Arm Holdings traded at $452 in June and just under $244 when Jim Cramer introduced this interview, a fall of 46%. Rene Haas spent the segment saying nothing in his order book has changed.
The market sold chip stocks on the idea that the frontier AI labs might slow themselves down. Haas's answer was that demand across the edge, automotive, robotics and the data center has never been higher, and that the thing limiting Arm is not orders but the ability of the supply chain to deliver.
"Demand is off the charts as we talked about earlier. And the big constraint is going to be supply."
Haas runs the company whose designs sit inside a large share of the world's processors, and he told Cramer his confidence in Arm's own $2 billion chip business has gone up at every earnings call this year.
The full segment is covered here so you can skip it.
Here are the 9 numbers that matter.
👤 Guest: Rene Haas, CEO of Arm Holdings
🎙️ Host: Jim Cramer, who presents CNBC's Mad Money and runs the charitable trust he discusses on air
🧩 Other segments: Nikesh Arora of Palo Alto Networks and Todd McKinnon of Okta
📰 Published: 16 September 2026 on YouTube (CNBC)
🔴 YouTube | ⏱️ length not available
Key Takeaways
The stock is 46% below its June high and still worth double what it was at Haas's last appearance
Arm's confidence in its $2B chip business has risen at each of the last two earnings calls
He said he is more confident in September than he was in July, and July was better than May
A humanoid robot carries hundreds of CPUs spread across multiple processing clusters
Head, torso and limbs each get their own, with clusters of 10, 20 or 60 CPUs
The shortage Haas expects is not wafers alone but substrates, testers and memory together
Customers are not objecting to Arm competing with them — they are asking it to ship faster
Haas called the collective-slowdown argument sensational and not grounded in logic
1. Down 46%, Still Doubled
Cramer set the interview up with the round trip the whole AI hardware group has made this year.
"The whole AI data center cohort peaked in June, then most of them bottomed near the end of July before rebounding like crazy." Some are still below their highs.
Arm is one of them. The stock reached $452 in June: "Now pulled back to just under 244. That's down 46%."
Cramer's point was the other frame. "It doubled from the last time I saw our next guest."
The proximate cause of the latest fall was the safety debate. "People are worried about a self-imposed slowdown in the data center come from the big frontier AI labs, which is why ARM got slammed on Monday" — Cramer said he sees a multi-year chip shortage regardless.
2. Demand at a Record
Cramer opened by asking Haas to be the evidence for the argument he had made at the top of the show: that some companies have demand strong enough to ignore a quarter-point move.
Cramer's framing was explicit. "And what I say is you have to look at companies where demand is so great that a quarter percent really doesn't stop the buyers." He said it makes little difference to Arm's customers whether rates are three and three-quarters or four.
Haas would not do the arithmetic on rates and said so. "I don't know how to exactly link what chairman Warsh did today in terms of our overall demand in terms of percentages but what I can say Jim is that demand for our technology"
What he would say is that it spans every end market Arm sells into — the edge, automotive, robotics and the data center — and that it has never been stronger.
"So, yeah, demand for us, it's just never looked better."
3. The $2B Number Improved
Cramer asked about the chip business Arm has been building without owning foundry capacity, and whether the target had come down.
The original disclosure was a visibility number, not a forecast. "So what we said in the earnings call, I think it was May time frame, that we had visibility to $2 billion."
Confidence in it rose between the May and July calls, on Haas's account, and has risen again since.
"Here I am in September and what I can tell you is Jim, I'm more confident today than I was on that July earnings."
Cramer called it an inflection, and Haas repeated the point rather than hedging it: "We're feeling really good about it."
4. Not Grounded in Logic
Cramer put the week's slowdown argument to Haas and said he is in the Jensen Huang camp: the companies will be responsible, there will be guardrails, and the build-out continues because demand is too strong.
Haas answered from product liability rather than from policy. "My view on it is that ultimately any company that's putting out a product needs to stand behind the quality, reliability, and safety." He called it "That's a kind of a fundamental table stakes issue."
He allowed two things that make AI different. "I think what's different here is that number one, AI has got this mystical quality about it that people get really afraid about." And: "And secondly, these are young companies. These are young companies developing products really, really fast."
His conclusion is that engineering discipline solves it and a collective pause does not. "But the answer is not let's all collectively tap on the brakes because I don't even know how one product solves that versus another."
Cramer's pushback was that people keep telling him the warnings must be true because nobody would say them otherwise, including the claim that everyone dies by 2030.
Haas answered with his own company as the counter-example. Arm ships thousands of products and 350 billion chips, and has to stand behind their quality; he asked what it would look like if its products caught fire.
The analogy he chose was a junior employee. "And if we had an intern who worked at the company 6 months and looked at the quality process we had and then came out and said all of our products are going to blow up and everyone's going to die. It's just irresponsible."
"So I found sort of the knee-jerk to all this a little sensational but not grounded in logic."
5. Hundreds of CPUs a Robot
Cramer asked how many processors a good robot or a self-driving car actually needs. The answer is the most concrete thing in the segment.
"A lot. And it starts with the distribution of where the CPUs are."
The architecture is several brains, not one. "You've got CPUs that are the brains of these machines." And: "So, literally in a robot, you'd have a couple of very large CPU clusters."
A cluster is itself many processors. "And when I say CPU clusters, these are chips that could have 10 CPUs, 20 CPUs, 60 CPUs."
The processors are distributed through the body. Something in the head, something in the torso, and then sensors, controllers and perception hardware out in the limbs.
"Hundreds and hundreds of CPUs in robots." Haas said an automobile is similar, with literally hundreds of CPUs.
6. Wafers Aren't the Only Limit
Cramer's follow-up was whether the world can physically make that many processors. Haas said the bottleneck is broader than the one everybody watches.
"Demand is off the charts as we talked about earlier. And the big constraint is going to be supply."
The obvious constraint is wafer capacity for logic chips. That is not where he stopped.
"But it's not just the wafers, Jim. It's the substrates. It's the testers. It's the memory. It's the entire supply chain."
He put a duration on it. "So we are looking at some really complex supply chain issues, I think, for the next number of years."
7. Can You Make More?
Cramer raised the awkwardness of a company that licenses designs to everyone now building and selling its own chips.
Haas said the objection he prepared for never arrived. "Our problems have been far more the latter than the former." The team had rehearsed answers to customers accusing Arm of encroaching on their space.
What customers actually ask is about supply. "The questions we get is, can you make more? Can you deliver them faster? When is your next generation coming out?"
"And there's plenty of room for everybody."
His example is Nvidia. "I mean, Jensen builds a CPU based on ARM. He's talked about fantastic demand for Vera." Haas added: "So, the world's pretty big and there's a lot of space for all of us."
8. No Longer a Phone Company
Cramer asked whether the market still thinks of Arm as a mobile business, and said that is why he thinks it is undervalued.
Haas conceded the history and said the mix is about to flip. "You know that's the history of the company but in very short time data center is going to be our largest business."
He said the data center business has grown at double-digit rates for several years.
The positioning claim is the broadest one he made. "Really the thing about ARM is we are the compute platform for AI. Every AI application is going to run through ARM one shape or another."
9. Beat Every Quarter So Far
The close was about how much weight to put on the language Haas had just used.
"We've been very conservative traditionally about how we talk about the numbers" — which is why he said the sequence of improving confidence should be read as meaningful.
He put his record behind it. "I hope people take that to the bank because we have essentially I think beat every single quarter since we've been public."
Cramer's takeaway tied the segment back to the Fed. "Well, that's why I want people to understand just because the Fed raises rates doesn't mean it's the end of the world." And: "There's some people have such strong demand that their stocks continue a pace."
Bonus Insights
Cramer disclosed a trade he regrets. "The trust had a great position." It made a lot of money in Arm and then, in his words, "we took profits and that was a big mistake".
Haas's standard for a technology company is the same standard he would apply to any manufacturer. Governments do not always get quality right, he said, and neither do companies — which is why he treats it as hygiene rather than as an AI-specific problem.
The ordering of the two interviews mattered to the guests themselves. In the Palo Alto Networks segment, Nikesh Arora cited Haas's view that the market is large enough for everyone as the reason he expects the same to hold in cybersecurity.
Haas's bottom line is that the AI hardware trade is being priced off a debate about whether the labs slow down, while the actual constraint on Arm's business is a supply chain — wafers, substrates, testers and memory — that he expects to stay tight for years, with robots and cars adding hundreds of processors each on top of the data center.
Products, Companies & Tools Mentioned
Arm Holdings (Haas's company: designs the processors, now sells its own chips, and expects the data center to become its largest business)
Nvidia (Builds a CPU on Arm and has talked about strong demand for it, which Haas uses as evidence there is room for both)
Watch the full episode:
If this was worth your time, send it to someone closer to the industry than you are.
Get the latest market chatter as it happens: