Implied forward EPS growth, AI cohort
Forward 12.9x on Micron with +173% expected EPS growth. Forward 51.8x on Lam Research with +7%. Same cohort. Same rally. Different reality.
Trailing P/E divided by forward P/E minus one gives the implied forward EPS growth the consensus has baked into today's price. The dispersion across the AI cohort is not subtle.
Micron +173%. AMD +144%. Broadcom +123%. Sandisk +121%. Nvidia +70%.
Lam Research +7%. KLA +5%. Applied Materials +14%. Meta +17%. Alphabet +9%. Amazon +3%.
Same 90-day rally. Same +50% to +70% return. Two different stories underneath.
Micron at $757 trades 35x trailing earnings. The forward multiple is 12.9x. The market expects $101 of EPS over the next year against $21 last year.
That is not a bubble multiple. That is a multiple that has compressed because the EPS line caught up to the price. Same arithmetic on Sandisk: $29 trailing, $169 forward, the multiple went from 53x to 24x not by sentiment shifting but by the model adding earnings.
Operating margin tells the same story. Micron 67%. Sandisk 70%. These are peak-cycle numbers. They will revert. But the price already prices reversion: forward 12.9x assumes margins compress and EPS still hits $101.
Lam Research went $174 to $295 over ninety days. The forward EPS estimate moved from $7.40 to $7.91. The multiple moved from 33x to 52x.
That is not the same trade as Micron. That is the multiple doing the work, not the model.
KLA, Applied Materials, the equipment cohort runs the same arithmetic. Forward EPS basically unchanged. Multiple compounded.
Equipment names are real businesses. Their order books reflect the AI capex flowing through. But the price already discounts six quarters of confirmation that has not arrived.
Western Digital is the cleanest tell. Stock +24% in fourteen days. Forward P/E 48.4x against trailing 28.7x. The market is openly pricing earnings DECLINE next year while the stock runs with the cohort. That is cohort beta with no underlying engine.
Three groups across two engines:
Fundamental-driven (chips): Micron, AMD, Broadcom, Sandisk, Nvidia. Multiples already discount the EPS the model expects. If margins hold, these names are flat-to-up on the print. If margins crack, the forward multiple was the wrong anchor and the names give back what the EPS line was supposed to deliver.
Multiple-driven (equipment + hyperscalers): Lam Research, KLA, Applied Materials, Meta, Alphabet, Amazon. Multiples compounded against EPS that did not move. These names are the ones that look like 1999 in arithmetic terms, even though the underlying businesses are profitable and growing. The multiple is the bubble; the company is not.
The trap: Western Digital. Multiple expanding while EPS forecast cuts.
Tuesday at 8:30 ET prints April CPI. Cleveland Fed nowcast 3.56% YoY. Consensus 3.6 to 3.8%. Hot above 3.8% pushes the September cut probability from 25.5% down toward 15%, and the multiple-driven leg of the cohort takes the move directly. Cool below 3.5% extends the multiple expansion.
Thursday after-hours prints Applied Materials Q1 fiscal. Street EPS $2.68. A beat with raise validates the equipment cohort's multiple. A miss with cautious guide forces the multiple back to fundamentals.
The fundamental-driven leg is regime-invariant near-term. The order books are signed. The chips ship. The ASP cycle has 12 to 18 months of grace before historical reversion. The capital that went long Micron at 12.9x forward is buying the model, not the cycle.
The cleanest expression of the AI capex thesis at present prices: Micron at forward 12.9x with the highest expected EPS growth in the cohort and 67% operating margin. The diversified version: a semicap basket like SMH or SOXX, which over-weights the fundamental names by index construction.
The second-derivative play that moved with the cohort but on a different engine: uranium. URA +28.64% over ninety days as data-center power demand priced into the spot market. Nuclear utilities ran softer.
The optionality-preserving move: Treasury bills at 3.68% costs $14 a week per $20K. That is a cheap option premium on Tuesday's CPI and Thursday's AMAT print, both of which can re-weight the matrix above.