← Back to blog
GeneralAbout 6 min read

Ten Downgrades in Four Sessions, and Not One New Number

Published Nov 17, 2026
Ten Downgrades in Four Sessions, and Not One New Number

Ten research desks moved Nvidia to Hold in four trading sessions. Bernstein halved its price target from $400 to $200. Morgan Stanley pulled the stock off its Top Pick list, Bank of America took it off the US 1 list, Baird cut to Neutral, HSBC trimmed its target. For a company that Wall Street spent four years describing as a monopoly, that is a remarkable amount of movement in a week.

Here is the part that gets lost: not one of them published a new number on the order book.

Nvidia closed Friday, November 13, at $162.85, down 12.75 percent for the week. It touched $158 intraday on Wednesday, the day the board announced Jensen Huang's resignation. Volume was 741 million shares on Wednesday and 363 million on Friday. That is not rebalancing. That is a crowd heading for the same exit at the same time.

A downgrade is a document about the analyst

Read the actual notes and a pattern shows up. The language is about uncertainty, not about demand. Truist and Craig-Hallum cited "uncertainty about the leadership transition." HSBC cited the leadership vacuum plus the unresolved question of export policy. Baird simply moved to the sidelines.

None of them said customers are cancelling. None of them said Rubin is late. None of them cut revenue estimates in a meaningful way. The consensus for the just-reported quarter stayed at $108.8 billion with $2.47 in earnings per share, and earnings revisions over the past ninety days are still positive at plus 5.2 percent.

So what changed? The range of outcomes. A founder-CEO premium is not a line item, but it is real, and when it disappears an analyst has to widen the band around their model. Widening a band is not the same as lowering the midpoint. It does, however, require a lower rating, because the rating is a statement about confidence, not just about price.

Trading floor screens in a dark room, rows of falling red lines with no readable numbers

The arithmetic that makes this awkward

At the November 11 close of $173.73, Nvidia traded at roughly eleven times consensus fiscal 2028 earnings of $15.79 per share. Eleven times, for a business growing revenue around ninety percent with a 74 percent gross margin.

AMD, on the same basis, traded at about thirty-one times its next fiscal year earnings. Broadcom traded at eighteen. Both are excellent companies. Neither is growing at Nvidia's rate.

You can argue the discount is deserved — customer concentration, circular financing, China exposure, key-man risk. But note that these risks existed on November 5, when the stock fell 9.3 percent in a single session, and on October 6, when it began a 24 percent slide that had nothing to do with the board. The market has been repricing Nvidia for six weeks. The resignation gave it a label, not a new thesis.

Then there is the rates problem. The ten-year Treasury yield rose 29.5 basis points during that week to 5.154 percent. At that level, every long-duration growth stock has to justify its multiple against a risk-free alternative that actually pays. A governance shock and a 5.15 percent ten-year arriving in the same week is how you get a 12.75 percent weekly loss on a company that is still guiding to ninety percent growth.

The investors who moved before the analysts

One detail worth putting next to the downgrade wave: some large holders were already reducing exposure before any of this happened. Duan Yongping's investment vehicle cut about 3.96 million Nvidia shares in the third quarter, a 63 percent reduction that left it with 2.32 million. ARK sold another 543,000 shares through two ETFs on November 16.

Neither sale was a reaction to the board. Both were filed in a quarter when the stock was still near its highs, which means the people with the longest holding periods were trimming into strength and the analysts with the shortest horizons were cutting into weakness. That is a normal division of labor, and it is also a reminder that by the time a rating changes, the position has usually already moved.

What actually converts a Hold into a Sell

Two triggers, both mechanical.

The first is $158. That was the intraday low on November 12, at the worst of the panic. If Nvidia closes below it after the November 17 earnings call, the analysts who moved to Hold are suddenly looking at a stock that broke the panic low on news they have already downgraded for. At that point, protecting yourself means going to Sell, because the story is no longer "uncertainty" but "the market rejected the explanation."

The second is the ten-year. If the yield holds above 5.10 percent, the math for multiple expansion gets harder every week. Nvidia can beat on revenue and still drift lower if the discount rate is climbing faster than earnings estimates.

Neither trigger has anything to do with whether Vera Rubin ships on time. That is the uncomfortable truth about this kind of selloff: the fundamental case and the price path have decoupled, and they stay decoupled until one of them gives.

The rotation is real, and it is not a verdict

AMD gained 18.93 percent that week and closed at $585.30 on November 12. Broadcom gained 18.16 percent. The two moves are nearly identical, which tells you this is a sector rotation rather than a stock-specific bet on AMD's roadmap.

Capital did not leave AI. It left the one name that had a governance problem and went to the two that did not. That is what a portfolio manager does at 9:31 on a Thursday morning when the compliance desk forwards a headline: you do not re-underwrite the industry, you reduce the single-name exposure and keep the sector weight.

William Blair upgraded AMD the same week, arguing that as hyperscalers diversify away from one supplier, AMD's data center roadmap moves from alternative to necessity. That is a reasonable case. It is also a case that has been true since AMD's Helios rack and Venice ramp were announced, and the stock only moved when Nvidia stumbled. Which suggests the upgrade was less about AMD and more about the hole that opened next to it.

What the ten notes actually tell you

They tell you the sell side has decided that the founder premium was load-bearing. Daniel Ives put the number at $300 with Huang and $225 without, and the stock is now below both, which means the market has priced in more than a CEO change — it has priced in a change in how Nvidia is valued.

They also tell you that nobody wants to be the last analyst with a Buy on a stock that is making governance headlines. Ten downgrades in four sessions is not ten independent judgments. It is a queue.

The useful thing about a queue is that it empties. Once everyone is at Hold, there is no one left to downgrade, and the next rating change has to be an upgrade. That does not mean the stock goes up. It means the pressure from ratings is mostly spent, and from here the price depends on things analysts cannot vote on: whether the fourth-quarter guide holds, whether Rubin ramps, and whether a permanent CEO shows up before the next earnings call.

Eleven times forward earnings is not a valuation that requires optimism. It requires the absence of new bad news. That is a lower bar than it looks, and a harder one to clear than any of those ten notes admits.

Related articles