Valuations of chip manufacturing exploded with the AI boom, skyrocketing the worth of companies like Nvidia and AMD. The semiconductor market is soaring and projected to continue growing with demand for AI.
A booming semiconductor market suggests greater access to—and demand for—AI tools over time. However, some businesses predict a global chip shortage could bottleneck access to AI solutions.
Large language models are still improving at an exponential rate. They are continually able to process more complex tasks as their context windows grow.
AI tools also face a slimmer chance of government regulation next year. An incoming Trump administration means AI regulation is less likely than under a Democratic administration, Kahn suggested.
“Restrictions around regulating AI will maybe open up to allow for more independent commerce to figure out what it really means—versus having the government tell you what it’s not going to be,” Kahn told FleetOwner.
New AI applications for fleets next year
AI technology will likely continue to improve in 2025, and AI solutions will become more widely available. But what improvements are important for fleet operations?
Paul Pallath, VP of applied AI for Searce, pointed to three areas where AI could change transportation in 2025—AI agents, multimodal data management, and video generation.
AI agents
Large language models ((LLMs) are one of the most potent AI developments in recent years. LLM developers are looking to make the technology even more impactful.
Agentic AI, sometimes called robotic process automation, may be the next major development for large language models. This technology places a single AI chatbot in front of the user for commands—but behind the chatbot lies an intricate stack of multiple AI tools.
Users can ask the agent to perform a data-centric task in a similar conversational way that they would ask an employee. Layers of LLMs could perform, review, and automate the processes required for the overall task. Getting an invoice number or reconciling multiple invoices could be done much more quickly—and simply—with the help of agentic AI.
AI giants like Google with Gemini 2 and Nvidia with Blueprints are already developing agentic AI to solve more complicated tasks.
“The premise of Gemini 2 is that it enables agentic AI at scale that allows us to now have autonomous bots … that manage multiple different intelligent bots that are task-specific large language models,” Pallath said.
Multimodal data management
One application already growing in popularity is using AI to consolidate disparate forms of data. Paper forms and video feeds contain helpful data—but joining them in a single spreadsheet for a fleet’s TMS can involve a lot of manual labor. AI-powered image and language processing could automate the process.
“LLMs are multimodal, which means that it can connect with data, which is textual information, audio information, and video information, and process it in one go,” Pallath said. “If I’m asking for something, and if this information is in three disparate sources, LLMs now have the capability to summarize all of these, take the entire context, and give me a response.”
AI is already turning video feeds into automated truck logging and driver distraction detection.
“There are pockets where AI is thriving. Video-based telematics is moving very aggressively in that space,” Pallath said. “There has been a monumental shift in the results of that, where you see increased accident reduction and increased driving risk reduction.”
In addition to videos, AI tools could also consolidate paperwork such as bills of lading or audio recordings of customer transactions into more accessible formats. Paired with AI agents, using and accessing several types of information could become much easier.
Video generation
Pallath also suggested that AI video generation could become much more useful in 2025.
Similar to large language models, AI video generators have entered an arms race. Companies like Google and OpenAI continue to launch competing AI video generators that appear more and more convincing.
Paired with technical documentation, Pallath said AI video generation could soon create training videos for maintenance operations.
link