Singapore – September 22, 2026 -- Enterprise spending on AI tokens is outpacing the infrastructure built to manage it, with McKinsey reporting that enterprise LLM spending tripled over twelve months even as per-token costs fell sharply. Goldman Sachs projects token consumption will rise 24-fold by 2030, exposing a gap in how companies track and finance one of their fastest-growing cost lines.
Accels upgrades its routing platform to cover 150+ AI models
Singapore-based Accels is expanding its Accels Router product to provide access to 15 leading model providers and more than 150 AI models through a single API. The platform, which has served customers in recent years, delivers low-latency routing across large language models.
CEO says token spending still runs on spreadsheets
"Token consumption is already one of the fastest-growing costs on the enterprise P&L," said Raymond, CEO of Accels. "But it is still managed with spreadsheets and manual reconciliation. There is no infrastructure for treating token usage the way finance teams treat any other material input."
Company plans to merge routing, billing and credit data into one system
Accels intends to combine routing data, billing data, and credit assessment into a single system rather than separate products. Under this model, a company's token consumption patterns -- volume, predictability, and cost -- would directly inform its credit profile and financing terms.
Roadmap extends beyond routing toward token-based financing tools
Accels is exploring additional capabilities aimed at how AI agents operate, spend, and scale, positioning token usage as both an operational input and a financeable asset. Raymond said the companies that win in AI will be those able to manage the economics of running agents at scale, not just those with the best models.