Ignition Portfolio Update #6
Week of August 8, 2026.
Saturday, August 8, 2026 · Two trades this week.
Let me start with a statement that will make some of you raise an eyebrow.
I bought SpaceX.
Rewind to Tuesday afternoon. SpaceX reported earnings for the first time in its life. Revenue nearly doubled, to $7.81 billion. EBITDA came in well above what anyone expected.
Then people found the capex line.
$18.4 billion of spending in a single quarter, $15.8 billion of that for AI compute. Capex numbers up from “only” $7.7 billion the quarter before, when Wall Street expected $13.2 billion. Add up half a year of this and you get $28.5 billion, which is nearly the size of the record amount the IPO raised in June. Not a sustainable amount of capex by any measure. But data centers don’t get built on their own, rockets don’t go to space by themselves, and the latest AI models don’t (yet) become exponentially smarter independently.
So on Wednesday the stock fell 13.6% to an all-time low. Below its own IPO price. Six weeks after listing.
I bought it there. Entry around $108 with half of my cash reserves.
And yes, I know how strange this sounds. I wrote a letter seven days ago saying the layers of the AI trade carrying circular financing, real debt, capex fatigue and giant revenue promises are exactly the layers I refuse to own. I remember writing it. So let me explain myself.
1.1 The Exception
Go back to why I sold all of my Micron in the first place.
I didn’t sell it because capex is bad, capex built every single company I own. I sold it because Micron’s capex, and CXMT’s, and Samsung’s, produces a commodity. And a 212-times-oversubscribed IPO in Shanghai proved that anyone with enthusiasm and a banker can fund more of that commodity.
Economics 101: high demand and low supply means high prices, high prices attract new supply, and new supply kills the prices. It’s just how it works!
But SpaceX’s capex isn’t creating commodities.
Nine million retail investors piling into a CXMT IPO can fund a DRAM plant or a GPU data center. But launch is not a commodity. It might be the hardest business in the public markets to copy, full stop. And Starlink sits on top of it, a distribution network whose barrier to entry is the launch monopoly underneath it.
So.
On Wednesday, at the low, the market was pricing SpaceX below its June IPO price.
People love to hate on SpaceX. Wednesday they got to do it at a discount, and I was on the other side of that trade.
Proudly.
Quick word on the SpaceX CFO’s claim that the AI compute pays for itself in under a year. Do I believe it? I don’t have to. Honestly, a payback claim like that, from a company spending $15.8 billion a quarter, belongs in the same drawer as every other promise I said I distrust. I bought the position without needing it to be true, because SpaceX exists on so many fronts.
And there was one more thing I couldn’t ignore. SpaceX is building its own power plants for a $16.8 billion chip factory. They’re targeting 15 to 20 gigawatts of compute by the end of next year, generated in-house. I spent last Saturday writing that the real constraint of this whole buildout is power, permits and transformers, and that the winners are whoever already controls generation. Then the most famous company in the world announced it plans to be its own utility, and its stock hit an all-time low the very same day.
Sometimes the framework hands you a clean answer wearing an ugly costume.
1.2 The Other Unlock
Thursday was supposed to be the day it all broke.
911.5 million shares came out of lockup. That’s about 43% more stock than the entire 638.9 million share IPO float. The freely tradable portion of the company more than doubled in one morning, from 4.9% of shares outstanding to 11.8%, with the price already sitting below the IPO level. Every mechanical rule about unlocks says that setup resolves lower, holders see the flood coming and sell first. Ask anyone.
Unless two things happen.
The first is that the influx of shares meets a huge demand wave and the tide brings the price up.
The second is simpler: people weren’t selling.
Maybe the first happened, maybe the second happened, very likely a combination of the two happened.
But on the day everyone said it would collapse, it closed up 6.1%.
Friday it rose another 15.8%, its best day ever, and finished the week up around 23% at its highest close since mid-July. The peers followed. The whole space sector had its best week in months.
I keep putting this next to CXMT in my head, because on the surface the two rhyme and underneath they’re opposites. Two weeks ago, supply arrived in the memory market. Funded wafer capacity, a listed company with every incentive to run its lines. I sold Micron, because supply of product is how a commodity cycle dies. This Thursday, supply arrived at SpaceX too. Except it was supply of shares, not supply of rockets. Nobody woke up Thursday with the ability to compete with the underlying business. And when a stock absorbs its own worst technical event, below its IPO price, with everyone positioned for disaster, and goes UP? The selling was exhausted. Simple as that.
Supply of stock is not supply of product. The first one changes who owns a business. Only the second one changes what a business is worth.
I got Wednesday’s price because the market briefly confused the two.
1.3. No Hike!?
Three things happened in macro this week. Together they explain most of the green on the heatmap below, which is exactly why I don’t fully trust the green on the heatmap below. Again, green doesn’t mean you’re smarter.
First, the jobs report. The US economy lost 23,000 jobs in July. Economists expected it to add 80,000. The unemployment rate fell to 4.1%, which sounds nice until you read why, people left the workforce. Wages fell too. Yikes.
It’s not surprising to anyone observing the trends of the job market, but for some reason it was to the analysts.
Going into Friday, markets had the odds of a September rate hike above 57%, with last week’s three dissents still ringing in everyone’s ears. By Friday afternoon the September hike was dead. There’s still a hike priced for this year, just not before December. Yields fell, the dollar dropped, and the S&P 500 closed at a record 7,757, up 3.6% for its best week since April. The Nasdaq rose 5.2%.
So the best week since spring was delivered by a report showing the economy shrinking its workforce. Bad news was good news, because the discount rate blinked. Make it make sense.
Second, gold. Up 7.2% this week to $4,404. That is not a normal weekly move in the world’s oldest asset. When the stock market is at records and gold is moving like that, the two markets are having very different conversations, and I’ve learned to at least listen in on the second one.
Third, the yen, the one I flagged last week when it was still a mystery intervention. Well, mystery solved. The United States and Japan confirmed the first joint yen-buying intervention since 1998. Treasury Secretary Bessent said it out loud and added they “will not hesitate” to do it again. The yen had touched 164 to the dollar, its weakest since 1986, and snapped back to around 156.
Now here’s the interesting part, the carry trade didn’t unwind. In August 2024 a yen rally this size set off forced selling all over the world. This time the emerging market carry indexes barely moved. Strategists are openly calling the euro and the Swiss franc the new borrowing currencies now. And Japan reportedly sourced its dollars without selling Treasuries, which is why the long end of the bond market didn’t convulse.
So the system passed the test. But look at what the test revealed. The trade that blew up the world in 2024 shape-shifted. It’s still here and waiting to unwind. And the Bank of Japan still has a September hike on the table, which is the one event that would ask this question a second time.
One intervention is a warning light. A joint one with the US Treasury attached, that’s two. I went into this week holding 17% cash because of the first warning light. I’m coming out holding 10%, because SpaceX at an all-time low was worth spending some of it on.
I’m watching my portfolio close and the macro environment closer.
I truly believe the chance of a major correction — even a crash — is significant, and I’m not willing to get caught off guard.
Once I smell blood in the water, it’s time to get out.
But not yet.
Before we continue:
Get access to my personal portfolio, watchlist, insights, stock picks, deep dives and more with the annual plan. Subscribe or upgrade now!
2. The Free View
16 positions, 9 themes, and roughly 10% cash.
SpaceX is new and REalloys is a quarter smaller. And for the first time, the largest theme in this portfolio isn’t power or nuclear. It’s space. SpaceX, Karman and Rocket Lab together are now about 21% of the whole thing, which happened partly by decision and mostly by price. I’ll rebalance this week.
Italicized companies belong to multiple buckets.
Power & Grid | GE Vernova ($GEV), Constellation Energy ($CEG)
Nuclear | Cameco ($CCJ), NuScale ($SMR), Energy Fuels ($UUUU)
Critical Materials | ReAlloys ($ALOY), Energy Fuels ($UUUU)
Space & Defense | SpaceX ($SPCX), Karman S&D ($KRMN), Rocket Lab ($RKLB)
Physical & Applied AI |Nokia ($NOK), Kraken Robotics ($KRKNF), Recursion Bio ($RXRX)
Fintech | Pagaya ($PGY), Adyen ($ADYEY)
Software | AppLovin ($APP)
AI Infrastructure | Nebius ($NBIS), SpaceX ($SPCX)
Biotech | Recursion Bio ($RXRX)
Cash ~10%
