Since the advent of the computer age, industries have been so awash in stored data that most of it never gets put to use.This data is estimated to be in the neighborhood of 120 zettabytes - the equivalent of trillions of terabytes, or more than 120x the amount of every grain of sand on every beach around the globe. Now, the world's industries are putting that untamed data to work by building and customizing large language models (LLMs).
As 2025 approaches, industries such as healthcare, telecommunications, entertainment, energy, robotics, automotive and retail are using those models, combining it with their proprietary data and gearing up to create AI that can reason.
The NVIDIA experts below focus on some of the industries that deliver $88 trillion worth of goods and services globally each year. They predict that AI that can harness data at the edge and deliver near-instantaneous insights is coming to hospitals, factories, customer service centers, cars and mobile devices near you.
But first, let's hear AI's predictions for AI. When asked, What will be the top trends in AI in 2025 for industries? both Perplexity and ChatGPT 4.0 responded that agentic AI sits atop the list alongside edge AI, AI cybersecurity and AI-driven robots.
Agentic AI is a new category of generative AI that operates virtually autonomously. It can make complex decisions and take actions based on continuous learning and analysis of vast datasets. Agentic AI is adaptable, has defined goals and can correct itself, and can chat with other AI agents or reach out to a human for help.
Now, hear from NVIDIA experts on what to expect in the year ahead:
Kimberly Powell
Vice President of Healthcare
Human-robotic interaction: Robots will assist human clinicians in a variety of ways, from understanding and responding to human commands, to performing and assisting in complex surgeries.
It's being made possible by digital twins, simulation and AI that train and test robotic systems in virtual environments to reduce risks associated with real-world trials. It also can train robots to react in virtually any scenario, enhancing their adaptability and performance across different clinical situations.
New virtual worlds for training robots to perform complex tasks will make autonomous surgical robots a reality. These surgical robots will perform complex surgical tasks with precision, reducing patient recovery times and decreasing the cognitive workload for surgeons.
Digital health agents: The dawn of agentic AI and multi-agent systems will address the existential challenges of workforce shortages and the rising cost of care.
Administrative health services will become digital humans taking notes for you or making your next appointment - introducing an era of services delivered by software and birthing a service-as-a-software industry.
Patient experience will be transformed with always-on, personalized care services while healthcare staff will collaborate with agents that help them reduce clerical work, retrieve and summarize patient histories, and recommend clinical trials and state-of-the-art treatments for their patients.
Drug discovery and design AI factories: Just as ChatGPT can generate an email or a poem without putting a pen to paper for trial and error, generative AI models in drug discovery can liberate scientific thinking and exploration.
Techbio and biopharma companies have begun combining models that generate, predict and optimize molecules to explore the near-infinite possible target drug combinations before going into time-consuming and expensive wet lab experiments.
The drug discovery and design AI factories will consume all wet lab data, refine AI models and redeploy those models - improving each experiment by learning from the previous one. These AI factories will shift the industry from a discovery process to a design and engineering one.
Rev Lebaredian
Vice President of Omniverse and Simulation Technology
Let's get physical (AI, that is): Getting ready for AI models that can perceive, understand and interact with the physical world is one challenge enterprises will race to tackle.
While LLMs require reinforcement learning largely in the form of human feedback, physical AI needs to learn in a world model that mimics the laws of physics. Large-scale physically based simulations are allowing the world to realize the value of physical AI through robots by accelerating the training of physical AI models and enabling continuous training in robotic systems across every industry.
Cheaper by the dozen: In addition to their smarts (or lack thereof), one big factor that has slowed adoption of humanoid robots has been affordability. As agentic AI brings new intelligence to robots, though, volume will pick up and costs will come down sharply. The average cost of industrial robots is expected to drop to $10,800 in 2025, down sharply from $46K in 2010 to $27K in 2017. As these devices become significantly cheaper, they'll become as commonplace across industries as mobile devices are.
Deepu Talla
Vice President of Robotics and Edge Computing
Redefining robots: When people think of robots today, they're usually images or content showing autonomous mobile robots (AMRs), manipulator arms or humanoids. But tomorrow's robots are set to be an autonomous system that perceives, reasons, plans and acts - then learns.
Soon we'll be thinking of robots embodied everywhere from surgical rooms and data centers to warehouses and factories. Even traffic control systems or entire cities will be transformed from static, manually operated systems to autonomous, interactive systems embodied by physical AI.
The rise of small language models: To improve the functionality of robots operating at the edge, expect to see the rise of small language models that are energy-efficient and avoid late
