Snowflake, AI, & Data Infrastructure: 4 Predictions for 2025
This past year has been transformative for the Snowflake community. Just six months ago, I wrote about the groundswell of AI innovation among Snowflake users, and that trend has only continued. In fact, I would say it’s accelerated.
From tens of conversations I’ve been having with CIOs and CFOs who are leading the charge in AI adoption, it’s clear they’re seeing significant results. These include cost savings, performance enhancements, granular observability into warehouse and query health, increased engineering efficiency, and more.
If this groundswell continues into 2025 (and I believe it will), we can expect to see some changes among the Snowflake community. Here are four of my top predictions, which I wanted to share in hopes that it will be useful as people think through their strategic planning for next year.
1. CFOs and CIOs will become more strategic about infrastructure spend
This past year, data infrastructure spend increased over 36% year-over-year. And that growth isn’t likely to slow down—some sources are projecting a 26% CAGR until 2028. So it’s not surprising that this line item is getting a lot of attention from CFOs.
There are several ways to address this problem. Let’s focus on Snowflake as an example, which is also one of the largest line items within the data infrastructure category for many customers. One solution is to have better observability for accurate capacity planning and for identifying where to manually optimize workloads (warehouse rightsizing, query routing, etc.). Another is to automate these optimizations, either with tools developed internally or external tools that are more mature, have built-in AI, and cost less to implement..
As such, my first prediction is that CFOs and CIOs will start thinking more strategically about their infrastructure spend. Specifically, they’ll realize that buying a third-party optimization solution offers more of a net positive for both the top and bottom line.
2. Shift from generalized to verticalized FinOps solutions
Until now, CIOs have relied on general-purpose tools for FinOps. However, these have limited capabilities and only observe and report: e.g. X team is spending Y dollars over-budget, or X% of the spend is due to this set of users.
But knowing that a team is overspending or using a platform inefficiently is only Step 1. To actually fix the problem, you need to know the root causes of those inefficiencies—which often vary by platform.
To address this problem, I predict that CIOs will start adopting more verticalized tools to gain more granular insights into their data infrastructure platforms, including Snowflake. Because of their platform-specific nature, these tools can offer detailed, AI-generated recommendations for improving efficiency, and an estimated impact of these changes.
3. The evolution of the engineer’s value
Until recently, the division of labor between humans and machines was clearly delineated: human engineers would handle open-ended tasks, while machines took care of well-defined, repetitive, and simple jobs. As AI has become smarter and more sophisticated (for example, reinforcement learning mimicking human learning and decision-making), the line between where humans add value vs. machines has become blurred.
But although AI agents can take on more open-ended problems than before, they aren’t error-free. What’s more, they lack the domain expertise and understanding of business constraints that human engineers have.
As more engineers become aware of this reality, I predict that more of them will stop resisting AI, and instead become its advocate and champion. More importantly, they’ll figure out how to add value on top of it: deciding where to automate (and where not to), designing safeguards to protect performance and spend, and vetting solutions to find the best ones.
4. Not all savings are created equal
When budgets tighten, CFOs and department leaders often apply across-the-board cuts and spending freezes. But this can often be counterproductive, as some purchases actually help to reduce spend—especially those that control data infrastructure costs.
Basically, there’s a cost to cost savings.
I’ll use an example from one of our customers at Keebo. They migrated from Snowflake to a less mature cloud data warehouse offering, only to regret that decision in three months because the vendor’s product gaps required the customer to spend months of engineering time trying to identify heavy hitter queries, users, warehouses, etc. So yes, the unit cost was cheaper, but a job that would have taken them five minutes on Snowflake took three hours on this less mature alternative.
As the C-suite becomes smarter and more strategic in managing costs, I predict more vetting of options and trade-offs when it comes to data infrastructure spends. After all, you don’t want someone making $200,000 per year spending three whole months just to save the company $50,000 on their Snowflake bill. That’s not a good use of resources.
Final thoughts
2025 will be a year of unprecedented opportunity for savings, efficiency, and growth. Much of this opportunity will be fueled by AI. But equally important are the visionary leaders who will recognize this opportunity and take steps to realize it.
The question is: where do you fall? Do you want to lead the charge in AI innovation? Or will you sit on the sidelines until all your competitors have run ahead?
When you’re ready for the future, here at Keebo, we’re always ready to help.