Replace is the wrong word for what is actually happening
Can RISC-V replace Arm? For anyone building IoT hardware the honest answer is no, at least not in the winner-takes-all sense the question implies. The two architectures already share the same silicon, and the real contest is over which one wins the next design, not who gets evicted.
The question assumes a fight that is not really happening
“Can RISC-V replace Arm” is the sort of headline that does well and explains nothing. It pictures a title bout where one architecture knocks the other out and walks off with the belt. Silicon does not work like that. Instruction sets are not football teams, and a chip is not obliged to pick a side.
Arm spent three decades becoming the default for mobile and embedded compute. That is not a position you overturn with a press release and a GitHub repo. There is an enormous installed base of Arm designs, a mature toolchain, and a mountain of software that assumes an Arm core underneath it. None of that evaporates because a cheaper, more open alternative turned up. For the licensing and business-model differences underneath all this, see RISC-V vs Arm: What’s the Difference?.
They are already living in the same chip
Here is the bit the versus framing misses entirely. RISC-V is not waiting outside Arm’s products trying to get in. It is already inside them, and inside plenty of other silicon besides. Nvidia has said it uses RISC-V cores as the management controllers in every GPU it ships, handling power, boot and system housekeeping, and reckoned it had shipped over a billion RISC-V cores in 2024 alone. Qualcomm has shipped hundreds of millions of RISC-V cores in its devices. Western Digital moved its storage controllers to RISC-V years ago.
In a modern system-on-chip, mixing instruction sets is normal engineering, not heresy. An Arm application core can run the operating system while a small RISC-V core minds the power management, a second handles security, and an accelerator built on RISC-V does the on-device AI. The user never knows and never needs to. The chip does not care that its cores speak different dialects, as long as they share a bus.
RISC-V does not have to beat Arm to matter. It is already winning the parts of the chip nobody argues about, the controllers and accelerators, while Arm keeps the application core. Coexistence inside one package is the actual state of play, not some far-off compromise.
Where Arm is genuinely hard to shift
In IoT, most compute is brownfield. There are billions of Arm Cortex-M microcontrollers already designed into products, with firmware written, tested and certified around them. A working design that ships and passes compliance is not one an engineering manager rips up for the fun of switching instruction sets. The cost of change has to be paid back somehow, and “the ISA is more open” rarely balances that ledger on its own.
Arm’s other moat is the soft stuff: the compilers, the debuggers, the middleware, the decades of Stack Overflow answers, the engineers who already know it. RISC-V’s tooling has matured quickly, but Arm still has the deeper bench. For an established product line with an Arm core that works, inertia is a feature, not a bug.
Where RISC-V quietly wins instead
The action is in new designs. When a team is starting a chip from a blank sheet, with no existing Arm firmware to protect, the calculus flips. No architecture licence, no per-chip royalty, freedom to add custom instructions, and no single vendor’s roadmap to be tied to. For high-volume, cost-sensitive IoT silicon that is a strong hand, which is why the newest connected microcontrollers increasingly ship with RISC-V cores rather than Arm ones.
It also wins the jobs Arm was never really fighting for: the subsystem controllers, the accelerators, the sovereignty-driven designs in markets that would rather not depend on a foreign IP vendor at all. Add on-chip AI to that list, since Nvidia extending its CUDA platform to RISC-V CPUs signals the architecture is being taken seriously well beyond housekeeping duty.
| Scenario | Which ISA tends to win | Why |
|---|---|---|
| Existing product, working firmware | Arm | Cost of change rarely pays back |
| New cost-sensitive IoT MCU | RISC-V | No licence, no royalty, high volume |
| Subsystem and control cores | RISC-V | Small, custom, invisible to users |
| Rich application processor with big software stack | Arm today | Ecosystem and toolchain depth |
| Custom AI accelerator | RISC-V | Extensible ISA, own your design |
| Sovereignty-driven silicon | RISC-V | No dependence on foreign IP |
So, can it?
Not in the way the question wants. RISC-V is not going to evict Arm from the world’s phones and the huge embedded base that already runs on it. What it will keep doing is winning new designs, colonising the controllers and accelerators, and sitting quietly inside Arm-based products doing the jobs nobody notices. Over time that is how an architecture becomes ubiquitous, not with a knockout but by turning up everywhere.
For anyone specifying IoT hardware, the practical takeaway is to stop thinking replacement and start thinking mix. Expect both architectures in your bill of materials, often in the same chip, and pick per subsystem on cost, tooling and how much you want to own the design. Where that leaves things by the end of the decade is the subject of The Future of RISC-V: 2030 and Beyond.
Last updated 20 August 2026. Sources: Nvidia use of RISC-V as GPU management controllers and over one billion RISC-V cores shipped in 2024, per Nvidia statements reported at RISC-V events; Qualcomm RISC-V core shipments per company and industry reporting; Western Digital RISC-V storage controllers per Western Digital; CUDA platform support for RISC-V announced at the 2025 RISC-V Summit in China, per RISC-V International.
