SpaceX, Anthropic, and the Opportunity Cost of Knowing

Single moments of commitment can define your life, sometimes with a price.

On Thursday June 12th, SpaceX will price the largest initial public offering in the history of capital markets, somewhere in the region of $1.77 trillion. By October, Anthropic will likely follow it past a trillion of its own, having filed quietly while sitting on a $965 billion private valuation and a revenue run rate that did not exist in its current form three years ago. For most people these are incredible numbers passing overhead, the way the price of a painting one will never own passes overhead. For a much smaller group they are the confirmation of a conviction held early and acted upon. And for a stranger third group, the one I belong to, they are the confirmation of a conviction held early and not acted upon, which is a different thing entirely, and the more interesting one to sit with.

In 2018 I was deep in the international relations modules of my doctorate, writing about the weaponization of space through the development of anti-satellite systems, and I was introduced, through someone working in Elon's office at Tesla, to the possibility of a role at SpaceX. I want to be precise about what I understood at the time, because the temptation in retrospect is always to grant oneself less foresight than one had, as a kind of consolation. I understood that space was not a frontier in the romantic sense but an infrastructure layer in the geoeconomic one, that the same logic of high ground that governs terrestrial conflict was migrating upward into orbit, and that whoever owned the communications and navigation layer would own something closer to a utility than a product. I knew this clearly enough to be writing about its military dimension. What I could not do was act on it, because I was on an F-1 visa and the International Traffic in Arms Regulations close the entire sector to non-citizens before any question of merit or fit arises. The door was not one I declined to walk through, in fact. Rather, it was a door that had been bricked over before I reached the hallway. But the knowledge was there, and it has now been assigned a number with (a great) many zeros after it.

The second instance is the one that actually implicates me, because in that case the door was open and I chose not to walk through it. In 2023, conversations with people at Anthropic began to take on a more serious texture. By then my reading of the situation was, if anything, stronger than it had been with space. I had followed the scaling work closely enough to hold beliefs about where the capability curve was heading, and more importantly I understood the strategic shape of what Anthropic would, in all likelihood to me at the time, become. That is a thesis about knowing and trust as the binding constraints on adoption, and I believed it. What stopped me was not doubt. It was that I had already made a different commitment, to Los Angeles, to building a life and a set of relationships and eventually my own equity here, and Anthropic at that moment was beginning its slow institutional drift back toward in-person work in the Bay Area. To pursue it seriously would have meant unwinding a prior bet whose value I trusted even though it was not yet visible to anyone, including me. So I let the conversations lapse, and last week's confidential filing has now put a number on that decision too, denominated in the same currency, with a similar quantity of zeros.

The reflexive frame for all of this is regret, and regret is both the wrong instrument and the less interesting one. It is wrong because it presumes a single axis of value along which the decision can be scored, and the entire texture of our wonderful lives is that it is lived along several axes at once, most of them not denominated in equity. We know from the behavioral literature that human beings do not optimize against one utility function in the way the older economics assumed; we are loss-averse and gain-seeking by turns, we weight the vivid over the probable, and we make our largest commitments not at the clean forks where the costs are legible but in the quiet execution of earlier decisions whose implications surface only later. The person who turned down these rooms was not blind to the opportunity. They were weighting something else more heavily, and the only question is whether that weighting was sound, and whether it continues to be.

This is the part worth dwelling on, because it cuts against the founder-mythology version of conviction, in which the visionary sees what others cannot and is rewarded for the seeing. The seeing is necessary but it is nowhere near sufficient. The genuinely rare condition is not ignorance and it is not prescience; it is the experience of knowing something to be true, holding that knowledge without distortion, and nonetheless organizing one's life around a different truth. The best secrets, in the sense that matters, are not the ones hidden from everyone. They are the ones in plain sight that almost no one will act upon, and the reason almost no one acts is that acting requires the dissolution of some prior commitment, and prior commitments are the most expensive things to dissolve precisely because their value has not yet been confirmed by anyone else. To act on a secret early is to pay, in advance and alone, a cost that the market will only later agree was worth paying.

What this means is that an opportunity cost of the SpaceX or Anthropic kind is rarely the record of a fork declined. It is almost always the implied price of a commitment kept. I held a thesis about Los Angeles before the city's gravity was obvious, and the cost of that thesis is now partly denominated in the equity of two companies I did not chase further. I hold a thesis now, in the deployment of my own capital, about the layer of instruments that make opaque systems legible and therefore ownable, the assay technology rather than the ore, and the same enormous public numbers that price my earlier abstentions are also, read correctly, the late confirmation of exactly that pattern: that the world will pay almost any sum for the thing that renders a previously illegible domain transparent enough to own. A $1.77 trillion valuation for the company that made orbit into infrastructure, and a near-trillion-dollar one for the company making machine cognition readable and governable, are not two separate miracles. They are the same proposition, confirmed twice, late.

So, I do not read these listings as a verdict against my judgment in 2018 or 2023. And many who are scolding themselves similarly should not either. A public price is not a retrospective scolding; it is information, and the only information it actually contains is the cost of the commitment I chose to keep instead. That cost is real and I have no wish to soften it. But the knowledge underneath it has held in both directions now, and knowledge that holds is the only thing that converts, eventually, into the kind of action that generates depth and growth.

The discipline the moment demands is not the reversal of old decisions, which would be its own form of unseriousness, but the narrowing of the distance between knowing and acting on the things that remain early, the things not yet priced, the instruments being built right now by people the public markets have not yet learned to value. Thursday's bell, and October's after it, are confirmations of patterns already complete.

And there is great beauty, indeed a lifetime of it, to know that the only moment that has ever mattered is the one before the confirmation arrives.