Why the AI revolution breaks all the old rules about consolidation
Past industrial revolutions have all progressed through predictable evolutions; AI likely will be very different.
For more than a century, new industries have followed a predictable evolutionary arc. A breakthrough technology appears, an explosion of new entrants follows and then consolidation inevitably takes over.
As examples, railroads once numbered in the hundreds before merging into a few dominant networks. The early U.S. automobile industry saw thousands of manufacturers reduced to a handful. Airlines, retailers and even the early Internet where Amazon absorbed or displaced countless smaller resellers, all matured through the same pattern – expansion, then contraction.
This path wasn’t accidental. It was structural.
Why industries consolidate: The weight of cost structures
In every previous industrial wave, scale conferred efficiency. Building railroads required massive capital and fixed infrastructure. Automobile manufacturing demanded factories, distribution networks and supply chains. Airlines needed hubs, fleets, maintenance facilities and logistics systems. As these industries matured, the companies with the largest capital bases, the most integrated operations and the greatest economies of scale survived. High fixed costs and infrastructure intensity drove consolidation.
In other words, the cost structure – heavy, centralized and incredibly expensive – rewarded size. Efficiency was inseparable from scale.
However, artificial intelligence breaks this logic.
AI does not follow the industrial rulebook because its most essential input, knowledge, is not a finite resource that scales linearly with investment. The traditional argument says that AI will consolidate because larger companies have more data, more computing power and more engineering talent. But this view is anchored in yesterday’s economics.
AI systems are increasingly being built with the assistance of AI itself. The knowledge required to design, tune and deploy AI models is becoming more democratized, not concentrated. Tools that once required elite expertise are coming within reach of any capable developer or even a small team because AI accelerates learning, experimentation and iteration.
Even more importantly, the underlying computing resources needed to run AI are broadly accessible. Unlike railroads or automobile factories, computing capability does not require ownership of multi-billion-dollar plants. Anyone can rent access; anyone can scale up or down on demand. Imagine if, during the rise of automobiles, every aspiring entrepreneur had free or inexpensive access to a fully equipped assembly line. The number of car companies would have been vastly larger. That is the world AI is entering today.
The economic pressures that once forced consolidation simply do not apply in the same way.
Efficiency will solve the compute problem
Today’s conventional wisdom insists that if the world wants advanced AI, it must invest trillions of dollars into new data centers and infrastructure. The narrative is that compute must grow, and only the largest technology companies can fund such expansion, ushering in an era of massive consolidation.
But this logic overlooks a key dynamic. AI is not just a consumer of compute, it is a generator of efficiency. Every year, models become cheaper to train and cheaper to run. Every breakthrough in architecture, sparsity, quantization, routing, distillation or optimization reduces the compute footprint required to deliver high-quality intelligence.
AI itself will help solve the compute problem that AI creates.
This is the ironic twist the industry has not fully internalized. While current forecasts assume a future locked into capital-intensive infrastructure, the more probable trajectory is that AI efficiency accelerates so quickly that the need for trillion-dollar data-center expansions diminishes. The market will not stand still waiting for giant buildouts. Instead, it will innovate around the bottleneck.
Next: An expansion of AI alternatives
Because the opportunity is enormous and the barriers to entry are lower than in any prior industrial revolution, alternatives will emerge, and quickly. Here’s what’s likely over the coming months and next several years.
In short, the competitive landscape of AI will expand, not contract.
AI breaks the pattern
AI is the first major industry in modern history where the forces pushing toward fragmentation may outweigh the forces pushing toward consolidation. Because computing capability is rentable, knowledge is leverageable and efficiency gains are compounding, we should expect a broad, diverse and sustained ecosystem of AI developers, not just a small handful of mega-firms dominating the field.
Every previous industrial revolution was defined by heavy infrastructure, capital intensity and consolidation. This one will not be. AI is the anti-pattern, and that is what will make this era unlike any other in the history of technology.
Jan Sevcik, FACHDM is co-founder and CEO at Medical Search Technologies, which specializes in assessing AI for healthcare applications.