Let me start with the numbers the models are dancing around. Best Buy's revenue went from $51.76B in FY2022 to $41.69B in FY2026 — a 19.5% collapse — and the "recovery" in the most recent fiscal year is $160M, or 0.4%. That is not stabilization; that is a flat line with a rounding error. Net margin compressed from 4.7% to 2.6% over the same span, and the operating margin sits at 3.3%, which for a company with 22.5% gross margin means SG&A and interest are eating roughly 19 points of gross profit. The quarterly data, however, tells a more nuanced story than the annual averages suggest: the last two quarters show net margins of 3.1% and 3.2%, up from 2.3% and 2.0% a year earlier, and the holiday quarter swung from a 0.8% margin ($117M NI on $13.95B revenue) to 3.9% ($541M on $13.81B). That $424M swing in a single quarter is either a structural improvement in inventory management and markdown discipline, or a one-time benefit from the AI PC and phone refresh cycle pulling higher-margin product through the registers. I suspect it's both, but the AI component is the part that will evaporate within 18 months.

The valuation math is where the prior models and I part ways. The synthesis pegs fair value at $68.53, implying a 24% downside from $90.27. I think that's too punitive. At $1.26B of free cash flow on a $19.03B market cap, the FCF yield is 6.6%, and the dividend yield is 4.2% with a 75% payout ratio. The P/S of 0.46x is cheap by any retail standard. The P/B of 6.46x looks expensive, but that's a buyback artifact: total equity is only $2.96B because the company has been aggressively shrinking its share count, which inflates ROE to 36% and ROIC to 44% — numbers that describe capital structure, not operating excellence. A more honest multiple is EV/EBITDA at 8.3x, which for a flat-revenue consumer retailer with a 4% dividend is unremarkable, not expensive. My own rough DCF — $1.1B normalized FCF (I'm haircutting the $1.26B because the holiday-quarter margin pop may not fully repeat), 2% growth, 9% discount, 2.5% terminal — lands around $95-105 per share before net cash. That's above the current price, not 24% below it. The synthesis model is clearly anchoring on the revenue-decline narrative and underweighting the margin recovery that is actually showing up in the last two quarters.

Where I do agree with the bearish models is on the structural ceiling. Revenue CAGR is -2.1%, earnings CAGR is -7.2%, and the "36.5% FCF CAGR" in the momentum data is a recovery-from-depression artifact, not a growth trajectory. Best Buy has no moat that Amazon, Walmart, or Target can't replicate with better logistics and lower overhead. The Geek Squad and trade-in services are real but too small a revenue slice to justify a multiple re-rating; the narrative layer's estimate that 35-40% of the price is "story" is, if anything, generous — I'd put it closer to 20-25%, because the AI hardware refresh is a genuine 12-18-month demand catalyst, not pure speculation. The sell-side framing of Best Buy as the "last-mile AI retail beneficiary" is the kind of story that gets written into a model, drives a 60% run from the $55 low to $90, and then gets quietly deleted when the refresh cycle normalizes.

The data point that no model flagged prominently enough is the insider selling. On July 13-14, 2026, insiders dumped roughly 900,000 shares — 336,006, 252,380, 107,534, 88,288, 62,072, 42,106, and 11,614 in a two-day cluster — at a price near the 52-week high of $91.27. That's approximately $81 million in insider disposals. Whether these are 10b5-1 pre-scheduled sales or opportunistic exits, the timing at the top of a 64% run from the low is not reassuring. The subsequent A-Award vesting of ~40,000 shares in July and August is routine compensation, not a counter-signal. A careful investor would treat this as management telling you the stock is where they think it should be, and that the AI-refresh tailwind is a known quantity they are monetizing.