By Romel Dhalla
Artificial intelligence is awesome.
Since ChatGPT came out, I have been using AI every day. At first, it was bedtime stories for my kids, dinner recipes, and answers to whatever question happened to be floating through my mind. That was then. Now, I use it constantly for business.
AI has become my part-legal, part-accounting, part-financial-advisor, and all-in-one assistant. I am still cautious with its advice, and I still check important work, but it has improved dramatically. What used to be a common hallucination, when AI seemed to lose track of itself or produce an answer that made no sense, is now far less common. Newer AI systems are better at reasoning through complex problems, checking their own work, and catching inconsistencies before they become obvious errors.
It has become so useful, so fast, and so capable that it has meaningfully improved my life. It helps me plan, organize, review documents, monitor tasks, check emails, write, think, test ideas, and manage details that would otherwise eat up hours of my day. I even ventured to make a short space-station video game based on one I played as a child. Totally amazing.
The scary part is not that AI has displaced a good deal of advisory work I might otherwise have sought. It is not even that it gives me much of that work instantly, personally, and at a high level of usefulness. The scary part is that we are still just at the beginning.
AI is still learning, still growing, still getting faster and better with every passing day. At times, the difference between what AI could reasonably accomplish two months ago and what it can do today feels like the difference between an elementary student and a team of PhD-level graduates.
I recently completed a major project in five hours that would likely have taken me three months of research, writing, modeling, and painful organization. It incorporated many variables at once, and frankly, it did the job better than I could have done it before.
Faster and better.
Scary.
And to cap it all off, the kind of advice and support I am receiving would once have cost tens of thousands of dollars. Instead, I get access to the great, expanding universe of AI for a monthly subscription fee that costs a little more than Netflix. That is genuine value, especially for young people, entrepreneurs, and anyone just starting out.
What a massive advantage.
I worked for a time with a young group on an AI business, and I am convinced they ran much of the company and its strategy through AI. How else could such young and inexperienced people do so much so fast? AI has enabled the hungry and the determined to do far more than their unaided human potential would have allowed. And again, we are only in the early stages.
I cannot really tell you where AI will go. I can only repeat what I have heard, seen, and increasingly believe: it will become far more capable. It will help run complex systems. It will automate enormous amounts of coordination, communication, analysis, and decision support. It will not just do some things faster than humans. It will do many things with a level of speed, memory, and consistency that no human bureaucracy could match.
For those saying AI will not replace professions, I think they are underestimating the speed of change. AI will replace, compress, or radically alter a large share of mental work currently done through screens, documents, spreadsheets, analysis, and communication. I think this has already started to affect lawyers and accountants, and it will affect financial professionals too.
The greatest impact will be felt by professionals who do not adopt AI systems to amplify their own efforts, increase their capabilities, reduce their costs, and engage more customers. The ones who embrace AI will become more powerful. The ones who do not will be left behind.
But AI has an Achilles heel: energy.
We live in a society dominated by our consumption of energy, yet we massively take it for granted. People simply do not appreciate what modern society becomes without abundant, reliable power. Think about a hospital running on diesel generators. How many MRIs will that energize, and at what cost?
We are embarrassingly rich in Manitoba because we have an abundance of hydroelectric power. We have some of the lowest-cost electricity in North America. In many other places, people have to make hard choices because energy costs have risen dramatically.
Some of that is because populations have grown and demand has increased. Some of it is because energy policy has become confused. We have pushed intermittent and unreliable sources like wind and solar without always being honest about the cost of backup power, transmission, storage, and grid reliability.
If people truly cared about reducing carbon while preserving modern living standards, they would take nuclear energy far more seriously. The refusal to do so makes parts of the environmental movement look far less scientific than it claims to be.
Unfortunately, there may be no stopping global warming in any simple sense. We will have to prepare for it, adapt to it, and build serious infrastructure around it.
But I digress.
The point is this: energy costs are going up, and guess what loves to hog energy?
AI.
So naturally, one has to wonder: how will we pay for it?
There are proposals for badly needed data centres that will provide the computational muscle AI needs to keep meeting relentless demand. But that muscle, like any other, needs energy to function. Data centres may increasingly be paired with nuclear power, including SMRs, or small modular reactors. Yet many nuclear projects still face intense public resistance because of fear, politics, cost, regulation, and the long shadow of rare but unforgettable accidents.
Nuclear energy is among the safest forms of large-scale power ever developed. It is dense, reliable, and carbon-free. It should be central to any serious conversation about AI, energy, and climate. But instead, too much of the conversation gets swallowed by slogans and panic.
So again, back to the point: AI needs energy. Energy costs are rising. Therefore, the cost of AI is likely to rise too.
The demand for AI is already so intense that it is not hard to imagine a future where computational power is effectively rationed by price. The organizations that can afford the best models, the most compute, the most reliable access, and the most integrated AI systems will be the big winners. Companies that cannot adapt, or cannot afford to adapt, may become obsolete quickly.
This also means many universities, small companies, independent builders, and young people pioneering the AI economy could end up with less access over time, not more. AI feels cheap today because we are still in the early phase. But the infrastructure behind it is not cheap. The spending commitments being made by corporations for AI, data centres, chips, and energy infrastructure are enormous. They are in the trillions.
Someone will have to pay for all of that.
And I am quite sure it will not continue to be covered by my monthly subscription fee at anything close to the level it is today.
The economic rationalization of AI will come upon us as quickly as AI itself came upon us. Perhaps that will slow some of the social upheaval. But probably not.
The wealth and income gaps we already live with may soon be stacked with another divide: asymmetric access to the knowledge, tools, and capabilities of AI.
The rich and powerful will have more of it. The poor and the small will have less. The people and companies who can afford intelligence on demand will move faster, think faster, and build faster.
That could play out in dangerous ways. I hope it does not. But we shall see, and soon.
Romel Dhalla, is President of Dhalla Advisory Corp., provides strategic corporate finance advice to companies and high net worth individuals. He was a portfolio manager and investment advisor with two major Canadian banks for 17 years. Contact him at romel@dacorp.ca. Any views or opinions represented in this article are personal and belong solely to the author and do not represent those of people, institutions or organizations that the author may or may not be associated with in professional or personal capacity, unless explicitly stated.