Jim Cramer Warns Off the AI Data Center Trade as Nvidia and Broadcom Split
Jim Cramer says he won't touch the AI data center trade yet, as Nvidia falls 5.22% to $169.35 and Broadcom rises 3.21% to $329.72 on November 10, 2026.
Why it matters
- On November 10, 2026, Jim Cramer said on CNBC he doesn't want to be 'that linked to the data center yet.'
- NVIDIA fell 5.22% to $169.35 while Broadcom rose 3.21% to $329.72 that morning.
- NVIDIA's Data Center segment generated $89.02 billion of $96.22 billion in quarterly revenue, about 93%.
- Broadcom's AI semiconductors reached 56% of fiscal Q3 revenue, up from 49% in Q2.
- NVIDIA's fiscal 2028 revenue estimates range from $506.48 billion to $778.47 billion.
Jim Cramer said on CNBC on November 10, 2026 that he's staying out of the AI infrastructure trade: "I don't want to make this trade because I still feel that you don't want to be that linked to the data center yet." He made the call as the two biggest chip names in the buildout moved in opposite directions. NVIDIA (NASDAQ:NVDA) fell 5.22% to $169.35 in Tuesday morning trading, while Broadcom (NASDAQ:AVGO) rose 3.21% to $329.72.
The split matters because both companies sit at the center of the AI spending boom, and investors are still arguing about how durable that spending is. Cramer's warning is a bet on caution, not a call that the boom is over.
What does 'linked to the data center' actually mean?
Being linked to the data center means a stock's earnings rise or fall with how much a small group of cloud providers and AI labs spends on servers. That customer base is narrow, and its spending decisions are lumpy. If a handful of buyers pause, the revenue disappears fast.
NVIDIA is about as linked as a company can be. In the quarter it reported in August, Data Center brought in $89.02 billion of $96.22 billion in total revenue — about 93%. Its supply obligations reached $279 billion, mostly memory bought for Vera Rubin. Days sales outstanding increased to 60 days from 45. Those are large commitments to make before anyone knows whether demand will hold.
Broadcom has more margin, and that margin is shrinking. AI semiconductors made up 56% of total revenue in fiscal Q3, up from 49% in Q2, according to the company's 8-K earnings release. Broadcom has six XPU customers and has set up financing vehicles for two of them, Anthropic and OpenAI.
CEO Hock Tan said most AI computing demand "originates from this concentrated group who develop state-of-the-art frontier models." He also raised a deployment risk: "Even as we ship the chips, will they be deployed on a timely basis?" That's the concentration problem in one sentence: a handful of frontier labs drive the numbers, and the physical buildout has to keep pace.
How much have the stocks already fallen?
NVIDIA is down 26.14% over the past month and 8.98% year to date. Part of that drop has nothing to do with data center demand. NVIDIA opened an investigation into its chief executive's relationship with a suspected Chinese spy, and the stock fell for a second day as investors weighed what it could mean, according to The New York Times and CNBC. Other chip stocks rallied during the same stretch, CNBC reported.
Broadcom has held up better. It's down 8.8% over the month and 4.21% for 2026. The divergence between the two names on November 10 — one falling, one rising — shows investors aren't treating AI chip exposure as a single trade anymore.
Why do the bulls disagree with Cramer?
The other side of Cramer's argument rests on valuation. NVIDIA trades at a trailing P/E of 36. Against consensus EPS of $15.9146 for the fiscal year ending January 2028, the multiple falls to about 11x. That EPS estimate has grown from $12.8148 90 days ago — analysts keep revising earnings up, which mechanically compresses the forward multiple.
Broadcom trades at roughly 17x fiscal 2027 consensus EPS of $19.4651. Management expects AI revenue of about $115 billion in fiscal 2027 and $230 billion in fiscal 2028, and says it has already secured supply for both years.
Jensen Huang sees the same demand picture his customers do. "AI has reached its inflection point. It's doing useful work. Its tokens are productive and profitable. Now, compute is revenue," he said.
The catch is that those low multiples only hold if the estimates turn out right, and the estimates vary widely. NVIDIA's fiscal 2028 revenue estimates run from $506.48 billion to $778.47 billion. Broadcom's fiscal 2027 estimates run from $114.72 billion to $200.12 billion. That spread shows how unsure analysts are about where AI spending is headed — which is exactly Cramer's point.
What should investors watch in the next earnings reports?
Three specific numbers will tell us whether the trade is breaking or just pausing:
- NVIDIA: whether revenue meets its guidance of $108.0 billion, plus or minus 2%, which assumes no China Data Center computing sales. Also watch how the $279 billion in supply commitments changes. Results will be posted on NVIDIA's investor relations site.
- Broadcom: whether fiscal Q4 revenue reaches about $34.8 billion and AI semiconductor revenue reaches $21.7 billion. Also watch the gross margin guide of about 73%, which is coming down as lower-margin XPUs make up more of sales.
- Deployment timing: Tan named "land, power, and shell" as major constraints. Delays at customer data center sites would push revenue into later quarters even if orders stay the same.
The takeaway is that the next round of reports will test whether the demand story survives contact with physical infrastructure.
What's the real question for holders?
For holders, the main question is concentration. NVIDIA gets most of its revenue from data centers, so its earnings power is a direct function of a few buyers' capex plans. Broadcom has its VMware software business, which grew 29% last quarter, as a cushion — but AI chips make up a larger share of its sales every quarter, so that cushion is thinning.
Cramer's position is essentially that the uncertainty is unpriced. When fiscal 2028 revenue estimates span $506.48 billion to $778.47 billion for one company, the forward P/E of 11x is only as solid as the middle of that range. Investors who disagree are betting Huang is right that compute has become revenue — and that Tan's "land, power, and shell" constraints don't delay the number that matters.
Source: 247wallst.com
More from Sophie Lindqvist
Show full bio
Staff writer covering regulation, policy and their impact on artificial intelligence and machine learning at AI In Context.
537 articles
Related articles
- Nvidia Trades 18% Down as Washington Adds a Governance Discount
- Jim Cramer urges caution after $63 billion AI financing deal
- Micron fell 17% and Cramer says the AI sell-off isn't done yet
- Nvidia Added $150 Billion to Its Buyback. One Strategist Says the Stock Is Still Cheap
- NVIDIA Rises 4% to $175.23 Despite Baird Cut and Huang Cloud