There Will Come a Time When All of This Seems Small

This Seems Small

8:49

SpaceX hit $3 trillion this week.

Let that sit for a moment. Not the number itself — numbers like that are hard to actually feel. But the context around it. Four days ago, SpaceX priced its IPO at $135 a share, raising $75 billion in the largest public offering ever recorded. By Tuesday it was touching $225. The market cap crossed $3 trillion before the ink on the prospectus was dry.

Anthony Pompliano put it simply: Elon Musk made more money in the last 24 hours than Warren Buffett made in his entire lifetime.

That's not hyperbole. That's arithmetic.

And here's the thing about that sentence that almost nobody is saying: it will seem like a modest footnote in ten years.

We are living inside what will eventually be described as the early innings. The seed round. The version 0.1 moment of the largest economic transition in human history — and because we're inside it, it feels enormous. The numbers feel staggering. The velocity feels unsustainable. But consider what's actually on the table.

SpaceX at $3 trillion is already the second-largest company on earth. OpenAI is filing for its IPO at a target north of $1.75 trillion. Anthropic — which just overtook OpenAI in private market valuation at $965 billion — is right behind. Three companies that barely existed in any meaningful commercial form five years ago are collectively approaching $6 trillion in combined market value before most of them have even started trading.

Goldman Sachs is projecting $765 billion in annual AI capital expenditure in 2026 alone. Growing to $1.6 trillion per year by 2031. McKinsey puts the global data center build-out at $7 trillion by 2030. This isn't venture capital optimism. This is infrastructure spending — the boring, physical, steel-and-concrete kind that happens when industries are already certain about where they're going. The certainty is the tell.

There's a specific kind of moment in technology transitions that historians can identify clearly in retrospect but is almost impossible to recognize from the inside. It's the moment when a technology stops being an experiment and becomes infrastructure. When the question shifts from "will this work?" to "how much do we need?" We just crossed that line.

The IPOs didn't signal a peak. They signaled institutionalization. When SpaceX and OpenAI and Anthropic go public, they stop being bets on the future and become positions in the present. Pension funds hold them. Index funds hold them. Sovereign wealth funds hold them. The capital that flows into those positions isn't speculative — it's allocative. It's the signal that the world's largest pools of money have decided this is real infrastructure, not an experiment. That's the same signal that preceded the buildout of the railroad networks. The electrical grid. The interstate highway system. The internet backbone. Every time, the numbers at the moment of institutionalization felt enormous. Every time, they turned out to be the small ones.

Here's what the insightful observer notices right now, while everyone else is staring at the $3 trillion headline: The compounding hasn't started yet.

Anthropic published internal data showing Claude went from a 3x research speedup to a 52x speedup in under a year. 80% of their production code is now written by Claude. That's not a trend line — that's an acceleration curve. The model that exists six months from now will be built faster by the model that exists today. And the model six months after that will be built faster still.

We are in the phase where each generation of AI capability is being used to compress the timeline to the next generation. The gap between frontier capability today and frontier capability in 2028 isn't linear. It's not even exponential in the traditional sense. It's compounding on compounding — recursive improvement applied to recursive improvement.

The $3 trillion SpaceX is the result of two decades of compounding in rockets and satellite infrastructure. The $6 trillion in combined AI market cap entering public markets this summer is the result of three years. What does the AI infrastructure layer look like after two decades of compounding at these rates? Nobody knows. And that's precisely the point.

The opportunities hiding inside this moment are not where most people are looking. Everyone sees the frontier models. The headline valuations. The IPO roadshows. Those are real, but they're also the most competed-for positions in the market. You're not going to outmaneuver sovereign wealth funds and hedge funds for shares in OpenAI on day one. What's less competed-for is everything the frontier needs in order to function.

Power.

The AI infrastructure buildout requires electricity at a scale that existing grids cannot immediately provide. McKinsey projects $5.2 trillion in AI-related data center infrastructure by 2030. That infrastructure needs power, and right now there is a significant gap between what frontier compute demands and what utilities can deliver. The people bridging that gap — on the site acquisition side, the utility coordination side, the modular deployment side — are operating in a space that is drastically undervalued relative to the model companies themselves.

The learning loop layer. Satya Nadella wrote last week that a frontier without an ecosystem is not stable. He's right. The companies that will generate disproportionate returns over the next decade aren't necessarily the ones building the models — they're the ones building proprietary compounding loops around the models. Sector-specific AI that gets smarter from domain-specific data. Industry workflows that accumulate context no general model can replicate. That's where leverage lives, and most of those companies don't exist yet.

The physical stack. Data centers. Cooling infrastructure. Custom silicon. The logistics of getting high-end compute from fabrication to deployment faster and more efficiently than competitors. These are unsexy problems with extraordinary economics attached to them. The sovereign gap. Andrew Curran's analysis is correct: the window for nations to reach the AI frontier has closed. But what follows from that is a massive consulting, advisory, and deployment market — sovereign governments that missed the model race but still need to deploy AI infrastructure, AI-enabled services, and AI-literate workforces. That's a generational services market that is just beginning to form.

The honest version of this moment is uncomfortable. The numbers today are large enough to feel like a ceiling. $3 trillion. $6 trillion combined. Hundreds of billions in annual CapEx. It reads like the top of something. The natural human response is to wonder whether we've missed it.

We haven't missed it. We're watching the ground floor get poured.

The people who made serious generational wealth from the internet didn't necessarily call the 1995 peak or time the 2000 crash. They identified the infrastructure layer that would be indispensable regardless of which companies won — and they positioned there. Bandwidth. Data centers. Cloud infrastructure. The picks and shovels. Those positions, held through the noise, produced the outcomes.

The same principle applies here, scaled up by an order of magnitude.

There will come a time — probably sooner than anyone thinks — when the numbers we're staring at today look like they were obviously too small. When $3 trillion for SpaceX looks like the quaint early valuation of the company that built humanity's off-world logistics layer. When the combined IPO value of OpenAI and Anthropic looks like the bargain-bin pricing of the companies that rebuilt every industry on earth.

The question isn't whether that time is coming. The question is whether you're paying attention right now, while everything still seems large.

---

Rich Washburn is a technologist and strategist working at the intersection of AI, infrastructure, and capital. He is Managing Partner and Chief AI Officer at Eliakim Capital, and CIO of Data Power Supply.