Snowflake adds OpenAI models with Microsoft integration

The integration

Many enterprises are increasing investments in developing AI tools, including generative AI applications, because of their potential to make workers better informed and more efficient.

To develop such applications that understand their unique business operations, enterprises need to combine generative AI models with their proprietary data, with data providing the "intelligence" in AI.

Because data's vital role in AI development, many data management vendors have added AI development environments. Snowflake is one. Others -- among many -- include archrival Databricks and tech giants Amazon, Google and Microsoft.

As part of development environments, vendors provide access to language models using native integrations that keep proprietary data secure and reduce the need for developers to configure complex connections.

Snowflake Cortex AI now includes integrations with models from seven AI vendors, including Snowflake's own Arctic LLM and models from AI 21 Labs, Anthropic, DeepSeek, Meta, Mistral and OpenAI.

Beyond adding another set of models to choose from, benefits of making OpenAI's models natively available in Cortex AI include the security that comes from not having to move data out of Snowflake and built-in data governance via Snowflake's Horizon Catalog.

By broadening the number of language models available in Cortex AI, Snowflake is providing developers with greater choice. However, even more significant is that Snowflake is providing choice alongside security and governance, which could attract customers and lead to growth, according to Andy Thurai, an analyst at Constellation Research.

"Models will become less and less differentiated soon enough, so that is not where the money is," he said. "By offering platform services -- such as data governance, security, access controls and building trusted AI on their platform -- companies can make money."

The integration also benefits Microsoft and OpenAI, Thurai continued, noting that Snowflake customers not using Microsoft could be attracted by the combination.

"From a Microsoft and OpenAI standpoint, this is a brilliant move," he said. "They want Snowflake customers from cross-region and cross-cloud to send their AI inference workloads without needing to go through long integration exercises."

Regarding the impetus for adding access to OpenAI's language models, Baris Gultekin, Snowflake's head of AI, cited customer feedback as a primary reason.

"What we keep hearing from customers is that they want to run the best models directly next to their data, within the security boundary of Snowflake, so they don't have to move or copy their data and can instead focus on driving AI-powered insights," he said.

In particular, having the opportunity to choose which model to use in development has been a request, Gultekin continued.

"Enterprises also want flexibility to choose their LLMs, because no single model excels at everything," he said.

While the integration provides joint Snowflake and Microsoft customers with added choice, the move does not yet make Snowflake's AI development capabilities as robust as those of its main competitors, according to Catanzano.

Unlike Databricks and Microsoft, which quickly pivoted following ChatGPT's launch to prioritize AI development, Snowflake didn't fully commit to AI development until Sridhar Ramaswamy was named CEO in February 2024.

Now, while providing strong security and governance, its model training and fine-tuning capabilities trail.

"Snowflake has made major strides in AI by integrating OpenAI, offering strong security and enabling cross-cloud inference," he said. "However, it still lags behind Databricks in native model training and fine-tuning, which remains a key differentiator for advanced AI development."

Thurai likewise noted that while Snowflake is adding features to improve Cortex AI, it has more to add before its functionality matches other AI development platforms.

"Snowflake has struggled a little bit to catch up on the AI wave," he said. "While they have a lot of corporate data, convincing enterprises to use them as a platform to access models has been a difficult task."

Plans

As Snowflake continues to expand its AI development environment, agentic AI will be a focal point, according to Gultekin. Unlike AI assistants and chatbots that can respond to questions from users, agents can act autonomously to proactively make suggestions as well as take on certain repetitive tasks previously performed by humans.

Agentic AI has become the biggest trend in AI over the past year, and that is expected to continue.

"We believe that AI agents will soon be essential to the enterprise workforce," Gultekin said.

Catanzano, meanwhile, suggested Snowflake work to address some of its shortcomings, such as custom model training. In addition, more partnerships with AI vendors, added developer tools and building industry-specific AI solutions would be wise.

"Strengthening these areas will help Snowflake match or surpass rivals like Databricks in AI development capabilities," Catanzano said.

Eric Avidon is a senior news writer for Informa TechTarget and a journalist with more than 25 years of experience. He covers analytics and data management.