Searching for the King of Kings

Christopher Nolan’s The Odyssey is full of stunning visuals.

But there is an image that I find unforgettable.

Agamemnon.

Whenever Agamemnon appears, he is the embodiment of absolute power. The king of kings. Imposing, reticent, magnificent. He has conquered Troy, defeated its armies and returned with the spoils of the greatest war of his age.

But his victory is already beginning to unravel.

It is a useful image for thinking about Quentin Solt's latest essay on Elon Musk and the AI wars.

I have followed Quentin’s arguments with interest, in light of our #TrojanHorseDebate. Quentin has a real talent for peering beneath the obvious layer of a technological system and asking where the economic power actually resides.

But I think there is a recurring problem in his analysis: he keeps looking for the king.

At some point Harvey and Legora were the Trojan Horse (and, implicitly, the winners). They would enter law firms, and, in the process, help commoditize the very expertise on which the firms depended.

Then the focus shifted to the foundation models themselves. Claude and ChatGPT would increasingly absorb the functionality of specialist applications and intermediaries. The Trojan Horse was no longer Harvey or Legora; it was the model sitting underneath them.

Now we have moved another layer down. In the latest essay, the model race itself is described as almost a distraction. The real contest (and the one that matters) is infrastructure: energy, compute, chips, distribution and the physical systems required to run AI at scale. Musk, Quentin argues, may not need to win the model race if he controls the infrastructure on which everyone else depends.

There is a lot that is persuasive here.

But there is also a conceptual problem.

A bottleneck is not necessarily a moat. And a moat is not necessarily a kingdom.

I have made this point before, but it is worth repeating: progress has a habit of moving the bottleneck.

Compute becomes more efficient. Models become smaller. Hardware changes. New sources of energy appear. Open systems improve. Capital floods into whatever has become scarce. What looks like an indispensable resource at one point in the technological cycle can become abundant at another.

Quentin himself recognizes many of these possibilities. His latest essay discusses efficiency gains, DeepSeek, alternative compute, Google, China and the possibility that artificial general intelligence arrives before some of the infrastructure now being built is fully deployed.

Those are not peripheral considerations. They are precisely why I am skeptical of turning today’s infrastructure constraints into tomorrow’s permanent economic hierarchy.

And this brings me back to the Trojan Horse.

The Trojan Horse was real. Troy really did fall.

But the interesting thing about the Trojan War is what happened after the apparent victory.

After ten years of warfare, with all the help from Athena herself and a host of other immortals, the victorious Greeks did not inherit an eternal kingdom.

Odysseus spent ten years trying to get home to Ithaca. Agamemnon returned to Mycenae and was murdered by Clytemnestra and Aegisthus.

The victory at Troy was real. It simply was not the end of history.

That is the lesson I think we should apply to AI.

I am not arguing that infrastructure does not matter. It matters enormously. Nor am I arguing that Musk will not be one of the great beneficiaries of the AI infrastructure buildout. He may well be.

I am questioning something more subtle: the repeated temptation to identify the current locus of scarcity and then extrapolate it into a permanent source of power.

The application layer was going to win.

Then, the foundation model layer was going to win.

Now infrastructure is going to win.

Perhaps.

But if technology has taught us anything, it is that the thing everyone is scrambling to own today is often precisely the thing that engineers, entrepreneurs and enormous amounts of capital will spend the next decade trying to make less scarce.

This is also why I find the Musk argument more interesting than the usual “who wins AI?” debate. Musk’s extraordinary record gives us very good reasons to believe that he can build things other people cannot. It does not, by itself, tell us that the things he is building today (even if we believe that he "playing tomorrow's hand") will remain the permanent bottlenecks of the AI economy.

And perhaps there is a final irony here.

Greek myths are grounded in a pagan worldview. Human society, mirroring the Olympic structure, is a pyramid (just like a traditional law firm). This is a world in which power is inherited, contested and violently lost.

Quentin, as a Christian, knows better than most of us that no kingdom of this world is permanent.

Agamemnon was, quite literally, the king of kings.

He was magnificent.

He won the war.

And then he went home and died.

I am not saying we should stop looking for the king.

My point is that we should be rather careful about assuming that, having found a king, we have found the kingdom.

P.S. One thing I should expressly acknowledge: Quentin Solt is asking an important question. Namely: what does the alluring promise being dangled in front of us by AI actually cost?

His follow-up post actually takes the metaphor one step further and casts Elon Musk as the Serpent. As fond as I am of biblical references, that might be a leap too big for my imagination (especially with Sam and Dario vying for the title). But even taking Quentin's analogy at face value, I will say this: we should (absolutely) be thinking about the cost of the apple, but it does not tell us who will own the orchard.