For those prioritizing the use and improvement of AI applications in 2025, now’s the time to examine a few trends that may be waiting around the corner.
Data, of course, plays a key role in fueling AI, which performs most effectively and delivers the greatest value when fed massive amounts of data. I believe the ability to harness as much data as possible—in real time, at scale, and with 100% uptime—is what differentiates the leaders.
At the same time, the three C’s of AI continue to rule, namely: cost, cost, and… cost. This, along with compute problems, is escalating with demand for AI, as seen, for instance, in increasing competition for data center space and rising cloud costs related to a surge in data volumes. The IBM Institute anticipates that the average cost of compute will climb 89% by the end of 2025, with generative AI being a key driver.
It’s important for your organization to figure out how it will offset these costs before you deploy AI in earnest. Apple, Google, and Microsoft, for example, are offloading compute to distributed consumer devices. A lack of planning may cause a major setback: According to Gartner, more than 50% of organizations attempting to deploy AI and generative AI make “cost-related missteps” that lead them to abandon these projects altogether. It’s important to go in with eyes wide open.
In that spirit, let’s take a look at what we might expect from the AI and data management landscape in the months ahead.
1. THE CRAZY RACE FOR AI DOMINANCE WILL CONTINUE
In 2025, I predict the fast and furious race for AI dominance will only heat up as it becomes harder for enterprises to differentiate their AI solutions. The ability to improve the accuracy of results is key to standing out. That’s why many companies have adopted retrieval augmented generation (RAG) and vector search to augment their AI models with enterprise-specific context and serve it up at scale.
In the coming year, those farther along in the AI race are poised to move beyond better chatbots to fine-tune their models to deliver output that is specific to the unique situations of a given enterprise and its customers.
2. GEN AI KNOWLEDGE GRAPHS AND GRAPH EMBEDDINGS WILL INCREASE
In 2025, companies considering GenAI applications or already using them may look to knowledge graphs and graph embeddings to create differentiated GenAI solutions. These could be brought together with sophisticated databases and mature tech like machine learning to increase the accuracy and completeness of GenAI responses and associated business outcomes in customer support, fraud detection, and other applications.
One of our customers, an online money transfer, billing, and payments system, is a good example of a company that is ahead of the curve, having combined its graph database with our high-performance database solution to reduce its fraud exposure by 30 times.
3. AI AND SUSTAINABILITY WILL RECONCILE THROUGH COST EFFICIENCY
The clash between AI and the environment is only getting worse, with Goldman Sachs Research estimating that data center power demand will grow 160% by 2030. As increasing public scrutiny over the environmental impact of AI bumps up against the race for AI dominance—and the desire to control costs—it will become more urgent than ever for companies to take steps toward sustainable operations.
These steps include: Caching responses to repeatedly asked questions submitted to large language models and accessing those through efficient vector search; handling as much of the computation of a model locally as possible; and shrinking compute footprint while seeking to gain processing power.
I believe the biggest driver for companies to become more sustainable, however, will continue to be the cost advantages of doing so. Vendors who build efficiencies into their solutions naturally are able to reduce their carbon footprint relative to others while delivering the same performance. As more data is applied to problems and accessed more frequently, the efficiencies required of data management systems will increase.
4. AI REGULATION MAY STALL—BUT THAT DOESN’T MEAN COMPANIES CAN REST EASY
California Gov. Gavin Newsom vetoed a bill on AI safety in September 2024 on the grounds that it would curtail innovation. And while some AI regulation has happened in the U.S. at the state level (including in California), Congress has passed no comprehensive AI laws. It’s clear that nothing is slowing down the AI train, but that doesn’t signal a free-for-all for companies.
To reduce the risk of compliance violations and prevent the spread of misinformation through hallucinations, companies should invest heavily in using data properly to avoid financial and reputational risks. Doing so requires robust infrastructure that can ingest massive amounts of data at extremely high speeds at scale while delivering accurate output.
5. CLOUD COSTS COLLIDE WITH THE AI “GOLD RUSH”
Companies are hunkering down on the cost of cloud computing, and this could collide with the AI gold rush—the race to dominate the AI market. In the coming year, companies will likely look for ways to offload compute costs and squeeze non-AI investments for every dollar to counter the fierce battle for resources.
Some tough decisions are on the horizon for many businesses: Will they divert money from SaaS, developer cloud instances, or other investments to put toward AI? Can they improve compute efficiency by paying more attention to where data is processed—in the cloud, on-premises, or at the edge—using the awesome compute power of smartphones, tablets, and laptops as they become enabled with AI-specific processors?
KEEPING AI COSTS IN CHECK
There’s much to look forward to in 2025 on the AI and data management front. Those who take the time to plan for these upcoming opportunities have the best chance of establishing themselves as leaders in the AI space. Keeping costs in check will continue to be crucial, of course. Fortunately, this can be done by coming up with creative ways to offset compute and optimizing infrastructure to efficiently handle the massive amounts of data required to extract the most value from AI.