AI, Markets, and the Cost of Progress
A few weeks ago, we introduced our Disruption Innovation & Growth+ (DIG+) strategy. In that piece, we framed artificial intelligence not as a passing technology cycle, but as the latest and most powerful chapter in a much longer process of digital transformation. One that is reshaping how decisions are made, how systems are controlled, and ultimately where durable economic value tends to settle.
When we introduced DIG+, we also made a commitment. That commitment was not simply to observe this shift, but to help our clients navigate it with clarity and confidence. Not just to survive what is shaping up to be a tectonic economic change, but to position themselves to thrive through it.
We said we would bring a clear-eyed, ongoing focus to this area. To separate signal from noise. To return to first principles when narratives get loud and markets move quickly.
Before getting into it, a brief warning. I have probably been, at times, rightly accused of being an incorrigible contrarian. In my experience, that label tends to appear when widely held assumptions go unexamined for a little too long.
In that spirit, this piece takes an alternative view on three ideas that have quietly become sacred cows in the current AI conversation:
🐄 That AI behaves primarily like software rather than infrastructure
🐄 That productivity gains automatically translate into higher margins and earnings
🐄 That the current AI infrastructure build is speculative excess rather than deliberate moat building
We think all three deserve a more careful look.
This note is the first in that series.
AI did not arrive out of nowhere
It is easy to talk about AI as if it suddenly appeared last year. In reality, it sits atop decades of prior change. Digitization. Connectivity. Cloud infrastructure. Automation. Each wave promised efficiency. Each wave delivered it, along with new forms of dependency and concentration.
AI differs in scale but not in structure.
My view is that it is best understood less as a tool and more as an economic operating layer. One that depends on physical systems as much as on digital ones.
That distinction matters because it shapes who ultimately benefits.
This also helps explain certain investment decisions we have made in line with the DIG+ framework. One example is our decision to make a significant private investment in transformer manufacturing. As intelligence scales digitally, constraints increasingly manifest in the physical world. Electricity must be generated, stepped down, conditioned, and delivered reliably before a single line of code can run. That layer is essential, even if it rarely captures attention. In periods of rapid technological change, durable returns have often accrued not to the applications that attract headlines, but to the companies that enable the system to function at scale. In a gold rush, owning the tools everyone depends on can matter more than chasing the gold itself.
AI feels like software. Its economics are physical.
A lot of the optimism around AI assumes software-like behaviour. Falling costs. Easy scalability. Broad margin expansion.
That assumption deserves scrutiny.
Modern AI depends on scarce, physical inputs. Specialized chips. Purpose-built data centres. Reliable access to large amounts of electricity. And, increasingly, dependable access to water for cooling and power generation. In many regions, power and water, not capital or demand, are now the binding constraints.
When something is capital-intensive and supply-constrained, it does not behave like an app. It behaves like infrastructure.
And infrastructure has its own rules.
Why AI feels cheap right now
Many clients have commented on how inexpensive, or even free-feeling, AI tools seem today. That is not accidental.
We are still in the adoption phase. Providers are encouraging experimentation and absorbing costs in order to embed AI deeply into workflows. We have seen this pattern before. Cloud computing. Mobile data. Digital payments.
But this phase does not last.
Once AI becomes embedded in critical processes, the economics shift. Access moves from being encouraged to being allocated. At that point, users either pay for priority or they wait.
That transition is structural. Not cyclical.
Where pricing power actually sits
This is where I think the conversation often becomes too neat.
The strongest pricing power in the AI ecosystem does not sit evenly with everyone who uses the technology. It sits with those who control scarce capacity. Compute. Power. Water secured sites. And with those who sell what is required to build and equip them.
Capital intensity creates leverage. It always has.
That does not mean AI users will not benefit. It does mean that the economic surplus will be shared, and in some cases contested, rather than freely passed through.
Productivity gains are real. So are the trade-offs.
There is no question AI will improve productivity. Some costs will come down. Some decisions will improve. Some processes will disappear altogether.
What tends to get missed is that productivity gains from infrastructure rarely come without new costs elsewhere. What usually happens is substitution. Internal costs fall. External costs rise. Over time, access is priced in line with the value being created downstream.
That is not pessimism. It is precedent.
Which is why I am cautious about the idea that AI adoption automatically translates into broad, sustained margin expansion across markets.
What this means for markets right now
Markets are currently trying to price long-term potential and near-term earnings at the same time. That is difficult, and it helps explain the volatility we are seeing.
Some businesses, particularly those that own infrastructure or sell AI-enabled products, will do very well. Others will see benefits offset by higher technology costs. And some will find that AI changes who they pay more than how much they earn.
This kind of dispersion is exactly what we would expect at this stage.
Why this matters to DIG+
When we think about DIG+, this is the backdrop we have in mind.
Our work here is not about chasing what is fashionable or reacting to every new headline. It is about helping our clients make sense of a large, complex, and still-unfolding change. One that will reshape how capital is deployed, how value is captured, and how risk shows up over time.
We are less interested in short-cycle excitement and more focused on where control, trust, and execution tend to concentrate as systems become more autonomous, more powerful, and more dependent on scarce physical resources. Power. Compute. And water.
AI is not a free lunch. It is a powerful capability layered on top of expensive infrastructure. That reality will shape outcomes in ways that are not always obvious in the moment.
This series is part of our commitment to you. To stay clear-eyed. To separate signal from noise. And to help you not just navigate this shift, but position thoughtfully within it, with the benefit of time, perspective, and discipline.
As always, our goal is not prediction for its own sake. It is preparation, and ultimately, better long-term decisions on your behalf.
Grant Colby
Managing Director & CEO
