Dive Brief:
- Snowflake launched an AI cost management feature on Tuesday in a bid to help enterprises navigate rising spend. Dynamic model routing will automatically select the appropriate AI model for tasks based on quality and cost, according to a press release. The capability is available within Cortex AI Gateway and other Snowflake AI products.
- Snowflake is also expanding customers’ access to frontier AI models, including the latest models from DeepSeek and Z.ai. The AI cloud company said along with expanded access to more efficient models, the new dynamic model routing feature further drives enterprise efficiency in translating models, compute and data into business outcomes, the release said.
- Enterprises are becoming much more rigorous about gaining meaningful business value from AI, Snowflake CEO Sridhar Ramaswamy said in the release.
Dive Insight:
Snowflake’s cost-control feature marks the latest vendor move to address executive concerns about AI spend.
Roughly three-fifths of IT professionals believe AI overspend has increased, despite token costs coming down, according to a Flexera study. Token costs might be dropping, but advances such as agentic AI demand significantly more tokens than generative AI. Gartner estimates that global token consumption will reach about 300 trillion tokens per day by the end of 2026.
As token usage rises, so do executives’ worry about the costs. The Linux Foundation launched the Tokenomics Foundation earlier this year to unite enterprises, hyperscalers and frontier model developers on developing frameworks and best practices to manage AI use at scale.
Vendors are responding to executive concerns about AI costs. Oracle rolled out token bundles for agentic AI in June. Meanwhile, AWS announced a public preview of its AWS FinOps agent to investigate cost anomalies.
The dynamic model routing offered by Snowflake aligns with vendor momentum across enterprise AI cost management. The capability is designed to help enterprises use expensive frontier models only when they improve outcomes across Snowflake’s AI products, according to the press release. The Cortex AI Gateway also updates routing decisions as model performance and pricing change.
“You don’t need the latest and greatest Opus models for summarizing Slack threads,” Ramaswamy told investors during the company’s Q1 2027 earnings call in May. “We are creating the controls that one needs in order to keep costs manageable as things continue expanding.”
