By Yotam Kramer
Despite the perception of many, as we enter 2025 AI is still in the very early stages of its adoption. In fact, while the technology has been in development for a number of decades, its application has only recently become a reality for the majority of enterprises.
While large language models such as the offerings from OpenAI may have taken much of the oxygen out of the room, it represents just one example of where AI can add value.
In fact, we’re already seeing the pendulum swing back to how to make money from AI.
Let’s take a closer look at the trends in AI, and key areas to watch in 2025.
Massive datasets need a different approach
Within AI, the prevailing narrative has long been “more compute power will yield better results.” However, the recent breakthroughs from DeepSeek will change this narrative.
We should also expect the narrative to change when approaching data. Here, CPUs, or central processing units, have long been the go-to for enterprises, yet this technology has been on the market for about 40 years. CPU-based massively parallel processing systems struggle with scaling, which means they often struggle with the complex and massive datasets of modern analytics. Meanwhile, platforms like Snowflake can mask a lot of challenges that only build up over time when ignored.
As CPUs reach their limits, we should expect to see greater adoption of graphics processing units in 2025. This may happen as organizations look for more efficient ways to process and analyze massive data volumes, leveraging the technology’s ability to handle thousands of parallel computations simultaneously, improving performance and cost efficiency for data-intensive tasks.
The year of the data scientist
While the profession may have been overlooked in years past, data scientists are in high demand, with a specialized skill set crucial to delivering AI goals for enterprises globally.
In fact, the median data science annual wage in the U.S. is $100,910, and the top 10% earn an average of $167,040 each year.
We should expect this trend to continue. This also means that running a team is increasingly expensive and competitive. As such, we’re likely to see a rise in companies turning more to AI for efficiency among its teams.
Expect data migration challenges to surge
AI hinges on access to data. Yet the task of collecting and processing this data — known as data migration — is a major headache for enterprises.
Due to their size and organizational complexity, enterprises work with massive data lakes. This often involves semi-structured and unstructured data stored on-site and on the cloud. To analyze this data appropriately it has to be moved around, yet many of these pipelines aren’t efficient. Using “schema-on-read” to process data can take hours and cause costs to skyrocket due to slow processing time.
This year will see this being tackled head-on. Enterprises will shift from their current focus on the end product to one that looks to cut costs and boost efficiency by processing data correctly at the source.
The biggest issue isn’t cost — it’s speed
Finally, 2025 will be the year that enterprises face up to the need for speed. Aside from the competitive edge that comes from faster analytics, speed is the most important metric to focus on to reduce overall running costs.
Finding faster, more efficient ways to migrate, process and handle data is the primary way that leaders can reduce costs without sacrificing quality or cutting the scope of work. These cost-saving initiatives make huge savings when the approach is applied across massive datasets and lakehouses.
Further, a survey from my company, SQream, found that 98% of respondents experienced machine learning project failures in 2023 despite having a budget of over $5 million for cloud and infrastructure. This highlights that investing in data projects exponentially increases the likelihood of successful results.
We should expect to see these four trends play a role in 2025. As enterprises focus on how to make money from AI, a much greater focus will be placed on how to make existing processes more efficient for better outcomes.
Yotam Kramer is the vice president of marketing at SQream Technologies. He is also a mentor at Techstars and has experience leading all marketing aspects and teams in B2B technology companies.
Illustration: Dom Guzman
Stay up to date with recent funding rounds, acquisitions, and more with the
Crunchbase Daily.
