NVIDIA Corporation · Technology
Survives a correction-sized 3-point cut to analyst growth.
At 46.0% margin the price requires 15.5% 3y revenue CAGR vs consensus 63.7%, a cushion of +39.0pt (+42.6pt at 1pt lower r) against a 32pt correction cut. Rule says on_top; chair keeps holds. The 39pt cushion clears a 32pt bust cut by only ~7pt, and the CEO resignation dated 2026-11-12 plus the worst news tone in the batch (10d -0.32) mean estimates are still moving.
Breaks if: Estimate cuts take consensus 3y CAGR below ~48% (cushion falls under the cut), or the interim leadership triggers customer order deferrals.. We re-check this against analyst estimates every evening.
10 years, at a 46% margin (analysts' margin, cut for cash conversion)
+37.4%/yr over 10 years once faded to terminal growth
points off 3-year analyst growth
50 published targets
Part of the AI epicenter
Does the profit analysts build their margin on actually arrive as cash? Measured over the last 3 annual statements filed before this test. After stock-based pay, 81% of reported profit arrived as operating cash, so the test cuts the analysts' 54% margin to 44%. That leaves a cushion of +38pt instead of +43pt, and the smaller of the two decides the group.
How much of reported profit arrived as operating cash. Below 0.8× is a red flag.
Profit not backed by cash, as a share of assets, latest year. Above 10% is a red flag.
Staff paid in shares instead of cash, which flatters margins. Above 10% is a red flag.
Dilution of each share's claim on the business. Above 3% a year is a red flag.
Non-operating gains in the latest year. Above 25% is a red flag.
After capex. Shown for context: heavy investment is not bad earnings, so it is never solved under.
From the priced-in reconstruction of Nov 21, 2026. Its figures use that day's price and free-cash-flow margin, so they can differ from the stress test above.
At the price the market sits at the bottom of the published model range — just 2% of 50 analyst targets are below it, versus a median of $296 (+136%) — meaning the price is paying for deceleration, a leadership vacuum and rising doubt about AI-spending circularity, and refusing nearly every bull case on the street.
Does Data Center demand actually decelerate to the ~11.6% per year the price requires, or hold near street estimates through calendar 2027? Unit-volumes via segment revenue history can measure whether Data Center growth is rolling over quarter by quarter — this is testable for the demand path, though not for customer credit quality on neocloud commitments, which nothing wired can measure.
A stress test of whether this narrative is still paid for by analyst forecasts, not a forecast that the stock will rise. The price-implied growth on its own has not predicted returns in our testing, and this list (sealed Nov 21, 2026) has not been graded yet. Cut sizes are assumptions. Not investment advice. Back to all names