S&P 500 futures and Nasdaq 100 futures are rallying ahead of the open, poised to continue to claw back some of last week’s losses.
The benchmark US stock index suffered its worst weekly loss since April 2025, but isn’t getting any help in repairing the damage from its second-largest constituent on Monday.
Nvidia is sharply lower after CEO Jensen Huang admitted to an affair following reporting that he had a romantic relationship with a woman the US believes to be a Chinese spy. However, traders are treating this like a uniquely Nvidia issue for now, with the VanEck Semiconductor ETFflat as of 8:30 a.m. ET.
Robinhood traders stepped up their single stock purchases as the market rebounded on Friday, with the likes of TeraWulf, Cloudflare, Opendoor Technologies, and Sandisk seeing notable jumps in volumes and net buys.
As a subsidiary of Robinhood, Sherwood Media is restricted from writing about any company in which Robinhood is or was a selling group member of the IPO during the regulatory "quiet period" for that company.
Honeypothole
We’ve spent all this time worrying about AI cybersecurity risks when our true vulnerabilities come from human frailty.
Nvidia CEO Jensen Huang put out a statement on Sunday apologizing to his family and Nvidia staff for infidelity in the wake of media reports alleging his romantic relationship with Alice Huang, a tech executive who the US government believes to be a Chinese spy. Nvidia’s board also announced the launch of an independent investigation into the matter.
Shares are down big in early trading.
Alice Huang works at Megaspeed, a cloud computing company that authorities in the US and Singapore have investigated amid concerns of chip smuggling in violation of US export controls.
Based on (very) thin trading on Polymarket this morning, prediction markets give about a 1 in 7 chance of a Jensen exit from Nvidia being announced before the end of the year. That outcome would be bad news for the stock, per Wall Street analysts, who also flag the risk of an increasingly challenging regulatory environment as a headwind for the shares.
From Yorkville Ives’ Dan Ives:
“This is a black eye moment for Jensen and will spark multiple congressional inquiries at a minimum and will create an overhang on the stock for a myriad of reasons. The key question investors will be asking is: Will Jensen remain CEO of Nvidia?….The investigation will be key and we would expect Jensen to continue in his role as CEO with no interruption in the near-term. There is clearly risk that Jensen could be removed as CEO depending on the Board investigation and this will add risk to the stock as he is so important to the Nvidia story and strategy globally.”
And Bernstein’s Stacy Rasgon:
“Substantive or not [headlines] reading ‘NVIDIA’s CEO got honeypotted by China’ are (at the very best) going to be a heavy distraction, and at the worst provide significant ammunition to both sides of the aisle, from China hawks (of which there are many on Capitol Hill) already displeased with the current US stance toward the country, to incoming Democrats unhappy with the close ties to Trump that Jensen has cultivated over the last few years. We believe demand for the company’s products seems unlikely to be affected, but it seems plausible to postulate increased risks (regulatory etc.) as AI buildouts become an increasingly political issue, and this news seems likely to keep NVIDIA (and its CEO) at the center of that debate.”
About a year ago, Jensen warned that China “will win” the AI race versus the US. He’s pushed to be able to sell dated or nerfed AI processors to China, arguing that it’s better for US national security to get the country more locked in to American platforms. At times, opposition to this argument has come from within the US; at other times, China has seemingly preferred to bolster its domestic chip capabilities rather than use Nvidia’s hardware.
Occam’s razor suggests a CEO advocating to be permitted to sell more product is a case of classic self-interest rather than a signal of nefarious intent. That being said, any attempts by Jensen to push for greater access to the Chinese market will now face greater scrutiny and skepticism —- and might just be a complete dead end.
“It may not be an accident that intelligence leaks have coincided with US midterm elections, when Congress is newly emboldened to pressure the administration to tighten controls,” writes 22V Research analyst Michael Hirson. “While both Washington and Beijing have incentives to maintain the truce on tech and supply chains, domestic politics in the US and bilateral dynamics are making it harder.”
I’ve spent a fair amount of time thinking about what’s priced in for Nvidia. And, well, this throws a different variable into the calculus.
On just the fundamental nuts and bolts, though, let’s assume Wall Street called for the chip designer’s earnings growth to completely stagnate in perpetuity, which would keep 12-month forward earnings per share estimates around $8.88. If Nvidia were to earn the “market” return (let’s call that 8.5%), it’d take two years for the stock to get back to its median forward P/E multiple (24.7x). Two years down the road is late 2028! And because we’re talking about one-year forward estimates, this would entail that analysts think Nvidia’s earnings run rate in the second half of 2029 will be the same as it was in the middle of 2026.
If that bottom-line outlook strikes you as too conservative, let’s pencil in sequential EPS growth decelerating to just 7% quarter-on-quarter from an average of 17% since the three-month period ended January 2025. Well, that would leave you with forward earnings per share estimates a couple pennies away from what’s implied at the 25th percentile valuation (17.2x).
Suffice it to say, Nvidia’s multiple implies a great degree of skepticism on the medium-term outlook for profits. And that, of course, is something you could have said for months.
Valuations are a poor timing tool. This applies both to avoiding stocks because they’re expensive, or buying them just because they’re cheap.
Price-to-earnings ratios might be of particularly little utility when the market is focused on the longevity of an unprecedented boom for a firm in a historically cyclical industry — and when the architect of this profitability suddenly finds himself in a publicly precarious position.
Dinged by Bing
During the noon hour on Friday, a hot headline across the Bloomberg Terminal dented the bounce-back in tech stocks and served as a shot across the bow for the health of the AI trade:
*MICROSOFT IS WINDING DOWN ORACLE DEAL TO RENT AI CLOUD CAPACITY
Per the report, Microsoft decided to end the pact to use Oracle servers to power an AI-enhanced Bing search engine because the older-generation chips “no longer fit the company’s specifications.”
[Editor’s note: Ouch. Not being good enough for Bing, that’s gotta hurt.]
On the surface, this is bad news. All we’ve heard about during the AI boom is demand outstripping supply, and here we have a major buyer of compute turning down supply!
In addition, we’ve consistently received color that the asset quality of older GPUs has been much stronger than once anticipated. The so-called “depreciation bears” have been in shambles. This is among the first high-profile anecdata that suggests…maybe not.
But this case requires some zooming out: Microsoft’s view of older chips is belied by the data. Rental rates for A100 chips (which are multiple generations outdated) aren’t nosediving, but actually firming up in recent weeks and well off their lows of the past year.
A reasonable outcome (likely, even!) is that another buyer valued this compute more than Microsoft, so Oracle was able to sell capacity at a higher price to a different customer.
We’ve remarked on how 2026 has been an “ask questions later” market; when it comes to potential negatives for Oracle, sometimes the questions might never even get asked.
If there’s a bearish AI trade takeaway here, it circles back to the first element we raised: supply and demand.
Microsoft likely doesn’t walk away from compute unless they’re optimistic about what their pipeline of capacity and completions look like. That’s not something that would be dampening rental rates now, but is something that would show up in the future if gigawatts grow enough.
On a related note, there have been a lot of debates online about what the shape of the compute forward curve means. A good chunk of it is in contango, that is, upward sloping. Here’s a screenshot from a Silicon Data presentation in October:
In a typical market, this structure would — perhaps somewhat counterintuitively — be considered bearish for prices. It’s a signal to producers to hold supply in reserve, because there’s no immediate need for it right now.
If demand were really hot, people would be willing to pay more now for a commodity than in the future. That’s backwardation — the downward-sloping curve. It’s a signal that if you have any available inventory, now’s the time to take advantage and sell it.
AI bulls have argued that the compute curve is different because there really isn’t any inventory. And that the shift higher across the curve, with longer-term compute has gained more, is therefore a signal that contango is bullish — a signal that suppliers of compute think prices are going higher.
Microsoft’s action hints at an alternative explanation: that the conventional reading of commodity curves might be appropriate here as well, and that traders are pricing in a more balanced compute market within the next 18 months or so.
Seen on Socials
Via Kevin Gordon on X:
