Global Markets
Home›Global Markets›Trade & Tariffs›AI inference costs fall to 2026 low amid price war and…
AI inference costs fall to 2026 low amid price war and open-source shift
Average inference prices ranged from US$1.16 to US$1.18 per million tokens from Aug. 6 to 8, down from US$2.04 on May 31 and US$1.45 in late July.
Businesses’ costs to run AI models have dropped to a yearly low, with pricing pressured by a global price war and growing use of low-cost Chinese open-source tools, according to research cited by SCMP Economy.
SCMP Economy, citing Jefferies and US research firm Silicon Data, said average inference prices for business AI models ranged between US$1.16 and US$1.18 per million tokens from Aug. 6 to 8, the lowest level recorded this year.
The decline marks a sharp step down from US$2.04 per million tokens on May 31 and US$1.45 in late July, the research showed. Silicon Data’s index tracks pricing across AI API providers and open-weight inference platforms used by software developers.
Jefferies analysts, including Thomas Chong, attributed the move to increasing emphasis on cost efficiencies across both US and Chinese tech ecosystems, adding that Chinese firms are pushing further on affordable computing through open-source approaches.