AI boom: Conjecture or Uncertainty?

Every major technological revolution follows a predictable ecological rhythm. This article covers how the irrational exuberance of the modern AI landscape mimics that same pattern.

Cycle of succession

When a patch of land is cleared(say by fire, landslide or volcanic eruption), it is a harsh environment, a somber state, Lichens are the first ones to arrive, Lichens are hardy opportunists, they survive on rocks, not only do they survive, they breakdown the rocks in this barren land into mineral soil(sand and silt), which has some essential minerals necessary for basic survival, the basic nutrients such as Potassium(K), Phosphorus(P) and Calcium(Ca). These opportunists do something nobody else asked them to do, in conditions nobody else can survive, using mechanisms(breaking rocks) that benefit only the future inhabitants they never meet. Lichens create conditions for value, without capturing that value themselves.

This is where things get interesting now. The minerals that were locked in the crystal lattice of stone for millions of years are now openly available. Now the wind and dust, carrying seeds of Fireweeds and Grasses, these seeds now have access to this mineral-rich soil, they play the volume game, it is just leaves and seeds, the essentials. These species don’t invest in wood or deep roots. This specific selection lets them grow and spread really fast. One may think that this growth is just on the surface; under the surface, it lacks roots. That is not the entire picture; the biological purpose of fireweeds is to make the ground fertile for something they will never become. They succeed by being temporary. They have short half-lives; they die out fast. And their decaying bodies leave behind a layer of organic carbon(humus).

While the land is crowded with such Fireweeds and Grasses, a tiny seedling of Douglas Firs arrives, and it looks unimpressive next to the towering full bloom of Fireweed. These seedlings have slow initial growth, as they invest in wood and root development, which takes time. It takes decades before canopy formation, but once these seedlings develop, they last fr generations to come.

Every major technology deployment follows this succession pattern; the question is never about whether the fireweeds will appear or clear, they always do, capital flow ensures it. The question is whether you are standing in the clearing watching the fireweed burn, mistaking their growth for permanence. And whether the capital you are deploying is Fireweed capital or Douglas Fir capital, because they require completely different soil conditions, timelines, and definitions of success.

Irrational exuberance with novelty

Investors usually try to assess the intrinsic value of assets, then make decisions around those assessments. Bubbles aren’t caused by the disruption potential of the underlying technology or financials; they are an outcome of the irrational exuberance applied to those developments. The changes in the price of any asset, as compared to the value of the underlying asset, are a result of collective changes in the investor psychology, governed by the forces around them.

Now, why do these bubbles get inflated so much? How come the investor psychology inflates the underlying potential of the “new” technology so much that the price starts deviating greatly from the value? The devil is in the word “new”, because the technology is new, there is no limit to its imaginative potential, as there is no history to restrain the imagination. That is why we usually see all the Tech bros espousing grand visionary narratives. They rarely talk about rational assessment of the value created by their tech; they like to give hour-long podcasts on why this is the next big thing(sometimes wearing black turtlenecks). And these limitless futures justify the sky-high valuation of underlying assets or technologies.

There is only one way to put limits on this awe, and that is the knowledge of history.

And what does history teach us? Bubbles are necessary; if people remained prudent, patient, and analytical, novel technologies would take decades to be built out. It is the very hysteria(the irrational exuberance) that causes these processes to be compressed into a very short period, with some of the money going into life-changing investments in the winners, but a lot of it being incinerated. A bubble has both technological and financial aspects. Those who crave technological progress are perfectly happy to see the investors lose money in their interests. The investors, on the other hand, would like to be a part of this technological progress but have no desire to incinerate their capital to help bring it about.

Succession Matrix: Capital, Tech, and Ecology

There are tons of capital flowing into the AI economy. But my problem is not with the capital flowing in; it is more about the intent and nature of that capital.

Uncertainty: A scenario where the future outcome is unknown, and it is impossible to calculate or assign probabilities because there is no precedent, no historical data, and the parameters themselves are unquantifiable (unknown unknowns).

A conjecture is when there is no data, no precedent, and no way to calculate probability. You form a structured conjecture using first-principles thinking, structural logic, and qualitative frameworks to outline how a new market or asset class might evolve.

When investing in AI, there is no historical baseline; on top of that, the Venture heuristics and financial models break down because the cost of compute is fluctuating dynamically, the commercialization timelines are unprecedented, and regulatory and governance guardrails are missing. It is hard to calculate the probability of a frontier AI lab achieving positive unit economics in five years because the structural parameters of the market change quarter by quarter. There is a data vacuum for traditional financial underwriting.

Because investing in deep-tech AI infrastructure sits in true Uncertainty, investors cannot rely on quantitative models. Instead, they must win through Conjecture, crafting a sophisticated, logic-driven investment thesis. To underwrite an AI deal, you must use first-principle thinking to map out structural vectors, for ex. If compute efficiency scales at $X rate, and enterprise trust requires local sovereign hosting, then the value will accrue to the middle layer (say, chip design or EDA) rather than the foundational models.

Now, applying this to debt financing, debt(unlike equity) has a capped upside but a massive downside. Equity investors (VCs) take conjecture bets because if the story comes true, they get 100X their money.

Debt lenders only get their interest back. Therefore, a debt provider should never fund a project based on conjecture, because they aren’t paid enough to take that level of unknowable risk

And it is really easy to confuse the two. When blinded by hype, even the most astute of capital allocators can mix the two up. If you are in this boat at the moment, take a pause and reflect. That financial model that you are so confident of, is it actually supporting the investment, or are you just speculating? That is the toughest thing to do in the world, to take an objective(neutral, non-conformist) stance when the whole world is going gaga over the Trillions just beyond the horizon.

Sources:

https://www.oaktreecapital.com/insights/memo/is-it-a-bubble

https://stratechery.com/2025/the-benefits-of-bubbles/

Chen, J., Blume, H. P., & Beyer, L. (2000). “Weathering of rocks and neogenesis of minerals associated with lichen activity.” Catena, 39(2), 121–146.

Carlota Perez’s Technological Revolutions and Financial Capital(2002)

Robert J. Shiller’s Irrational Exuberance (2000)

Frank Knight’s Risk, Uncertainty, and Profit (1921)