Notes on value, dislocation, and mean reversion
Axia Research is a running record of investment theses, quick reactions to market moves, and portfolio updates — written in public, in real time.
Before earnings season I'm considering spending 1-2% of my portfolio on puts on the Roundhill Memory ETF (DRAM), the most volatile corner of the AI trade. It's insurance: if the trade unwinds, the puts give me cash to buy the names I like at lower prices.
Wall Street expects the tech sector's operating cash flow to roughly double to $2.4 trillion by 2028. The companies that would have to pay for those AI services are forecast to grow far more slowly. Both forecasts can't be right. We look at the four places the money could come from, explain why the law of large numbers and a $2 trillion Anthropic IPO make AI's builders a poor risk/reward, and identify three dislocated sectors (health care, staples and consumer discretionary outside the mega caps) where we'll look for our next AI-beneficiary thesis.
A deep-value case that PayPal's "left for dead" narrative is overstated, its fastest-growing businesses are hidden inside a single-digit-to-low-double-digit multiple, and a buyback alone can nearly double EPS within a few years — with real external validation from a $53 billion buyout offer the board turned down.
Core CPI services came in softer than expected — a same-day reaction, not a position change.
A framework for finding margin of safety in sectors the market has temporarily given up on — and a worked example.
Putting on a small semiconductor hedge against a fully-positioned book, plus a new mean-reversion position — logged here for the record.