Hardware/Explainer

What Is RISC-V? The Open Chip Architecture Challenging ARM and x86

Unlike Arm and x86, anyone can build a chip on RISC-V without paying royalties. Qualcomm's Ventana buyout, Nvidia's RISC-V cores and a new Chinese server chip show why that's starting to matter.

The RISC-V open instruction set architecture logo
Image: RISC-V International.

RISC-V is a free, open computer chip architecture that anyone can use to design processors without paying a licensing fee — unlike Arm, which charges royalties, or x86, which only Intel and AMD can build on. That openness is why Qualcomm just bought a RISC-V server-chip startup, why Nvidia already ships billions of small RISC-V cores inside its GPUs, and why Chinese chipmakers are racing to build RISC-V server silicon. Here's what RISC-V actually is, how it differs from Arm and x86, and where the open architecture is already showing up in 2026.

The short version:

  • RISC-V is an open, royalty-free instruction set architecture (ISA) — the basic rulebook a processor follows — maintained by the nonprofit RISC-V International.
  • Anyone can design and sell a RISC-V chip without a license fee, unlike Arm (licensed) or x86 (Intel and AMD only).
  • Qualcomm acquired RISC-V CPU designer Ventana Micro Systems in December 2025 to build out its server and data-center ambitions.
  • Nvidia has used custom RISC-V microcontroller cores inside its GPUs since the Turing architecture, for tasks like video codecs and power management — not the main graphics compute.
  • China's SpacemiT and the XiangShan project unveiled the 46-core "Vital Stone" V100 RISC-V server chip on October 19, 2026.
  • RISC-V is already common in embedded and microcontroller silicon; it remains far from mainstream in phones and laptops.

What Is RISC-V?

An instruction set architecture, or ISA, is the set of basic commands — add, load, branch, and so on — that software uses to talk to a processor's hardware. Arm and x86 are the two ISAs that run almost every phone, laptop and server today. RISC-V is a third option: an open ISA that started as a research project at the University of California, Berkeley in 2010 and was first published in 2011.

What makes it different isn't the instructions themselves — RISC-V follows standard "reduced instruction set computer" design principles, similar in spirit to Arm — but who gets to use it, and on what terms. RISC-V's ratified specifications are released under royalty-free, open licenses. A startup, a university lab, or a chip giant like Qualcomm can all design a RISC-V processor and sell it without paying RISC-V International a cent. The ISA is also modular: there are three base instruction sets (RV32I, RV64I and RV128I for 32-, 64- and 128-bit designs), and chipmakers bolt on optional standard extensions, or even reserve their own custom instructions, depending on what the chip needs to do.

In practice, that modularity is the whole point. A tiny sensor chip might implement only the base integer instruction set to keep die size and power draw minimal. A server CPU aimed at AI workloads might add the ratified "V" vector extension for parallel math, plus a cryptography extension for secure boot and encrypted memory. Because the base ISA never changes once ratified — RISC-V International explicitly won't modify a finished specification, only add new extensions on top — software written for a basic RISC-V chip keeps working on a much more capable one years later, the same stability guarantee Arm and x86 offer, just without the licensing step in between.

That stability-without-a-gatekeeper model is also why RISC-V has become a talking point in chip-supply politics, separate from its technical merits. Because nobody needs an export license or a licensing deal with a single company to design a RISC-V chip, it's attractive to any country or company worried about depending on a foreign-controlled architecture — which is part of why Chinese chipmakers in particular have leaned into RISC-V server and AI silicon over the past two years, alongside genuine engineering interest in its flexibility.

Who Actually Controls RISC-V?

RISC-V International is the nonprofit that maintains the specification. It's incorporated in Switzerland — the organization relocated from the US in 2020, citing concerns about trade regulations affecting an international standards body — and has received operational and strategic support from the Linux Foundation since 2018. RISC-V International says it takes no commercial stake in products built on the ISA and doesn't take political positions.

Governance runs through member companies, which sit on a Board of Directors and a Technical Steering Committee and vote on ratifying new extensions. Only members get to use RISC-V International's trademarked compatibility logo. But — and this is the key distinction from Arm or x86 — membership controls who shapes the standard, not who's allowed to build chips with it. Once a specification is ratified, it's public, free and permanent; RISC-V International says it won't support forked or alternative versions of the ISA, which is meant to keep the ecosystem from fragmenting.

RISC-V vs. Arm vs. x86

The practical difference shows up in licensing and who can build what. For a deeper look at how Arm and x86 themselves compare on performance and power, see Pandromeda's Arm vs. x86 explainer.

ArchitectureWho can build chipsLicensing costWhere it dominates today
x86Intel, AMD (via cross-license)Closed, no outside licensingPCs, most servers
ArmAny company that pays Arm Holdings for a licenseRoyalty per chip, upfront license feesPhones, tablets, most laptops
RISC-VAnyone, with no license requiredFree and royalty-freeEmbedded controllers, microcontrollers, emerging AI/server silicon

That "anyone, for free" model matters most to companies that want a custom chip without paying a licensing tax on every unit, or that want to add their own instructions for a specific job — which is exactly what's driving the three developments below.

Qualcomm's Bet: Buying Ventana Micro Systems

In December 2025, Qualcomm announced it had acquired Ventana Micro Systems, a Cupertino-based RISC-V CPU design startup founded in 2018. Ventana held seats on RISC-V International's Board of Directors and its Technical Steering Committee, and had been developing server-class RISC-V chips — including a previously disclosed design built around multi-chip clusters with up to 192 RISC-V cores and clock speeds approaching 4.0GHz. Qualcomm's own announcement frames the deal as deepening its RISC-V CPU expertise and complementing its existing Arm-based Oryon CPU line, rather than replacing it. Financial terms weren't disclosed, and Qualcomm hasn't said whether its future data-center chips will end up using Ventana's RISC-V designs, Oryon's Arm-based cores, or some mix of both.

Nvidia Already Ships Billions of RISC-V Cores — Just Not for Graphics

Nvidia's use of RISC-V is easy to misread, so it's worth being precise: the company isn't running its main GPU compute on RISC-V. Since its Turing architecture, Nvidia has used small, custom RISC-V microcontroller cores — by outside estimates, somewhere between 10 and 40 per chip depending on complexity — to run background functions like video encode/decode, chip-to-chip interconnects, power management and security. These RISC-V cores replaced Nvidia's older proprietary "Falcon" microcontrollers, which couldn't run a full operating system or address memory the way a modern design needs to. Nvidia has added more than 20 custom instruction extensions on top of the base RISC-V ISA for this purpose. RISC-V International has estimated that Nvidia shipped on the order of a billion of these cores in 2024 alone — a figure that reflects volume of small embedded cores, not a shift of GPU compute itself to RISC-V.

The newer development is on the server side. At the Hot Chips 2026 conference in August, Nvidia presented a RISC-V host-CPU interoperability profile — a published list of requirements (including the RVA23 CPU profile, RISC-V boot and runtime services, and PCIe-level coherency) that a RISC-V-based server CPU would need to meet to pair cleanly with Nvidia GPUs over NVLink Fusion and run CUDA workloads. That's a specification for compatibility, not an announced Nvidia chip — it's aimed at RISC-V CPU makers (SiFive has been named as one early partner) who want their processors to work alongside Nvidia's AI hardware.

China's Push: SpacemiT's 46-Core Server Chip

On October 19, 2026, Chinese RISC-V chip designer SpacemiT and the open-source XiangShan processor project jointly unveiled the "Vital Stone" V100, described as a server-class RISC-V processor. The chip mixes two core designs: six XiangShan "Kunminghu V2R2" cores from the open-source XiangShan project and 40 of SpacemiT's own X100 cores, for 46 cores total, built on a 12nm process with clock speeds of 2.2–2.4GHz and a 120W TDP. It supports eight-channel DDR5-4800 memory and PCIe 5.0 with 52 lanes. SpacemiT says the X100 core scores over 9 points per GHz on the SPECINT2006 benchmark and calls the V100 the first chip fully compliant with RISC-V's server platform specifications — both company claims that haven't been independently verified yet. There's no public pricing, availability date, or server-vendor commitment so far, and the design has reportedly changed since SpacemiT first described a 64-core version of the project in early 2025.

Chipmakers like SpacemiT also use custom chiplet design techniques to mix and match cores like this on one package — a broader trend across the chip industry that isn't unique to RISC-V.

Where RISC-V Is Already Big — and Where It Isn't

RISC-V's real-world footprint today is lopsided. It's already dominant in parts of the chip world most people never see: embedded controllers, IoT devices, and — as Nvidia's deployment shows — the small management cores tucked inside much bigger chips. Plenty of widely circulated claims about RISC-V's overall "market share" across all computing are unconfirmed or come from secondary blogs rather than RISC-V International or independent analyst firms, so it's worth treating any specific percentage with caution.

What's clearly still true: RISC-V has essentially no presence in mainstream phones or laptops, where Arm and x86 remain entrenched, and even its push into servers — via Qualcomm's Ventana deal, SpacemiT's V100, and Nvidia's interoperability work — is still in the early-deployment stage, with no major RISC-V server chip in broad commercial availability yet.

Why This Matters Now

The timing isn't a coincidence. AI data centers want chips tuned to very specific workloads, and that favors architectures where a company can add custom instructions without a licensing negotiation. An open, royalty-free ISA lets a hyperscaler, a national chip program, or a startup design exactly the processor it needs — which is a big part of why Qualcomm, Nvidia and Chinese state-backed chip efforts are all investing in RISC-V in parallel, even though their specific goals (CPU product lines, GPU support silicon, and server chips, respectively) are quite different from each other.

It also matters for cost. Arm charges licensing and per-chip royalty fees that scale with volume, which is a real line item for a company shipping tens of millions of units, or for a cloud provider building custom silicon purely for its own data centers rather than to sell. Skipping that fee is one reason several hyperscalers have experimented with RISC-V for narrow, high-volume internal uses — storage controllers, network interface chips, AI-accelerator management logic — even while their main CPU fleets stay on Arm or x86 for now. None of this makes RISC-V a drop-in replacement for either architecture; it mostly means companies now have a genuine third option to reach for when the economics or the engineering favor it.

What's Next

Watch three things over the coming months: whether Qualcomm says anything concrete about shipping a Ventana-derived server chip (a design called "Veyron V3" has been mentioned as arriving as early as late 2026 or 2027, though Qualcomm hasn't confirmed a product timeline since the acquisition); whether SiFive or another RISC-V CPU maker actually demonstrates a chip meeting Nvidia's new interoperability profile; and whether SpacemiT publishes independent benchmark results or lands a server-vendor customer for the V100. None of those are guaranteed to pan out on schedule — this is still an architecture proving itself in new markets, not one that's already won them. For now, the AI accelerator market that's pulling RISC-V into servers is still being fought mainly on Arm and x86 silicon, as Pandromeda's Intel Arc Pro B70 vs. Radeon AI Pro R9700 comparison shows.

Frequently asked questions

Is RISC-V free to use?

Yes. RISC-V International's ratified specifications are released under royalty-free, open licenses, so any company can design and sell chips using the ISA without paying a licensing fee, unlike Arm's per-chip royalty model or x86, which only Intel and AMD can implement.

Who controls RISC-V?

RISC-V International, a nonprofit incorporated in Switzerland with operational support from the Linux Foundation since 2018, maintains the specification. Member companies vote on ratifying new extensions, but the ratified specs themselves are free for anyone, member or not, to use.

Does my phone or laptop use RISC-V?

Almost certainly not for its main processor. RISC-V remains rare in mainstream phones and laptops, which are still dominated by Arm and x86. It's already common inside embedded controllers, and it's expanding in AI-accelerator management cores and emerging server chips.

Why did Qualcomm buy Ventana Micro Systems?

Qualcomm said the December 2025 acquisition deepens its RISC-V CPU engineering expertise and complements its Arm-based Oryon CPU line, positioning it for future data-center and edge chips built on the open ISA.

Does Nvidia use RISC-V in its GPUs?

Yes, but not for the main graphics compute. Since its Turing architecture, Nvidia has used custom RISC-V microcontroller cores for the GPU System Processor and tasks like video codecs, chip-to-chip interconnects, power management and security.

Is RISC-V replacing Arm and x86?

Not yet. It's growing fastest in embedded and specialized silicon and in new server bets from companies like Qualcomm and SpacemiT, while Arm and x86 still dominate phones, laptops and most of the data center.

Sources

More on RISC-V →RISC-VChip ArchitectureQualcommNvidiaOpen Source Hardware
Mara Lindqvist
Written byMara Lindqvist

Mara Lindqvist edits the hardware desk. She covers graphics cards, processors, memory and storage, the foundries and chip designers behind them, and what the numbers on a spec sheet mean for people choosing a PC. Specifications in her stories come from manufacturer spec pages and datasheets.

More from Hardware

See all