On July 31, the US Treasury sold euros from its reserves and bought yen alongside Japan. First time since 1998. Japan's side was roughly $75 billion, the US side maybe $5 to $10 billion. The yen had just hit 163.73 per dollar, its weakest in forty years, and snapped back to 157.57. Two governments standing against anyone short the yen. It looked like the trade of the year. Three weeks later the yen is back near 159. Most of it is gone. The reason is boring. US 10-year yields about 4.69%. The Japanese equivalent yields about 2.85%. As long as that gap exists, money borrows cheap in yen and buys something better elsewhere. That doesn't care what was announced. Goldman's read was that the size was historic but the US role was symbolic, and that intervention buys time rather than fixing anything. Intervention is a signal. Policy is a price. What was actually tradeable? Not the pop. The pop was positioning unwinding, and it reversed. What mattered was what the intervention forced next. Odds of a Bank of Japan hike in September went from 21% to around 81%. That repricing happened quietly over two weeks while everyone argued about the headline. Same pattern everywhere. Earnings, Fed statements, policy announcements. The first move is flows. The durable move comes from whether the underlying incentive changed. Two things to take from this: 1. When a headline moves an asset, find the mechanism. If nothing changed about why money flows, the move is rented. 2. Watch the second-order reprice. Rate expectations, credit spreads, forward curves. Slower, and right more often. 3. If the BOJ hikes in September, the world's cheapest funding currency gets more expensive, and that touches far more than USD/JPY. That's the thing to watch. Not the press conference.

Quick market breakdown for the group. The S&P finished July down 0.13%. Sounds like a nothing month. It was the first negative July since 2014, ending eleven straight positive Julys. The cause was the semi ETF: dropping 22% — its worst month since December 2002. Sandisk fell 47% and is still the index's best performer for 2026. Micron fell 29%. The trigger was capex, not demand. TSMC raised its spending forecast alongside a beat, and the market read higher spending as a threat to future cash flow rather than a sign of confidence. This wasn't panic selling. It was leverage dying. Korea's Kospi hit circuit breakers two days running, down 8% in a session, SK Hynix off nearly 13%. And Leopold Aschenbrenner's fund went from $45 billion to roughly $10 billion in weeks. He ran up to 400% leverage, long AI infrastructure and short software. Both sides moved against him simultaneously, which removed the hedge the pair trade was supposed to provide. Leverage removed his ability to wait and find out. That's the takeaway — being right eventually is worthless if you can't survive the interim. New record high on August 4, Dow above 54,000. Earnings drove it: 86% of S&P companies beat estimates, the best rate since 2021.One caveat. Blended growth reads 47%, but strip out one-time gains at Alphabet and Amazon and it's 29%. Friday's jobs report was bad. Payrolls fell 23,000 against an expected 83,000 gain, and May and June got revised down a combined 103,000. Normally that means cuts are coming. But inflation is still above target, the Fed held 9-3 last month, and several officials have said openly they'd consider hiking in September if prices don't cool. CPI lands Wednesday at 8:30 ET. That print decides which of those two stories we're in.

First earnings report since the June IPO. Clean sweep on the estimates: - Revenue $7.81B vs $6.93B expected — up 92% YoY - Loss per share $0.09 vs $0.26 expected - All three segments beat (Connectivity, AI, Space) - Gross margin 55.3%, up from 43.9% a year ago - Starlink at 12M subscribers, doubled YoY - AI segment posted positive adjusted EBITDA for the first time Capex came in at $18.4B for the quarter. A year ago, it was $2.8B. $15.8B of that went into AI infrastructure alone. This caused Cash Flow to reach -$25B. However, they can absorb it because of the IPO that brought in over $100B. The most important remark of the report was that the CFO said AI compute is hitting a sub-one-year payback. If that's true, $18B/quarter is cheap, and the valuation makes sense. If it's really three years, this is a completely different company and $1.55T enterprise value has a long way to fall. What I'm watching in Q3: does capex hold near $18B or decelerate? Does Starlink ARPU hold $66 for a second straight quarter? And the lock-up expiry — roughly $116B of shares — will probably drive the price more than fundamentals will.

SPCX is down 23% since its IPO five weeks ago. Here are some key things to watch this earnings season. 1. Starlink subscriber growth Went from 4.6M to 10.3M subscribers in about a year. That pace has to hold. The entire bull case leans on Starlink roughly doubling revenue toward $20B — which means user growth has to keep outrunning the next metric. 2. ARPU (revenue per user) They're intentionally dropping price-per-user to grow faster and fend off Amazon's Kuiper and AST SpaceMobile. That's a smart move — right up until subscriber growth stops outpacing the price cuts. If that flips, revenue growth stalls even while the subscriber count still looks good on the surface. This is the metric people scroll past. 3. Starship launch cadence Elon has publicly said there's "genuine risk of bankruptcy" if Starship can't reach a launch roughly every two weeks. Not a side project — it's the load-bearing assumption under future launch economics and the orbital-data-centre story. Watch the real cadence number, not the roadmap slide. 4. XAI's burn rate. Before the February merger, xAI was burning an estimated $9.5B a year against ~$210M of revenue. That merger is why SpaceX posted a ~$5B GAAP net loss. This earnings call is the first real look at whether that gap is closing — or whether SpaceX just inherited someone else's cash bonfire. 5. Capex pace vs. the war chest There's a reported $119B "Terafab" capex plan on the table. The IPO raised $85.7B. Watch how fast they're actually deploying that capital — it tells you how much runway is genuinely left before they need to raise again or slow down. 6. Free cash flow This is the one that ties everything together. Guidance points to roughly $5B in free cash flow if satellite capex peaks and rolls off on schedule. That number tells you whether Starlink's profits are funding the moonshots sustainably — or getting swallowed by them faster than they can replenish. Why this matters: At ~70x trailing revenue on a loss-making company, there's almost no margin of safety in this price. A PE firm would love to own Starlink on its own. The problem is you can't — it comes bundled with two unproven, capital-hungry bets. This earnings report is the first real data point on whether that bundle is starting to pay off or starting to strain.

Quick timeline: back in 2024, Apple and OpenAI announced a partnership to build ChatGPT into the iPhone. Sam Altman even showed up at Apple HQ for it. Good vibes, big handshake. Then things turned. OpenAI bought Jony Ive's startup for $6.4 billion and started building its own hardware. In the process, it started poaching Apple's people — at least 10 engineers have jumped from Apple to OpenAI. Now Apple is suing OpenAI over stolen secrets: A senior Apple engineer, Chang Liu, allegedly kept his work laptop after leaving for OpenAI and used it to download confidential hardware files. So what should investors take from this? 1. Timing. OpenAI has been circling an IPO. A federal lawsuit alleging systemic theft is exactly the kind of headline that ends up in a prospectus risk section, and discovery could drag on for a year or more. 2. The hardware bet gets shakier. If Apple wins even a narrow injunction, OpenAI could be blocked from using specific designs or manufacturing techniques right before its planned device launch. For a company betting heavily on a physical product, that's not a small risk. 3. This points to a bigger pattern. Apple isn't framing this as one bad actor — it's framing it as systemic. If that holds up, it says something about how aggressively AI labs are recruiting out of hardware incumbents, and how exposed that strategy is to exactly this kind of suit. Bottom line: this doesn't kill OpenAI's hardware ambitions, but it adds real legal risk and a timeline drag right when the company can least afford either.