A 47% crash did not make Lam cheap. What would?

A record $8.22 billion quarter, and $38 billion gone the next morning

Lam Research ($LRCX) reported the best quarter in its history on 21 October 2026 and the market punished it anyway. Revenue for the quarter to 27 September 2026 came in at $8.22 billion, up 22% on the June 2026 quarter and 54% on the $5.32 billion of September 2025. GAAP gross margin reached 52.3%, operating margin 40.0% - the first time Lam has crossed 40% - and non-GAAP EPS was $2.20 against a $2.18 estimate. Operating cash flow was $3.50 billion, capital spending $258 million, so free cash flow was about $3.24 billion in 13 weeks. The quarterly dividend went up 27% to $0.33.

The guidance is where the sale started. For the quarter to 27 December 2026 management guided revenue to $8.60 billion plus or minus $400 million, gross margin 52.5% plus or minus 1 point and EPS $2.33 plus or minus $0.15. That beat the $8.55 billion sell-side consensus but fell short of the $8.9 billion some buy-side investors expected. Two reasons: component shortages push part of the December shipments into the March 2027 quarter, and China, which jumped from 26% to 35% of revenue in the September quarter, is expected to fall sharply in December. The stock closed at $230.15 on 22 October 2026, down 11.76% from $260.81 - roughly $38 billion of market value gone in one session on a quarter that beat on every line.

The company paid $329 a share in September. Its CEO sold at $319

The 10-Q shows Lam bought back 2.344 million shares for $1.01 billion in the quarter to 27 September 2026, an average of $329.17 per share. On top of that a $1.0 billion accelerated repurchase delivered 2.3 million shares up front and settles in the December quarter; $3.03 billion of authorisation remains, and the long-term policy is to return 85% of free cash flow to shareholders.

In the same window insiders made 65 sales worth $152.5 million and zero purchases. The CEO sold 30,000 shares at $319.26 on 9 September 2026, a director sold 9,557 shares at $314.74 the same day, and the chief legal officer sold 5,000 at $302.46 on 31 August 2026. At $230.15 the September buyback is 30% under water within one quarter. A buyback works like any other purchase: it adds value per remaining share only if the price paid is below what the business is worth. At 65.8x earnings - the multiple on 28 June 2026 - the company was buying at the top of its own cycle while the people running it were selling.

Why the model stops at $163.25: 24 times earnings for a business that shrank 14.5% in fiscal 2024

Trailing four-quarter GAAP EPS is $6.71 ($1.26, $1.45, $1.81 and $2.19 for the quarters to December 2025 through September 2026). That puts $230.15 at 34.3x trailing earnings and the FPI fair value of $163.25 at about 24.3x - close to the 23.4x the market paid on 29 June 2025 when the stock was $97.20, and far from the 13.7x of June 2022 or the 65.8x of June 2026. Against the December guide annualised to $9.32, the price is 24.7x and the fair value 17.5x.

Two things pull a DCF for this company in opposite directions. Return on invested capital of 49.9% and a net margin of 32.3% - against an industry median of 5.7% and 10.9% - are the inputs that lift any cash flow model. A free cash flow yield of 2.36% (annual free cash flow as a share of market value, roughly what a savings account paid in 2026) and a revenue history that fell 14.5% in the year to June 2024 when NAND spending collapsed are what cap it. Equipment makers earn their best margins at the point where customers all build at once, which is exactly when earnings look least sustainable - the logic explained on the cyclical stocks page. On the peer leg, Lam's EV/EBITDA of 28.4x and P/S of 11.0x sit above the semiconductor median (27.6x, 5.9x) and above Applied Materials (26.2x), so the 30% relative weight does not rescue the number either.

Street vs model: even the lowest target on Wall Street is $127 above the fair value

Wall Street counts 39 buy, 10 hold and 1 sell. The median target of $347.50 is 51% above the price and 113% above the fair value; the mean is $365.60, the range $290 to $500. On 22 October 2026 Needham cut from $390 to $290, Morgan Stanley from $385 to $339 and Mizuho from $365 to $360, all keeping buy-equivalent ratings; Evercore ISI set $315 on 20 October. Nine targets issued in the month to 22 October average $355.44. Cantor Fitzgerald kept $500 and raised its 2027 EPS estimate to $10.75.

The gap between $347.50 and $163.25 is a disagreement about which year to capitalise. Management put 2026 wafer fab equipment spending in the mid-$150 billions and gave a preliminary 2027 view of more than $200 billion, with memory systems revenue up 120% on September 2025. The median target equals about 32x Cantor's 2027 estimate - a multiple that treats a year of record spending as a plateau. The 50% DCF leg of the FPI blend normalises across the cycle; the 20% consensus leg is the only component pulling the fair value up, and at a 29.1% premium it is not enough.

Three numbers for the next 96 days: $8.2 billion, 35% and $283

Upside case: December revenue at or above $8.6 billion, the delayed shipments landing in the March 2027 quarter, gross margin holding the 52.5% range and 2027 equipment spending firming above $200 billion. Then $230 is 24.7x a $9.32 run-rate and the cycle argument weakens. Downside case: China's share dropping well below 35% without other regions filling the hole, gross margin at the 51.5% bottom of the range, or the accelerated repurchase settling in December with no follow-on - which would show the 85% payout was a peak-cycle promise.

Signals to watch:

• Revenue above $8.2 billion for the quarter to December 2026, due 27 January 2027 (estimate $8.65 billion, EPS $2.35).
• The 200-day average at $283.29 against the 50-day at $304.53; RSI(14) at 25 says the selling has been one-way since the 30 June 2026 peak at $438.50.
• $163.25 - the level at which the model's 29.1% premium disappears.