- Analysis suggests Singapore’s AI strategy focuses on skills and applications but hasn’t fully addressed “AI sovereignty.”
- Author Marcus Loh identifies Singapore’s AI economy as dependent on three strategic factors: frontier models, AI tokens, and semiconductor chips.
- Approximately 70% of Singaporean businesses now use AI to some extent, mostly relying on closed models like GPT, Claude, and Gemini.
- These systems offer high performance but create dependency on foreign companies and governments.
- The article warns that policy changes or access restrictions from a few major AI providers could simultaneously affect banks, ports, education, and SMEs in Singapore.
- The author proposes a national “model access” strategy, including supplier diversification and long-term agreements for critical sectors.
- Another risk is the sharp increase in AI token costs as businesses deploy large-scale agentic AI.
- OpenRouter analysis shows GPT-5.5 prices doubling compared to GPT-5, while actual costs rise 50-90% when using reasoning and large context windows.
- Gartner predicts agentic workflows could consume 5-30 times more tokens than current AI queries.
- Uber reportedly exhausted its 2026 AI budget in just months due to heavy reliance on Anthropic’s Claude.
- Nvidia VP Bryan Catanzaro stated that in some research groups, compute costs have exceeded personnel costs.
- China currently consumes over 140 trillion tokens daily, a 1,000-fold increase in just two years.
- The author suggests Singapore treat AI tokens as a strategic resource similar to oil, water, or food reserves.
- The third risk is dependency on US GPUs, as most AI workloads in Singapore run on Nvidia chips subject to Washington’s export controls.
- The US HR 2683 Act allows for controlling cloud access to AI GPUs as a form of technology export.
- Singapore currently has no public commitment from the US regarding prioritized GPU access or long-term chip supply.
- The article argues that if compute, model access, and token pricing are out of control, Singapore’s AI economy could become strategically vulnerable.
📌Sovereign AI will become a survival issue for Singapore in the AI era. While the nation may successfully train its workforce and promote AI adoption, it still faces the risk of dependence on frontier models, token pricing, and US-controlled GPUs. The author argues that Singapore needs to view AI as strategic national infrastructure similar to oil or electricity, including token reserves, guaranteed model access, and building long-term semiconductor relationships to avoid future disruptions or geopolitical pressure.
