AI predictions for 2025

Getting your Trinity Audio player ready...

It’s nearly impossible to go online, read a newspaper, or watch television without hearing about artificial intelligence (AI) in some way, shape, or form. In November 2022, when OpenAI launched ChatGPT and made AI accessible to a general audience that lacked a technical or academic background, for the first time in history, the AI genie left the bottle, and there are no signs of putting it back in.

Since then, we’ve seen enormous progress and innovation in the AI industry. From the text-based chatbots that kicked off the wave to generative image and audio AI and advancements in the foundational models that underpin all AI applications, a lot has happened since 2022. That being said, what’s next?

So much progress has occurred in the last two years that it can be difficult to imagine what else could unfold in the AI industry in the coming year. Lucky for you, I look at AI news daily so that you don’t have to! Based on what I’ve seen in the last year, here are the trends, innovations, and products I expect to gain traction in 2025.

The rise of AI agents

AI agents are coming, and you could argue they’re already here if you consider the iterations primarily available to developers and early-access users. AI agents—or intelligent agents—are AI model interfaces that execute tasks autonomously. Acting like personal assistants, they can complete multi-step processes with minimal, ideally zero, user input after receiving initial instructions.

In my opinion, launching agents will be the next significant leap forward for many AI service providers.

Some companies have already unveiled their AI agents. For example, Anthropic’s “computer use” feature for its AI chatbot, Claude, allows the AI to interact with computers as humans do—clicking buttons, typing, and planning complex tasks such as organizing trips or creating websites. Similarly, Microsoft (NASDAQ: MSFT) has rebranded its Copilot product as “Microsoft AI Agents,” which integrates with external systems with enhanced memory to enable more complex workflows than its predecessor.

As soon as AI agents exit their walled gardens and closed betas, I’d bet good money that tech giants like Google (NASDAQ: GOOGL), Meta (NASDAQ: META), and OpenAI will race to release their agents. These agents are essentially the AI dream the public was promised and told stories about when consumer AI became popular: having a personal assistant in your pocket or on your screen, handling mundane tasks so you can focus on more critical work, or just relax. AI agents are likely to become a reality and gain traction in 2025, meaning that the AI dream people were promised is nearly manifesting.

Text-to-Video AI: Generative tech’s next leap

The next big leap in generative AI will be text-to-video capabilities, another innovation that is much closer to being a reality than we might think. We’ve already seen text-to-text chatbots, text-to-image generators, and text-to-audio applications, but text-to-video has remained elusive—until now.

In 2024, several companies, particularly OpenAI with its product “Sora,” began to tease and eventually launch their generative AI video tools. But before OpenAI was able to release Sora to the public, rumors of its imminent release prompted Google to preview its own generative video product, “Veo,” which is currently accessible to a select group of users.

Regardless, with tools as powerful as generative video, businesses will gain the ability to create high-quality multimedia (think commercials) at unprecedented speed and low cost, while individual creators could produce studio-quality content (think short films and movies) without needing significant resources. This is bound to have a significant impact on the creative world, leveling the playing field in an industry known for its barriers to entry and high costs.

That being said, I think we are going to see generative AI video in a lot of different places. I’m even going to go ahead and predict that we will see a Super Bowl commercial entirely created with AI in 2025.

Re-education campaigns by AI service providers

As innovations like AI agents and generative video emerge, AI service providers will probably need to re-educate their user base, so I expect to see that unfold in 2025. This could come through explicit tutorials, workshops, or implicit methods like blog posts and videos showcasing their products in action.

Currently, there’s a significant gap between how consumers use AI and its full potential. Many people treat AI chatbots as glorified search engines, asking questions and receiving direct answers. While this is valuable, AI can do so much more—from automating workflows to creating software prototypes. For users to maximize these capabilities, they need to understand how to prompt AI effectively.

Re-education campaigns will be essential for AI companies to encourage wider adoption of advanced features and, more importantly, to get potential users to understand the value in their offerings and the best ways to use them so that they can convert more users into paying customers. That takes us to an increasingly important element in the AI industry: profit.

The AI industry’s 2025 focus: Profitability and infrastructure

AI service providers are burning billions of dollars to train and run their offerings, but these services aren’t generating enough revenue to offset costs. This has led investors to scrutinize AI companies and demand clearer paths to profitability. In 2025, I expect this trend to continue. Investors will continue to be very vocal about the lack of profit they have seen from their AI investments and question their portfolio companies about when they can expect a return or even just profitability from the company.

Infrastructure investments—such as computer chips and data centers—have proven to be a money maker. Because they generate significant amounts of revenue, you can expect more investors to turn to infrastructure and to turn away from many of the AI startups we have seen entering the space looking for money. Infrastructure has been a money maker because companies can lease computing power to other businesses, creating a reliable profit center.

As all companies continue to hunt for profitability and face investor pressure to find it quickly, you can expect to see AI companies doing all that they can to achieve it. This probably means that they will get creative and open new revenue channels—like infrastructure plays—as well as cut costs, such as laying off employees and finding ways to run their business more efficiently.

AI’s energy solution: The shift to nuclear power

Even though infrastructure plays generate significant amounts of revenue, AI has an around-the-clock demand for computing power and, because of this, consumes substantial amounts of electricity. One area in which companies are looking to become more efficient is within their data centers.

To manage their intensive computing needs more efficiently, some AI companies are exploring nuclear energy to power their data centers. Nuclear power is extremely efficient and cost-effective in the long run. Microsoft, for example, has signed a deal to bring the decommissioned nuclear plant Three Mile Island back online so it can purchase power from the plant to run its data centers.

While nuclear power does look promising, companies aren’t going to reap its benefits anytime soon—even if they are making deals and investing today. It can take over a decade and significant amounts of money to reactivate decommissioned plants, which means in the short term, the switch to nuclear is going to be just another cost in an already very costly industry. However, investment in the future needs to begin today, which is why I believe we will see more companies take steps to switch to nuclear energy beginning in 2025.

Slowing progress and innovation in the AI sector

Admittedly, this is less of a prediction from me and more of a prediction from Google CEO Sundar Pichai, which I happen to agree with. During the New York Times annual DealBook summit, Pichai said, “I think the progress is going to get harder. When I look at [2025], the low-hanging fruit is gone,” adding: “The hill is steeper … You’re definitely going to need deeper breakthroughs as we get to the next stage.”

This is probably the truth.

In my opinion, growth in many areas often follows a stair-step pattern: periods of rapid progress followed by plateaus. With chatbots and image generators already mainstream and the next major consumer innovations, such as AI agents and generative video, on the way, the industry will probably be stuck in limbo for a while, focusing on refining existing models and tools rather than releasing new, groundbreaking products.

To add insult to injury, AI has become a space where you need very deep pockets to compete, which pretty much limits participation to tech giants and very wealthy individuals at this point. This means the actual pool of serious competitors in the space is much smaller than it appears, making unpredicted innovation a little less likely than it would be if there were more participants.

But! As history shows, breakthroughs can come from unexpected places—we’ll just have to wait and see what 2025 brings.

In order for artificial intelligence (AI) to work right within the law and thrive in the face of growing challenges, it needs to integrate an enterprise blockchain system that ensures data input quality and ownership—allowing it to keep data safe while also guaranteeing the immutability of data. Check out CoinGeek’s coverage on this emerging tech to learn more why Enterprise blockchain will be the backbone of AI.

Watch: Transformative AI applications are coming