RISC-V is quietly eating the low end of computing while x86 looks away

In 2026, RISC-V crossed an estimated 25 percent of global market share across application processors, microcontrollers, and AI accelerators combined — the first time in twenty-five years that the x86/ARM duopoly has had to share the label “pillar of computing” with a third architecture. Nobody in the x86 world is losing sleep over it, and that's exactly the point. RISC-V didn't win by beating Intel or AMD at high-performance desktop computing. It won by taking territory nobody with a royalty-bearing instruction set architecture was willing to defend cheaply enough.

The instruction set is free, and that changes the math for everyone below the flagship tier

ARM licenses cost real money per unit at volume, and Arm Holdings' pricing model has only gotten more aggressive since its 2023 IPO put shareholder pressure on royalty rates. For a flagship smartphone chip, that licensing cost is a rounding error against the value of the device. For a $0.30 microcontroller going into a light switch, a thermostat, or a garden sensor, that same royalty can be a meaningful fraction of the bill of materials. RISC-V's open, royalty-free instruction set architecture removes that line item entirely, and at high volume — the kind semiconductor companies shipping billions of embedded units care about — removing a per-unit royalty is a bigger deal than any performance benchmark.

This is why RISC-V's strongest gains haven't come in flagship phone chips or laptop processors, where ARM and x86 still dominate. They've come in microcontrollers, where RISC-V has already displaced ARM's Cortex-M line across several product ranges, and in embedded and IoT silicon, where the volume is enormous and the per-chip margin is thin enough that a free instruction set is worth a genuine design effort to adopt.

Automotive is the segment to watch

Infineon, one of the largest automotive semiconductor suppliers in the world, has committed to RISC-V for portions of its automotive microcontroller roadmap — a decision with outsized signal value because automotive chip qualification cycles run five to seven years and suppliers don't switch architectures on a whim. Once a RISC-V design wins a automotive design slot, it's locked in for the life of that vehicle platform, which is exactly the kind of durable, high-volume commitment that turns an architecture from “interesting alternative” into “default choice for the next RFP.” China's domestic chip ecosystem is moving in the same direction for reasons that are as much geopolitical as economic: Xiangshan, an open-source RISC-V core project, and Ruyi OS represent a deliberate bet on an architecture that isn't subject to US export control the way advanced ARM and x86 licensing can be.

What RISC-V still isn't winning

None of this means x86 or ARM are under near-term threat in the segments that actually generate the revenue: server CPUs, high-performance laptops, and flagship mobile silicon. Tenstorrent's Ascalon-X core and a handful of application-class RISC-V designs are pushing into that territory, but they're competing against architectures with decades of compiler optimization, software ecosystem maturity, and operating system support that a from-scratch instruction set can't replicate overnight. RVA23, the RISC-V profile that standardizes the application-processor feature set, is a necessary step toward that competition — without a stable, well-supported application profile, software vendors won't invest in optimizing for RISC-V the way they have for ARM's AArch64. That standardization work is real progress, but it's infrastructure, not yet market share.

The lesson for hardware buyers and chip designers

The RISC-V story in 2026 isn't “open source beats proprietary” as a general principle — it's a much narrower and more useful lesson: in high-volume, cost-sensitive, long-lifecycle segments, a royalty-free instruction set has a structural cost advantage that compounds at scale, and that advantage is now large enough to move automotive and industrial suppliers who don't switch architectures casually. If you're specifying silicon for a high-volume embedded product, RISC-V now deserves a real bill-of-materials comparison, not just an interesting-but-immature label. If you're evaluating application processors for anything user-facing, the ARM and x86 ecosystems still have a multi-year software and tooling lead that a raw architecture comparison won't capture.