![JMA – Ninja & Deep Jandu [Official MV] Kaptaan](https://i.ytimg.com/vi/-dsfIh319s0/maxresdefault.jpg "JMA – Ninja & Deep Jandu [Official MV] Kaptaan")
Vijayant Rai, Managing Director, India, Snowflake
Snowflake is using dynamic model routing to help enterprises cut the cost of running AI workloads by recommending models based on the complexity and economics of each use case as companies increasingly move AI applications from experimentation into production, according to Vijayant Rai, Managing Director, India, Snowflake.
The company recently announced Cortex AI Gateway, which allows customers to choose the right model on the Snowflake platform for the right use case. Rai said the need to manage AI costs is becoming increasingly important as adoption accelerates.
“If you want your AI to scale, it has to be cost-efficient. For the outcome you are getting, the input cost has to make sense — an important parameter for Indian organisations. CIOs and CFOs are watching with a hawk’s eye to see if they are paying the right amounts for the outcomes they’re getting,” Rai said.
Rai said Snowflake’s AI Data Cloud allows customers to keep their data on the platform while accessing AI and LLMs from providers including Anthropic and OpenAI, as well as open-weight models such as DeepSeek and GLM. Recently, Snowflake also announced support for the latest models from DeepSeek-V4-Flash 0731 and GLM-5.3.
Dynamic model routing analyses usage patterns for a particular use case and recommends the most appropriate model. For instance, instead of using a frontier model, enterprises could opt for an open-weight model that delivers comparable accuracy at one-third the cost.
“With dynamic model routing, we’re making recommendations; it gives them the option of which models to use for what inferencing, based on cost efficiency. For a simple thing to be done at scale, you might want to use an open-weight model over a frontier model because the latter is too expensive. For some deep research work like complex enterprise planning, you might want to use a frontier model,” Rai explained.
The move comes as Snowflake sees customers shifting from AI experimentation and pilots to production AI use cases.
Rai observed a shift in adoption cycles, moving from earlier experimentation to pilots to real AI use cases that customers are getting into production.
In India, a majority of Snowflake customers are using its AI features, he said, with the company seeing particularly strong adoption among digital-native businesses and startups.
“In India, we have digital natives, the startup and unicorn community, who are one in the cloud. They are comfortable with the latest technologies and use them at scale. This part of the industry is way ahead in adoption cycles. On the other hand, traditional enterprises are moving from integrating pilots and chatbots for different parts of the business to looking at enterprise context and building an enterprise-scale, agentic framework,” he noted.
Snowflake expects AI adoption to further expand its India opportunity beyond its traditional data-cloud business.
“We are getting bigger in India from a revenue standpoint; AI adoption has only increased our market. Earlier, the market opportunity was largely around data transformation. Traditional organisations are moving from on-premises systems to the cloud, etcetera, and getting those benefits. Along with all this, which is breaking data silos and moving to the Snowflake AI Data Cloud, we also have the opportunity of AI. That is an exponential increase in our total addressable market,” Rai shared.
Published on September 15, 2026
