Falling AI token costs could fuel next leg of Singapore semiconductor rally: analysts

The significant drop in AI token costs, with the Large Language Model Token Expenditure index falling to US$0.9665 by August 31 from a May peak of US$2.0651, is the core development here. This reduction in operational expense for AI models could directly stimulate wider adoption of AI applications and agentic AI. For Singapore, this translates into a potential boost for its semiconductor equipment suppliers like AEM, UMS, and Frencken, whose shares have already seen substantial gains this year. The immediate implication for Asia is a potential acceleration in AI infrastructure investment, as the economics of deploying AI improve. While the share price rallies for these Singaporean firms are notable, the sustainability hinges on actual enterprise adoption of AI solutions, not just speculative bets. The real test will be whether falling token costs translate into tangible increases in orders for chip manufacturing equipment in the coming quarters.