The AI Winner Might Not Be the One With the Best Model
What Apple’s strategy reveals about where AI power really lives
We have been conditioned to hunt for power in the wrong place. In the current AI arms race, the prevailing logic is one of “compute-determinism” which is the belief that the firm with the largest training run, the most expansive budget, and the tallest bars on a benchmark chart will inevitably own the future. It is a tidy, persuasive, and likely incorrect theory.
Apple’s recent maneuvers to use Alphabet’s (Google) proprietary model in their devices suggest a different reality. By treating frontier models as modular inputs and as components to be licensed, they (Apple) are demonstrating where the true leverage in the age of intelligence resides.
It lives in the ownership of where attention is captured.
The Economics of Capture
The AI market is exhibiting the classic symptoms of a capital-intensive industry in its “theatre” phase. Enormous sums are being deployed to prove technical superiority, yet the economics of capture are migrating elsewhere. While frontier models are undeniably improving, they are also converging. The gap between “state-of-the-art” and “good enough” is narrowing faster than the costs of achieving the former are rising.
For the “Model Labs,” the math is becoming structurally unforgiving with exponential increases in capital expenditure ($CapEx$) that are yielding linear gains in utility and its prices collapsing.
Apple understands that, and it knows that who owns attention, defaults, and trust ultimately wins.
The Path of Least Resistance
By embedding Alphabet’s model Gemini at the operating-system level, Apple is positioning AI as infrastructure. Once intelligence is surfaced through system-level defaults within its notifications, voice assistants, and wearables, user choice shifts from explicit to implicit. Behavior follows the path of least resistance.
This is where leverage compounds. At this layer:
The Model Recedes: It becomes a “back-end” concern and is replaceable, negotiable, and increasingly invisible.
The Relationship Remains: The entity that governs the interface retains the data, the identity, and the “platform tax.”
Risk is Decoupled: Capital-heavy bets on frontier capability carry massive exposure to cost inflation. Conversely, those at the interface benefit as intelligence cheapens and their margins improve without a proportional increase in reinvestment.
From Invention to Intermediation
The market remains fixated on the “research lab” model of success because breakthroughs make for better headlines. But the history of computing suggests that power eventually settles with the System Architects. The victors will be those who act as the governors of how that power is applied, constrained, and trusted.
The question for investors is no longer “Who has the best weights?” but rather, “Who controls the moment the user reaches for help?”
The Bottom Line
The history of technology is a graveyard of innovators who built the foundation only to be eclipsed by the gatekeepers who built the toll booth.
We are entering the “Commodity Intelligence” phase and in this world, the raw ability to reason is simply utility, like electricity or bandwidth.
So, if you do not own the interface, you are simply an upstream supplier and suppliers are always the first to be squeezed.
All the zest, 🍋
Cien