Industry executives and experts share their predictions for 2025. Read them in this 17th annual VMblog.com series exclusive.
By Eduardo Mota, senior cloud data architect – AI/ML specialist, DoiT
Artificial intelligence (AI) and machine learning (ML) are changing faster than organizations can adopt them. The cloud continues to supply that much-needed flexibility to adapt infrastructure and advance. But because different models require different resources, locking workloads into specific infrastructure is detrimental to innovation. That said, buckle up. It’s still the Wild West in AI and its raucous reign shows no indication of settling down in 2025.
Advancing infrastructure and models
Developments are going to come fast and furious in infrastructure and model sizes. An emphasis on graphics processing units (GPUs) is underway and companies like AMD and Intel are in a race to optimize their chips for the GenAI era. This will continue to push innovation in chip architecture and manufacturing in the year ahead. What’s more, advancements in specialized models will become apparent, marked by the development of expert ones for common tasks such as programming, data processing with structured query language (SQL) and others.
Keeping up
New AI models are coming out regularly and businesses must be able to produce the same quality of content that competitors do. Close-sourced models remain the best to work with and are only available in the cloud, which is where those workloads belong. This will allow easy technology changes, process large amounts of data, reduce carbon footprints and even offload some infrastructure management, all of which will grow in importance in 2025.
It’s all about the data
In the year ahead, we will see organizations take a step back in their AI/ML strategy to examine their data roadmap. Although generative AI (GenAI) is powerful, the differentiator will be the data being utilized to enhance it. This will lead to a closer look into data strategy in order to distill only what’s required to create real value.
New use cases
Naturally, we can expect more sophisticated and impactful use cases to emerge, especially data processing at a larger scale. Context windows will continue to grow with models, and as such, the ability to get insights from large datasets will become routine. Graph databases will also be getting closer with GenAI, which will reduce latency and unearth relevant information much faster.
Regulation and hesitation
We’re going to see a bigger push for regulation in data collection and usage of GenAI. As the value of data increases, organizations will need to guard it better and try to collect more, which is where we’ll see more regulatory clauses come through. We’ll see organizations in general taking a step back as AI’s limitations – its inability to think or reason specifically – are shared, emphasizing that a human is needed to get the best out of the models.
Sleepless in security
There’s an increasing proliferation of bad actors finding holes in improperly configured infrastructure which will result in more than a few sleepless nights for chief information and security officers. The cloud was touted as “secure as can be,” and while that might be one of its capabilities, most organizations don’t have the bandwidth or expertise to properly configure and maintain a truly secure environment.
AI will cause security concerns to increase as bad actors can now generate realistic artifacts to trick employees. GenAI, in particular, will need a security perimeter where it can operate that eliminates any access to critical data. This is easier said than done due to the nature of GenAI and being unable to test every scenario. With this in mind, a well-crafted AI security strategy will be critical to ensuring data is safe.
AI is the way
Gartner has said generative AI is now in the “trough of disillusionment,” mainly because of the overhype it received. For AI engineers and data science practitioners, GenAI has already proven very useful in solving complex problems that once seemed unsolvable. However, at the same time, it’s not that science fiction AI we see on TV, and it is clear many technological issues need to be solved and their impacts discussed.
Although this may feel disillusioning to some, most all involved with AI agree it just keeps getting better and its full potential will be realized. The catalyst will be a re-evaluation of expectations as well as adopting a similar approach to AI that an entrepreneur uses when it comes to new business opportunities: Implement fast, in small steps, and adjust to user feedback.
The heights and depths of AI, and the many actions that take place in between, have left many saying “ugh.” But making revolutionary ideas a reality will do that to you, especially in 2025.
##
ABOUT THE AUTHOR
Eduardo Mota is senior cloud data architect – AI/ML specialist, at DoiT. An accomplished Cloud Architect and Machine Learning Specialist, he holds a Bachelor of Business Administration and multiple Machine Learning certifications, demonstrating his relentless pursuit of knowledge. Eduardo’s journey includes pivotal roles at DoiT and AWS, where his expertise in AWS and GCP cloud architecture and optimization strategies significantly impacted operational efficiency and cost savings for multiple organizations.
