Arm’s quarter shows how it’s carving a lucrative path in the crowded CPU resurgence
Arm Holdings shares fell Wednesday evening despite the chip designer reporting a better-than-expected quarter and giving an upbeat outlook for its data center CPU business. Revenue for the company’s fiscal 2026 fourth quarter ended March 31 increased 20% year-over-year to $1.49 billion, ahead of the LSEG-compiled analysts’ consensus estimate of $1.47 billion. Non-GAAP earnings per share (EPS) increased 9% to 60 cents, beating the 58 cents expected. ARM YTD mountain Arm Holdings YTD Shares of Arm dipped roughly 6% in after-hours trading, giving back about half the gains they had during the regular session. We pointed out in Wednesday’s Morning Meeting for Club members that this could happen — great numbers and a possible pullback in the stock because of the run-up ahead of the print. It’s exactly what happened. The stock closed at a record high of $237 — padding out year-to-date gains to 117%. Bottom line When we started a position in Arm last month at around $170 per share, we wanted to ensure the portfolio had exposure to the data center CPU market. See, the artificial intelligence revolution has evolved in a major way over the past six months. At first, everything was about having the best graphics processing units (GPUs) to train large language models. Then the focus shifted to inference, and now those workloads are evolving again, from handling human-generated prompts to supporting continuous, agent-driven tasks. While GPUs still have a critical role to play in the future of AI, the once left for dead central processing units (CPUs) are having a major moment. This CPU renaissance was confirmed when Intel reported two weeks ago. Intel CEO Lip Bu Tan said on the April 23 earnings call that the CPU-to-GPU ratio in AI racks used to be 1-to-8. But with the rise of agentics, it’s more like 1-to-4 — and in the future, it could be parity, meaning 1-to-1. In other words, a lot more CPUs are needed than a few years ago. Advanced Micro Devices told a similar story on its earnings call Tuesday night. Quantifying how big the CPU market is getting, AMD CEO Lisa Su said she now expects the CPU server total addressable market to grow at a greater than 35% clip annually, reaching over $120 billion by 2030. In an interview with Jim on CNBC on Wednesday, Su said , “Agents are really driving tremendous demand in the overall AI adoption cycle.” It’s hard for a stock to go up three times on the same information, so we’re not surprised to see Arm give back some of its recent parabolic gains. However, we thought the post-earnings call solidified our thesis. Arm-based CPUs represent more than 50% share among top hyperscalers. AMD and Intel may claim they have the market share edge, but Arm pointed out on the call that the three largest AI accelerator providers pair their chips with Arm-based ones. Nvidia ‘s Rubin GPUs are integrated with Vera (Arm-based) CPUs; Google has its Tensor Processing Units (TPUs) with its Axion (Arm-based) CPUs, and Amazon has Trainium with the Graviton (Arm-based) chips. All three are also portfolio names. “Whether it’s Nvidia, whether it’s Amazon, whether it’s Google, the very largest and most prevalent accelerators by volume are the TPU, it’s Trainium, and it’s Rubin. … Those all connect to Arm,” CEO Renee Haas explained on the call. TPUs from Alphabet ‘s Google are co-designed by fellow Club name Broadcom . Why we own it Chip designer Arm is at the center of the CPU revival. The move from AI training to running the models has reignited demand for central processing units. Arm has lucrative licensing and royalty businesses for its chip architecture, which is widely used by major hyperscalers. In March 2026, Arm unveiled the next chapter in its story: the company’s first in-house data center CPU, designed specifically for agentic AI workloads. Competitors : Advanced Micro Devices , Intel Most recent buy : April 20, 2026 Initiated : April 20, 2026 The biggest players in AI are increasingly favoring Arm-based CPUs over traditional x86 processors, an architecture dominated by AMD and Intel, because of their performance advantages and greater efficiency. While Arm’s business model has traditionally centered on collecting upfront license fees and royalties tied to chip shipments, the new leg to the story is the development of its own chip. The customer response to the ARM AGI CPU looks terrific. When introducing its first-ever in-house data center CPU at its Arm Everywhere event back in March, the company said it had a line of sight to more than $1 billion of demand over the next two years. It hasn’t even been two months, and management has already doubled this view. They now see over $2 billion of customer demand across fiscal year-end 2027 and 2028. However, they did soften this upbeat guide slightly by noting they are maintaining the initial $1 billion outlook because they have to line up the supply chain capacity to meet the demand. Concerns over these supply constraints are what caused the stock to give up its initial pop after hours. As we said when we first added Arm to the portfolio, the company has a great sales pitch with its CPU. It believes hyperscalers could potentially reduce AI data center capital expenditures (capex) by up to $10 billion per gigawatt. That’s everything, given the market’s focus on free cash flow. The longer-term target is still $15 billion in fiscal year-end 2031, and these sales are not expected to cannibalize Arm’s existing business, which is an important push back to a bear thesis. “The primary reason we did this,” Haas said, in reference to developing its own chip, “was that our customers asked for it. At the end of the day, we are responding to customer demand in a market.” The bottom line is that demand for Arm-based data center CPUs is off the charts and supportive of strong double-digit revenue growth for the foreseeable future. The story gets even better with the success of its in-house chips, and now it’s up to management to navigate a tight and complex supply chain environment to over-deliver on its goals. We’re maintaining our price target of $250 and hold-equivalent 2 rating, given the recent parabolic move in share price. In the short time since we put Arm into the portfolio, the stock has gained nearly 40% as of Wednesday’s close. If the after-hours move holds, we’ll be giving back some of that advance. But the rally in Arm shares our April 20 initiation and in 2026, for that matter, has been nothing short of incredible. Commentary As for the quarterly results, License and Other revenue grew about 29% year over year to $819 million, beating Street estimates. These revenue streams are from the upfront license fee the company collects from customers who want access to its CPU architecture and designs. Royalty revenue increased 11% year over year to $671 million, but that actually missed what the Street expected. However, the shortfall was probably due to the smartphone market. This piece of the business still grew year over year, but there’s weakness in the end market due to the memory shortage. More importantly, the company saw an accelerated ramp of Arm-based server chips by all major hyperscalers, as well as increased deployment of data center networking chips. We were also pleased to see Arm’s gross margins and operating margins come in better than expected. Arm’s current revenue streams are all from licenses and royalties, creating some extremely attractive gross margins. They were 98.32% on a non-GAAP basis in the quarter. (GAAP stands for generally accepted accounting principles. Non-GAAP, sometimes referred to as adjusted, strips out one-time factors in hopes of delivering an apples-to-apples comparison from quarter to quarter.) Non-GAAP operating margins were better than expected, too, and we should see more operational leverage in the future as cost growth decelerates from a 26% compound annual growth rate (CAGR) in fiscal year 2024 through fiscal year 2026 to a mid-teens CAGR from fiscal year 2026 through fiscal year 2031. Outlook Arm provides guidance on a quarterly basis. For the first quarter of fiscal year 2027, the company expects revenue of $1.26 billion plus or minus $50 million, meaning a range of $1.255 billion to $1.265 billion. That’s slightly better than the consensus estimate of $1.25 billion, according to LSEG. (However, that would be lower sequentially as fiscal Q4 was $1.49 billion.) The company expects non-GAAP operating expenses of $760 million, which is a little higher than the FactSet consensus estimate of $742 million. Non-GAAP earnings per share are expected to be 40 cents, plus or minus 4 cents, meaning a range of 36 cents to 44 cents. This is above the consensus estimate of 36 cents, according to LSEG. (Jim Cramer’s Charitable Trust is long ARM, NVDA, GOOGL, AMZN. See here for a full list of the stocks.) As a subscriber to the CNBC Investing Club with Jim Cramer, you will receive a trade alert before Jim makes a trade. Jim waits 45 minutes after sending a trade alert before buying or selling a stock in his charitable trust’s portfolio. If Jim has talked about a stock on CNBC TV, he waits 72 hours after issuing the trade alert before executing the trade. THE ABOVE INVESTING CLUB INFORMATION IS SUBJECT TO OUR TERMS AND CONDITIONS AND PRIVACY POLICY , TOGETHER WITH OUR DISCLAIMER . NO FIDUCIARY OBLIGATION OR DUTY EXISTS, OR IS CREATED, BY VIRTUE OF YOUR RECEIPT OF ANY INFORMATION PROVIDED IN CONNECTION WITH THE INVESTING CLUB. NO SPECIFIC OUTCOME OR PROFIT IS GUARANTEED.
About the Author
