Last Thursday, roughly $100 billion of SpaceX stock became sellable for the first time. Some of it is mine.

Everyone I spoke to asked some version of: are you taking profits?

Five days later, here’s the state of play.

• The stock rose about 6% on Thursday, closing at $114.92
• It rose almost 16% on Friday, closing at $133.11

As I write this on Tuesday it’s trading around $132, a day after closing at $138.74, its first close back above the $135 IPO price in weeks.

The market priced a supply event and got a demand event.

I’m still not selling. Not a share. And the more interesting story from last week is that the people who could finally sell mostly didn’t either.

Why I’m not selling

I wrote my first check into SpaceX in 2019, at around a $30 billion valuation, when Starlink had zero customers and serious institutional investors still treated space as a curiosity.

I’ve held through every version of “too early,” “too expensive,” and “too much capex” since.

And all through last week I ignored the chart and stayed on the calendar.

The honest risks haven’t changed.

The spending is enormous: over $18 billion of capital expenditure last quarter, most of it poured into AI infrastructure, and that number is what sold the stock off on earnings day. Adjusted EBITDA came in at nearly double what Wall Street expected, but the company remains unprofitable on a net basis. More unlock tranches arrive between now and early December, and share supply will keep shoving the price around.

If your horizon is measured in quarters, this is a hard stock to hold.

My horizon isn’t. I’m underwriting the next five years, at minimum.

Last Thursday’s tell

The lockup was designed with a performance condition: an extra 455.8 million shares would only unlock early if the stock closed at least 30% above the $135 IPO price ($175.50) on five of the ten trading days into earnings. It never got there.

The people who wrote the structure tied their earliest exit to a meaningfully higher price.

Then the gates opened, and the holders with the lowest cost bases in the company (early employees and early investors sitting on years of gains) looked at the exit and declined it. The stock rose 23% into the largest share unlock in the company’s history.

What I think the market is missing

At today’s price, the market is valuing a very good launch-and-broadband business with an expensive AI side project.

Revenue grew 92% last quarter. Starlink subscribers doubled to 12 million and, per Musk, the service reaches a $20 billion run-rate this year. And the stock still spent six weeks underwater because of the capex line.

But the capex is the point: a vertically connected stack where each layer makes the next one more valuable.

Starship collapses the cost of getting mass to orbit.

Starlink stops being just internet for people and becomes the nervous system for machines since every AI agent and robot coming online needs bandwidth no ground network can carry.

The compute layer fills that capacity: contracted AI cloud customers today, and eventually data centers in orbit, running on sunlight that never sets, with no permitting, NIMBYs, or grid to fight.

The software and intelligence layer sits on top of all of it.

• The rocket carries the compute
• The compute serves the intelligence
• The intelligence drives the machines
• The machines demand the network
• The network needs the rocket

Every turn of that wheel makes the next turn cheaper, faster, and harder for anyone else to match.

No other company on earth owns the whole loop, and the weekend gave us the scale of the ambition in numbers.

The entire world’s bandwidth today runs somewhere between 2,000 and 8,000 terabits per second. On Saturday Musk confirmed the plan: V3 satellites roughly 10x more capable than the ones flying today, 10x more of them launched, and a total constellation exceeding 100,000. That’s a system with over 100x Starlink’s current bandwidth, which would move 10 to 50x more data every second than the entire internet does right now.

Musk’s floor case: even if revenue per gigabit falls tenfold, communications revenue alone clears $200 billion a year.

Who’s coming online: 2 billion people without regular internet, and a machine population (AI agents, robots, sensors on and off the planet) that will consume enormous amounts of bandwidth. The plan is to build the capacity for the internet that’s coming.

The trillion-dollar line

On August 4, its first earnings call as a public company, SpaceX moved its trillion-dollar annual revenue target forward, from 2031 to 2030.

When a well-meaning holder on X suggested over the weekend that he must have meant cumulative revenue, Musk corrected: “actual revenue in the year 2030 itself, not cumulative.” In the same post he conceded it’s “always possible that we flounder along the way or are delusional,” but that these are the real internal projections.

On the Starship launch ambition: “Probably a year from now we will be doing at least one flight a day, possibly more.”

What I’m watching over the next 18 months: the December run-rate print, the first successful ship catch, how quickly Starlink V3 actually deploys, and whether the 2027 gigawatt targets stay on the board.

Five years out

Stand in 2031 and look back.

If Starship is flying daily, Starlink is carrying a meaningful share of the world’s traffic, and gigawatts of compute are live and paying, then this 2026 IPO summer (the drawdown, the lockup panic, the capex complaints) will read as a classic early-infrastructure shakeout.

The same pattern as every layer before it: reusable rockets were impossible, Starlink was unviable, and this buildout is “too much capex.”

I didn’t write multiple checks into SpaceX since 2019 to sell in week nine of public trading.

My plan for the rest of the unlock schedule

Here’s the supply calendar between now and year-end (dates approximate, per the reported schedule):

• Around August 21: the next tranche, roughly 319 million shares
• September 10 and 24: two more releases of similar size
• October 9 and 24: two more, roughly 328 million each
• Two trading days after Q3 earnings (late October or early November): the largest single release, up to roughly 1.3 billion shares by some counts
• December 8, day 180: the final scheduled unlock, taking the tradable float to around 40% of the company

Two notes on that list. Unlocking makes shares eligible to sell; it doesn’t sell them. Thursday proved the difference. And Musk’s own stake, along with certain other holdings, stays restricted into mid-2027. The largest holder on the cap table couldn’t join the exits this year even if he wanted to.

Between now and December, the price will do what prices do when new supply meets a story that’s mid-build, and none of it will tell me anything I need to know. The December run-rate number will. The first ship caught out of the sky will. Starlink’s subscriber curve will.

Those are the numbers I’ll check, on the calendar they arrive on.

The things that would change my mind are physics and customers: like a heat shield that never gets solved, a launch cadence that never arrives, subscribers that stop doubling.

Until then the position does what it’s done since 2019: it sits, and it compounds.

I bought a decade and I intend to let it pass.

One more thing, because it’s the part I actually work on every day. Whatever SpaceX becomes, it has already accelerated the layers beneath it: the manufacturing, propulsion, sensing, and software that every constellation is built from.

That layer is where Cape Fear Ventures invests.

Richard

I own SpaceX and have since 2019. This letter explains my decision, not a recommendation for yours. Nothing here is investment advice or an offer of any security.