The financial press is currently tripping over itself to crown Arm Holdings as the undisputed king of the silicon renaissance. They see a quarterly earnings beat and mistake a toll booth for an engine. The narrative is simple: Arm is winning because it’s everywhere, from your pocket to the data center, and the shift to v9 architecture is a goldmine.
They are wrong.
Arm isn't winning a race; it’s managing a legacy estate that is increasingly under siege by the very "partners" it claims to empower. What the spreadsheets call "growth," I call a desperate squeeze of a captive audience before the walls close in. If you think Arm’s current trajectory is a sustainable ascent, you aren't looking at the instruction sets; you’re looking at the marketing deck.
The v9 Royalty Trap
Everyone is buzzing about the shift from Armv8 to Armv9. The "bull case" is that Arm gets to double its royalty rates because the new architecture is more complex and essential for AI.
Let’s be clear: Armv9 isn't a revolutionary leap in computing power. It’s a licensing maneuver. By bundling SVE2 (Scalable Vector Extension) and tighter security features, Arm has forced a migration. If you’re a chip designer today, you aren't buying v9 because it’s a "game-changer" (to use a tired cliché I despise); you’re buying it because the ecosystem demands it for modern Android and iOS compatibility.
This is rent-seeking, plain and simple. I’ve seen this movie before in the software world. When a platform holder realizes they can’t innovate fast enough to justify new sales, they tweak the terms of the lease. Arm is currently extracting more value per chip, but they are doing so by burning the goodwill of their biggest customers. Apple, Qualcomm, and MediaTek are not "partners"—they are high-net-worth prisoners looking for a way out.
The RISC-V Ghost in the Machine
The "lazy consensus" ignores the open-source specter in the room. The argument usually goes: "RISC-V is too fragmented; it lacks the software ecosystem to threaten Arm."
That’s a 2018 argument. In 2026, it’s a lie.
The move toward RISC-V isn't about saving a few cents on royalties. It’s about sovereignty. When Nvidia tried to buy Arm, the industry woke up to a terrifying reality: their entire product roadmap was owned by a single entity that could be sold to a competitor. Even though the deal failed, the trauma remains.
Companies like Meta, Google, and Amazon are now designing their own custom silicon. They are starting with "boring" stuff—NICs, SSD controllers, and management processors. But they are building the muscle memory to move to the main CPU. Every time Arm hikes a royalty rate to please Wall Street, the ROI on switching to RISC-V looks better. Arm is effectively subsidizing its own replacement by being too greedy, too early.
The Myth of the Data Center Conquest
You’ll hear analysts rave about Grace, Graviton, and Cobalt. They point to these Arm-based server chips as proof that the x86 duopoly (Intel and AMD) is dead and Arm is the new standard.
Wait.
Arm doesn't build these chips. They provide the blueprints. The real value—the branch predictors, the cache hierarchy, the interconnects—is being designed by the cloud giants themselves or by specialized shops. Arm is getting a fraction of the value created in the data center while taking a disproportionate amount of the PR credit.
Furthermore, the data center is a brutal environment for a licensing-only model. Hyperscalers don't want "standard." They want "optimized for our specific workload." Arm’s business model relies on a degree of standardization to maintain high margins. The more Amazon or Microsoft tweaks the architecture to fit their specific AI training needs, the less they need Arm’s generic "off-the-shelf" cores. They’ll take the Instruction Set Architecture (ISA) license, sure, but they’ll stop buying the Cortex designs. That’s a margin killer that nobody is talking about.
The AI PC is a Marketing Hallucination
The latest quarterly hype is centered on the "AI PC." The idea is that NPU-heavy (Neural Processing Unit) laptops will spark a massive refresh cycle, and since Windows on Arm is finally "good," Arm will devour the laptop market.
This ignores the fundamental physics of software.
The Windows ecosystem is a sprawling, chaotic mess of legacy x86 code. Emulation has improved, but it is never free. For the average corporate buyer, the risk of a mission-critical Excel macro or a proprietary VPN client breaking on an Arm-based Surface is a non-starter.
Intel and AMD aren't sitting still. They’ve integrated NPUs into their own chips. The "power efficiency" lead Arm once held is evaporating as x86 moves to 3nm and below. Arm’s advantage wasn't some magical property of its instruction set; it was a head start on mobile-first design. Now that the playing field is leveled on process nodes, the x86 giants are fighting back with the one thing Arm can’t replicate: 30 years of total software dominance.
The Hidden Cost of the "Total Compute" Strategy
Arm is trying to move away from selling individual "cores" to selling "subsystems." They want to sell the whole compute block—CPU, GPU, and NPU—pre-integrated. They call this "Total Compute Solutions."
In reality, this is a defensive play.
Arm’s GPUs (Mali/Immortalis) have always been "fine," but they aren't world-beaters. By bundling them into a subsystem, Arm is trying to lock out third-party IP providers like Imagination Technologies or even the internal teams at companies like Samsung.
When you start forcing bundles on your customers, it’s a sign of weakness, not strength. It means your individual components can no longer win on their own merits. I’ve talked to engineers at major Tier 1 mobile vendors who are furious about this. They want the best CPU and the best GPU, even if they come from different places. Arm is telling them, "Take our mediocre GPU, or pay a premium for the CPU alone." That is a recipe for a revolt.
The China Factor: A Ticking Time Bomb
The competitor article likely glossed over Arm China. It’s a mess. A semi-autonomous entity that Arm doesn't fully control, operating in a geopolitical powderkeg.
China accounts for a massive chunk of Arm’s revenue. Yet, the Chinese government is aggressively pushing for domestic alternatives to avoid Western sanctions. They aren't moving to x86. They are moving to RISC-V and homegrown architectures. Arm is essentially being used as a bridge while the Chinese tech industry builds its own independent infrastructure. To price Arm as if its China revenue is "safe" and "recurring" is financial malpractice.
The Valuation Disconnect
At current multiples, Arm is priced as if it’s a high-growth software-as-a-service (SaaS) company. It’s not. It’s an intellectual property firm tied to the cyclical, capital-intensive, and volatile hardware industry.
Every dollar of Arm revenue requires a customer to successfully design, manufacture, and sell a physical piece of silicon. If there’s a glut in the smartphone market, Arm suffers. If a 3nm fab goes offline, Arm suffers. SaaS companies have 90% gross margins and can scale with the click of a button. Arm has to fight for every design win in a multi-year cycle.
The market has priced Arm for perfection. But we are entering an era of "Good Enough" computing. If a $20 RISC-V chip can do 90% of what a $50 Arm chip can do, the market will shift. It happened in servers with x86 versus RISC/UNIX in the 90s. It will happen again.
The Reality of the "Compute Resurgence"
There is no "CPU resurgence." There is an AI explosion. And in the AI world, the CPU is the assistant, not the star. The heavy lifting is done by accelerators—GPUs and TPUs.
Arm’s role in the AI era is to be the "manager" of the system. It’s an important job, but it’s not where the value is being captured. Nvidia is capturing the value. The cloud providers are capturing the value. Arm is just providing the scaffolding.
The current quarterly results are a reflection of a specific moment in time: a post-pandemic recovery combined with a one-time architectural shift (v9). It is a peak, not a plateau.
If you want to understand the future of the industry, stop looking at Arm’s royalty reports and start looking at what the lead architects at Apple and Tesla are doing. They aren't trying to figure out how to use more Arm IP. They are trying to figure out how to use less.
The industry isn't "carving a path" for Arm. It’s building a bypass.
Stop buying the hype. Arm is a brilliantly managed legacy business that has reached the limits of its monopoly power. The next decade won't be about the "resurgence" of the CPU; it will be about the democratization of the instruction set. And in that world, a high-priced, proprietary gatekeeper is the first thing to be disrupted.
The toll collector's job is great until someone builds a new road.
The road is already under construction.
