The quarterly tape is the most arresting thing in this file and the prior models are largely glossing over it. Revenue went $1.76B → $1.88B → $1.88B → $1.70B → $1.90B → $2.31B → $3.03B → $5.95B across the last eight quarters, with net income swinging from -$1.93B in Mar-2025 to +$3.62B in the April-2026 quarter — a 60.8% net margin on a NAND business. That is not a cyclical recovery; that is either a data error, a one-time gain (litigation, tax, divestiture, purchase accounting), or the single most violent NAND up-cycle in industry history. Annual figures haven't caught up: the trailing FY ending Jun-2025 still shows -$1.64B NI on $7.36B revenue. The two TTM-flavored ratios (P/E 64, EV/EBITDA 53) are computed off a window that already includes the explosive Q4 print, while P/S 0.93 and P/B 0.74 are computed off stale annual denominators. The synthesis model's "39.5x sales" claim is simply wrong on the inputs — at $290.79B market cap against TTM revenue of roughly $13.2B (sum of the last four quarters), P/S is closer to 22x, and against an annualized run-rate of $23.8B from the Q4 print it's ~12x. Still expensive, but the model is anchoring on the wrong denominator.

The Market Forces note that "SanDisk ceased to exist in 2016" is also wrong in the other direction — Sandisk was spun back out of Western Digital in early 2025, which actually explains the violent quarterly discontinuity (spin-related accounting, inventory revaluation, NAND ASP recovery hitting a newly-levered standalone entity all at once). The Pre-Flight model correctly smelled "cyclical hardware turnaround" but then the narrative_platform classification (confidence only 0.5) drove the rest of the stack to suppress profitability metrics — exactly backwards for a memory cyclical, where you absolutely want to look at margin trajectory. Gross margin from -114% in Mar-2025 to 60.8% net in Apr-2026 is consistent with a NAND price spike layered on a fixed-cost manufacturing base post-spin; it is not consistent with steady-state earning power. Operating cash flow of $84M on $7.36B revenue with -$120M FCF tells you the FY25 annual is a trough snapshot already obsolete.

A careful contrarian would argue three things. First, the 60.8% net margin quarter is almost certainly non-recurring — either a deferred tax asset release, a one-time gain on the spin, or peak-cycle NAND ASPs that mean-revert in two-to-four quarters; mid-cycle net margins for NAND pure-plays historically top out around 15-20%. Second, at $1,963/share and $290B market cap, even crediting the $5.95B quarter at face value gives a forward P/E around 20x on hypothetical $14-15B annualized earnings — which only works if you believe peak earnings are sustainable, which they aren't in memory. Third, insider activity is quietly bearish: every transaction in the last 60 days is a sale, F-InKind (tax withholding on vesting, neutral), or a gift — zero open-market buys with the stock at all-time highs after a 5x revenue acceleration. That is what insiders do when they think the print is the peak.

I dissent from the synthesis verdict's reasoning but agree with its direction. "Disconnected from fundamentals" is right; "39.5x sales on a money-loser" is the wrong math. The correct framing: SNDK is trading at roughly 12x peak-cycle revenue and ~20x peak-cycle earnings in an industry that historically trades at 1-2x trough revenue and 8-12x mid-cycle earnings. Apply Micron/SK Hynix-style mid-cycle multiples (10-12x normalized EPS, where normalized assumes ~$8-10B revenue at 15% net margins = $1.2-1.5B) and you get a fair value market cap of $15-20B, or roughly $100-135/share — an order of magnitude below spot. Even being generous and assuming the new Sandisk holds $20B revenue at 25% net margins through-cycle gives $5B normalized earnings × 15x = $75B, or ~$500/share. The $1,963 price requires believing NAND has structurally re-rated into a software-like business, which the bear narrative (Samsung, Hynix