DeepSeek, The New AI Cold War, And Its Implications For ASEAN

From the US perspective, the DeepSeek phenomenon has set strategic alarm bells ringing. While it weakens US sanctions aimed at restricting China’s AI advancement, it emboldens emerging economies (this includes economies in the ASEAN region) to bypass US-based AI monopolies.

A relatively unknown Chinese company recently unveiled DeepSeek, an open-source and cost-effective AI large language model (LLM). The release has signalled a significant shift in the global AI landscape as it has directly challenged the entrenched dominance of US-based AI giants such as OpenAI (GPT), Meta (Llama) and Google (Gemini).

Developing a unique AI model is costly, especially if model is training from scratch. The costs depend on various factors such as infrastructure, data collection, training duration, fine-tuning, and deployment. DeepSeek has also demonstrated successfully that with proper management high-caliber AI models can be developed at a fraction of the traditional costs.

The release of DeepSeek caused tremors in the US Stock market. Nvidia, a leading AI chip manufacturer experienced a record one-day loss of over $500 billion in market value. The fear was that DeepSeek’s low-cost AI models might reduce the demand for high-end GPUs. Other major tech firms, including Microsoft, Alphabet (Google’s parent company), and Meta, also saw steep decline in their stock prices.

The US has employed restrained technological strategies that has restricted China to access to advanced semiconductor technology from firms like Nvidia GPUs and banning exports to entities involved in China’s AI development. In response, China pursued AI self-reliance which is exemplified by DeepSeek’s model. Their LLMs have been trained using just 2,000 Nvidia H800 GPUs—hardware that the US still permits China to import but at reduced capability. The success of DeepSeek has significantly undermined US efforts to cripple China’s AI ecosystem. Instead of retaliating directly by imposing similar counter measures, China has chosen to implement alternative approaches that could successfully circumvent US-imposed technological chokepoints.

If ASEAN nations pivot toward China’s AI ecosystem, it could dilute US digital influence in the region. Hence to counterbalance DeepSeek’s growing reach, the US may look to strengthen AI partnerships with ASEAN via Quad and AUKUS digital initiatives. It could expand access to OpenAI and Google APIs for ASEAN startups and look to impose AI governance frameworks with an implicit aim to limit the spread of Chinese-trained AI models.

While US is wary of these developments, for Beijing, DeepSeek is not merely a technological success—it is as well a geopolitical instrument too. China can export DeepSeek-based AI models to ASEAN nations at a fraction of cost creating digital dependencies and reinforcing Beijing’s technological dominance in Southeast Asia. By countering the Western-centric AI models that dominate global discourse, China could position DeepSeek as a ‘non-western’ AI alternative.

Historically, ASEAN nations have pursued a kind of strategic hedging, that is, avoiding exclusive alignment with either the US or China. However, DeepSeek’s rise presents some strategic choices. For example, countries like Cambodia, Myanmar that has already aligned with Beijing to a good extent, could embrace DeepSeek and integrate it into government and corporate AI applications. Other countries such as Indonesia, Malaysia, Thailand, Vietnam could proceed via AI partnerships. Here DeepSeek could be applied for low-cost AI applications while they could continue to maintain US partnerships for cutting-edge AI research. This way, it continues to mirror ASEAN’s broader “neutrality strategy” in geopolitics. There is a chance that Singapore, the Philippines, and others may resist Chinese AI models, citing concerns over bias, misinformation, and surveillance risks. Nonetheless these nations could adopt EU-style AI regulations while ensuring Chinese-trained LLMs do not dominate their digital ecosystems.

ASEAN nations will have to move with full steam if they have to accomplish the task of developing their own competitive LLMs. They must reduce dependence on US and for this they gave to invest heavily in cloud-based AI services within domestic AI clusters. Also, nations in this region could create AI models that are bespoke for local languages (i.e. Bahasa Indonesia, Bahasa Malaysia, Thai, Khmer, or for that matter, Vietnamese). This way they would in a position to move way and stop their reliance on English-trained models. A regional ASEAN AI initiative that works collaboratively with ASEAN nations could pool resources to develop an indigenous LLM for Southeast Asia.

With DeepSeek we are seeing a fundamental shift in AI geopolitics that allows for a cost-efficient AI development outside the US-dominated AI ecosystem. Here I see a good opportunity as well as challenges for Southeast Asia. While the Deepseek phenomenon lowers the barrier to AI entry, making regional AI independence more feasible, it also creates a new arena for US-China digital competition within ASEAN. ASEAN governments would be forced to carefully navigate AI dependencies, ensuring that they do not become over-reliant on either Washington or Beijing. If ASEAN nations seize this moment, it could transform passive consumers into active contributors in the AI revolution. However, if they fail to act, they risk becoming mere battlegrounds in the digital Cold War between the US and China.

First published at EurasiaReview on February 9, 2025