July 2026 · Interim Note
A Good Week for the Book, for Two Unrelated Reasons
This week produced two genuinely good headlines for the book, and I want to be precise about why they're not the same story wearing two hats, even though the tape reaction looked similar.
Two headlines, two theses
On Thursday, Micron announced a $250 billion, multi-year commitment to U.S. memory manufacturing. MU rose 4.5%, and SanDisk — a separate holding in the same memory corner of the semiconductor sleeve — rose harder, up 7.6%. That's a supply-side and demand-signal story: a chipmaker committing real capital because it believes memory demand from AI infrastructure is durable enough to justify a decade-long buildout. I wrote it up in full on the News page.
On Friday, Meta confirmed it will start manufacturing its own custom AI chip in September, targeting 14 gigawatts of compute by 2027. META rose 4.7%. That's a different story entirely — vertical integration, a company betting it can own more of its own supply chain rather than depend entirely on Nvidia and cloud vendors. Full writeup here.
Both happened against the backdrop of live U.S.-Iran conflict headlines and a Nasdaq that still finished the week up on both sessions — a reminder that geopolitical risk and company-specific capital-allocation news can move the same tape in opposite directions on the same day.
Why I'm keeping these separate, even though the P&L doesn't care
It would be easy to file both under "good week for the AI-infrastructure theme" and move on. I don't think that's accurate. Micron's news is about memory-specific supply discipline; Meta's is about one company's decision to internalize part of its supply chain. If I treat every green day across MU, SNDK, and META as confirmation of one undifferentiated thesis, I lose the ability to tell later which specific bet is actually working and which one is just riding sentiment. The July letter's clustering disclosure is about correlated risk — it isn't a license to also treat correlated gains as a single data point when they aren't.
No position changes from either headline. Both are multi-year commitments whose real read-through won't show up for quarters, not days.