Every generation of technologists gets to live through at least one moment that feels like the ground shifting beneath them. The trick is recognizing it for what it is while it is happening, rather than years later when the conclusion is obvious to everyone. When I first encountered artificial intelligence, my immediate reaction was that this was not a product at all. It was something categorically different — a character change.
Products Versus Character Changes
It helps to be precise about the distinction. The technological milestones we tend to celebrate were, for the most part, product-oriented revolutions. The personal computer was a revolution. The smartphone was a revolution. The internet was a revolution. Even 3D printing, in its own way, qualified as one. In each of these cases, the early adopters were genuinely thrilled by a piece of new technology. But strip away the excitement and, at the end of the day, what they were excited about was a product — a thing you bought, held, or used.
AI is different. For the first time in decades, the change is not about a new device or a new platform you adopt. It is about the nature of what is being delivered. AI is providing intelligence to the masses. That is not a feature of a product; it is a shift in the underlying character of what technology can do. The last time something of comparable weight arrived was the internet, and even that was ultimately a means of connection rather than a source of cognition.
The Total Addressable Market Is Mankind
Once you frame AI as the distribution of intelligence rather than the sale of a product, the implications become almost dizzying. The natural next question is simply: what else can this do? And the honest answer is that the range of uses appears endless. Intelligence is a general-purpose input. It does not belong to one industry, one demographic, or one use case. When you try to draw a boundary around the total addressable market for something like that, you find there isn't one. The TAM is mankind.
That realization is what made me willing to bet heavily on the space. If a technology's potential reach is the entire human population and nearly every task they perform, then the conventional ways of sizing an opportunity stop applying. The downside of underestimating it is far larger than the downside of leaning in.
Picking the Themes That Benefit Most
Conviction in a thesis is only the starting point. The harder work is translating it into specific exposure. Rather than trying to own the abstraction of "AI," I started by identifying the thematics I wanted exposure to — the areas I believed would be the biggest beneficiaries of the shift. From there it became a matter of finding the companies positioned at the center of those themes.
One of those was Palantir, which I was fortunate to find early. Another was Nvidia. I had mentors who pointed me toward Nvidia well before it became a household name, though they described it to me as a gaming chip company. The crucial insight was seeing past that label. There was no reason its chips had to remain confined to gaming. The same hardware could power data centers, and from that recognition the rest of the trajectory unfolded. Spotting the transition before the market fully priced it in is what separates an early position from a crowded one.
The One Thing I Got Wrong
I had a deep comfort that AI would prove bigger than the revolutions that preceded it, precisely because it wasn't a product. That conviction held up. But conviction in the destination is not the same as accuracy about the timeline. The single thing I did not anticipate was the insatiable demand for AI arriving this quickly. The appetite for intelligence, once it became available, did not build gradually — it surged.
That gap between being right about the magnitude and being surprised by the speed is itself a lesson. When a genuine character change arrives, the temptation is to assume it will diffuse at the measured pace of previous product cycles. It doesn't. The very thing that makes it a character change — its relevance to everyone, everywhere — is also what compresses its adoption curve. Recognizing the shift early matters, but bracing for how fast it can move may matter just as much.