us for your print and/or digital copy.

P/S: The Edge is also available on Apple's App Store and Android's Google Play.