The new border: Why server farms are the battleground of AI sovereignty

In May, US authorities accused a Bangkok-based firm, linked to Thailand’s AI initiative, of transferring Nvidia-powered servers to Chinese buyers. Despite Thailand’s efforts to build chips, data centers, and establish itself as a key player in the AI landscape, it now faces federal scrutiny. The infrastructure that symbolized the country’s ambition has become evidence against it.

Traditionally, selecting between cloud providers like AWS, Alibaba Cloud, and domestic operators was driven by cost and latency considerations. However, export controls and licensing have evolved beyond mere categories, becoming a contentious issue. In October, the Senate passed a defence bill with the GAIN AI Act, requiring chipmakers to fulfill US orders before shipping advanced chips to China.

The bill's momentum dwindled when the White House intervened, leading to mixed messages and contradictory decisions. While some advanced chips were blocked, others were cleared for shipment to China. This legislative inconsistency has resulted in consequences beyond mere fines – companies lose the essential infrastructure supporting their AI strategies.

The situation isn’t limited to Thailand alone; Malaysia also faces similar challenges. Prime Minister Anwar Ibrahim recently froze approvals for new data centers unless they demonstrated a clear AI benefit, acknowledging the need to prioritize projects that can withstand scrutiny. This move reflects the broader trend of regional powers cautiously navigating the complex landscape of AI development.

Malaysia has become a hub for major cloud providers, including AWS and Google, alongside Alibaba. Similar trends can be observed in countries like Thailand, which welcomed its first Google Cloud region in the same year. Singapore, too, has updated its AI strategy, emphasizing AI as a national capability. Vietnam has taken a more stringent approach by enacting an AI Act, set to come into effect by March 2024.

However, Indonesia’s approach presents a unique case study. The country is planning a sovereign AI fund to position itself as a regional AI hub. Despite the involvement of foreign entities like Huawei and an Indonesian tech giant, the plan lacks a swift implementation timeline, raising questions about the feasibility of realizing these aspirations.

While these developments may appear aimed at achieving AI sovereignty, the reality is far more complex. The reliance on foreign chips and the strain on existing infrastructure undermine true sovereignty. These efforts can be seen as a form of diversification, but it remains unclear if this strategy is truly effective. The regional approach, rather than aligning with a single bloc, focuses on forging alliances with multiple countries simultaneously.

Governments in Malaysia, Indonesia, and Thailand are increasingly seeking to balance their relationships with both the US and China, using AI as a tool to showcase progress without taking definitive stands. However, domestic conglomerates, particularly those controlling critical resources such as energy, land, and water, play a significant role in shaping the AI landscape.

Their interests often align more closely with licensing pipelines and water rights than with the US-China rivalry. The geopolitical frame, while real, serves as a convenient cover for domestic power struggles that would persist even without the influence of major chip suppliers. In conclusion, the concept of AI sovereignty remains elusive.

While hedging strategies offer governments temporary respite and the opportunity to present progress, the underlying challenges related to domestic capacity and enforcement persist. For businesses, this presents a complex situation where local infrastructure and regulatory clarity are crucial for the successful adoption of AI technologies.