Arm Holdings Chief Financial Officer Jason Child is capitalizing on surging demand for the company's own processor designs and newly launched AI chips, which have doubled the UK-based firm's stock price this year. The rally has bolstered the acquisition firepower of the SoftBank-owned company, now valued at roughly $300 billion.
Arm has historically favored smaller deals, such as last year's $265 million purchase of networking startup DreamBig. In an interview, Child indicated the company will likely continue this approach, though he did not rule out larger transactions. Arm is no longer merely licensing chip blueprints to the likes of Nvidia and Apple; it now sells finished, self-developed chips directly. This shift has opened new growth avenues but also introduced a fresh set of challenges.
"Delivering finished silicon is far more complex than simply licensing a design," he said. Arm announced in March that it would commercially sell AI-specific CPUs, with Meta and OpenAI already onboard as customers. The company posted $4.9 billion in revenue last fiscal year and projects customer demand for the new chip to exceed $2 billion in fiscal 2027 and 2028 combined. The AI chip arena is now crowded with entrants including Microsoft, Meta, OpenAI, and Anthropic.
Child noted that Arm's timing for the AI CPU launch is opportune. The viral popularity of the open-source AI agent framework OpenClaw has fueled a boom in AI assistants. These agents require cloud providers to deploy Arm-based CPUs in large numbers, relying on software tools to execute long-running tasks, which demands close coordination between memory and CPU subsystems. In servers used for training large AI models, Arm CPUs also work alongside Nvidia GPUs. The surge in CPU demand benefits chipmakers like Intel, AMD, and Nvidia but intensifies supply pressure on underlying components such as memory.
As a newcomer to direct chip sales, Arm can secure only limited wafer capacity from foundries like TSMC. "The market opportunity is far larger than we initially expected when we launched this business," Child said of the AI chip venture. "But the challenge is that we now face various supply constraints head-on." Child, who joined Arm in 2022 after stints at Amazon, also discussed semiconductor supply chain shortages, whether Arm might follow Intel's lead in issuing new shares, and the company's relationship with parent SoftBank. (This interview has been edited for length and clarity.)
Q: Arm traditionally operated a light-asset model with high gross margins, yet now you're selling finished chips. What new financial risks does that entail?
"When you manufacture chips, you have to queue up for capacity. The standard logic foundries use to allocate capacity is: they look at how much you received last year and give you a small increment based on overall capacity expansion. If you're starting from zero, your share of capacity is very small. Ramping up takes years. We need to gradually secure our own foundry capacity, memory capacity, and so on over several years. That's a real challenge. Another hurdle is rapidly rising prices: demand far outstrips supply. That means without substantial capital, it's tough to compete in this space. The big spenders today are mostly large cloud providers, backed by strong free cash flow and highly profitable business models. That makes it extremely difficult for smaller players to enter, which is why you see companies like Groq seeking large deals. To achieve volume production, you must raise massive capital and credit lines, taking on huge capital expenditures. So our financial guidance is conservative, giving ourselves several years to scale gradually."
Q: How do you view the current CPU shortage?
"Compute demand keeps climbing. The biggest issue is actually how much electricity the companies pouring money into capital expenditures can secure. Chips are just one piece; you need the full system, data centers to house the hardware, and of course power to generate tokens. From what I observe, compute demand is virtually insatiable. The practical constraint is: even if you buy the chips, you still have to deal with severely tight memory supply; you need to ensure chips can be manufactured, which mainly depends on TSMC's capacity; and finally you need to confirm power availability. We're just one link in the chain, but right now, there's no sign of compute demand slowing."
Q: What is the most critical supply chain bottleneck, and how can Arm address it?
"Memory shortages rank at the top. Meanwhile, TSMC has repeatedly discussed wafer capacity and output limits, given the limited number of fabs that can meet the advanced process requirements and operate at full utilization. Many large model companies and cloud providers are planning data centers in the 1-5 gigawatt range. That's an enormous draw on electricity. The follow-on question becomes: where will these data centers be located, and can they secure the necessary power? Everyone predicts power will remain tight for years to come."
Q: Arm is no longer purely an IP licensor and is more deeply involved in full-chip development. How are your investment priorities shifting?
"When planning our business layout, we prioritize which capabilities we can build in-house and which must be filled through acquisitions. I look at M&A in two dimensions: first, whether a deal helps expand the total addressable market; second, whether it strengthens in-house capabilities to improve gross margins. In semiconductors, it's well known that no company can do everything from end to end entirely on its own. Nvidia has come relatively close to full-stack in-house development, and Apple is similar. Broadcom, Nvidia, and Apple sustain high margins precisely because they control many critical parts of the value chain. We're still in the early stages of transformation, and most of our acquisitions so far have focused on expanding the total addressable market. Our latest deal was DreamBig, a networking technology company. Arm hasn't had deep expertise in networking, so this acquisition directly broadens our market. We mostly target small and mid-sized private companies. Over the past 15-20 years, Arm has completed about 20 acquisitions, mostly in this category, and the strategy has worked well for us. We rarely acquire large public companies. Of course, we continuously evaluate various potential opportunities."
Q: Reports suggest Arm made a late bid for Cerebras last year and considered a major merger with Marvell Technology. How will you screen future M&A targets?
"We evaluate every opportunity. Some come to us unsolicited; others we pursue actively. Essentially, all possibilities are on the table. Large-scale mergers are within our capability. SoftBank holds 87% of Arm, and Masayoshi Son is our chairman. Looking at SoftBank's history, it acquired Graphcore; at the time, SoftBank evaluated a broad range of XPU special-purpose processor companies, including Graphcore, SambaNova, Cerebras, and Groq. All of these were on the review list. Arm and SoftBank maintain close ties, and Arm CEO Rene Haas now also serves as CEO of SoftBank Group International. How does Arm collaborate with companies within the SoftBank ecosystem that can complement its work?"
"Rene's dual role as CEO of SoftBank Group International makes sense. That operating group oversees all assets under Son's vision for artificial super intelligence (ASI). Rene's core job is to develop chips and systems to fulfill part of the Stargate project Son has outlined. Son holds the top-level vision, handling capital allocation and group strategy; Rene drives execution. This setup is highly convenient for us, making projects more efficient. There's a designated person within SoftBank overseeing related operations, so we don't need layer upon layer of approval. We can go directly to Son. As you can sense, the AI industry is iterating at unprecedented speed, presenting enormous opportunity, but it also puts pressure on the entire system. So we need to build a unique operating collaboration model."
Q: Even highly profitable cloud giants use high stock prices to issue new shares. Could chipmakers, or even Arm, follow suit?
"There's nothing to announce publicly right now. But looking at market patterns: when the broader market declines, companies tend to buy back stock; when share prices are at historical highs and valuations are elevated, they're more inclined to issue new shares. These options are certainly on our list of considerations. Thanks to our relationship with SoftBank, which is experienced in capital operations, we've chosen to collaborate on several high-risk, capital-intensive projects: licensing technology to SoftBank, providing design services, and developing custom products for them. SoftBank has now become a significant customer for some of our AI products, which will be disclosed later. This model allows us to avoid heavy debt. Without a partner like SoftBank, this path would be very difficult."
Q: Chip companies are increasingly forming financing partnerships with Wall Street institutions, spawning novel and complex financing structures. What's your take?
"The scale of capital being deployed right now is unprecedented. I was at Amazon from 1999 to 2001, living through the dot-com boom. Back then, e-commerce had a bright future, but the actual user experience was poor; it took a week or two for online orders to arrive. Many companies raised money but didn't fully absorb the capacity. The hardware and capex invested back then took 15-20 years to be fully utilized. The current AI wave is different: capital is put to work immediately. The real core question is how business models will evolve and where returns on investment across industries will ultimately land. Looking at capital flows, the money behind it is mostly from cash-rich giants like Amazon, Meta, Google, and Microsoft. Even in a worst-case scenario, the downside is large asset write-downs, but the companies themselves won't collapse."
Comments