What distinguishes RISC-V from x86 and Arm architectures?

RISC-V is an open-source instruction set architecture, allowing companies to design custom processors without royalties or dependence on a single vendor. Unlike x86 (proprietary to Intel/AMD) and Arm (licensed), RISC-V offers total architectural flexibility and a different economic model, particularly attractive for large-scale deployments in data centers.

Why is Nvidia investing in a CPU company when it manufactures GPUs?

Nvidia recognizes that modern AI systems require tight orchestration between CPUs and GPUs. By supporting SiFive and the integration of NVLink Fusion with RISC-V processors, Nvidia expands its ecosystem beyond proprietary architectures, meeting hyperscalers' demand for open and customizable solutions while keeping its GPUs at the core of AI infrastructures.

Is porting CUDA to RISC-V crucial for adoption?

Absolutely. CUDA is the dominant software ecosystem for parallel computing and AI, with thousands of optimized applications and libraries. A CUDA port to RISC-V would allow developers to migrate their existing code without major rewriting, significantly reducing the barrier to adoption and accelerating the transition to open architectures in data centers.

Are RISC-V processors actually more energy-efficient?

The customization allowed by RISC-V enables targeted optimizations for specific workloads, eliminating superfluous elements present in general-purpose architectures. This approach can indeed improve energy efficiency, although gains depend heavily on the specific implementation and the suitability of the architecture for the application. Efficiency promises must be validated by real-world, large-scale deployments.

When can we expect to see RISC-V widely deployed in data centers?

Gradual adoption has already begun with pilot deployments at some hyperscalers. SiFive's massive funding and Nvidia's support accelerate this trajectory, but widespread adoption will take several years for the software ecosystem to mature, for critical applications to be validated, and for existing infrastructures to progressively evolve towards these new open architectures.

RISC-V: The Rise of Open Source Chips for AI and Data Centers

Technologie • written by Nova

5 min read 05/10/2026

RISC-V open source processor chip for data centers and artificial intelligence

Early in 2026, SiFive closed a $400 million Series G funding round. Led by Atreides Management, this oversubscribed round brought together Nvidia, Apollo Global Management, Point72 Turrow, T. Rowe Price, and existing investors. The Californian startup's valuation now stands at $3.65 billion. This financial operation sends a strong signal: the RISC-V architecture is establishing itself as a credible alternative to proprietary standards in high-performance computing infrastructures.

The capital injection has a precise objective: to make RISC-V the open standard for processors dedicated to artificial intelligence in data centers. SiFive plans to expand its global engineering teams, accelerate the development of its scalar, vector, and matrix CPU IPs, and consolidate its software ecosystem with Red Hat Enterprise Linux, Ubuntu, and an anticipated CUDA port.

Open Architecture vs. Proprietary Giants

Unlike Intel's and AMD's x86 architectures or licensed Arm, RISC-V is based on an open-source instruction set. This characteristic allows companies to design custom processors without paying royalties or depending on a single vendor. For hyperscalers – the cloud giants operating thousands of servers – this flexibility represents a strategic advantage.

SiFive is capitalizing on this opportunity by offering solutions tailored for agentic AI workloads, these new models capable of orchestrating complex tasks autonomously. The Performance P870-D processor for data centers and the XM series of AI accelerators illustrate this ambition. These products directly integrate Nvidia's NVLink Fusion interconnect, enabling high-bandwidth coupling between customizable RISC-V CPUs and Nvidia GPUs.

Illustration: RISC-V: The Rise of Open Source Chips for AI and Data Centers - Technology

Nvidia's Strategic Shift

Nvidia's participation in this funding round is noteworthy. The GPU manufacturer, long an advocate of proprietary architectures, now seems to be betting on the open ecosystem. By supporting SiFive, Nvidia facilitates the integration of its accelerators with RISC-V processors, thus offering hyperscale customers an alternative to closed solutions.

This strategy responds to growing market demand. According to Forbes, hyperscalers are demanding energy-efficient and architecturally flexible processors to manage expanding AI workloads. The native integration between RISC-V CPUs and Nvidia GPUs via NVLink Fusion materializes this vision of a hybrid ecosystem.

Patrick Little, CEO of SiFive, confirms: "Hyperscale customers have made it clear that it's time to accelerate the availability of open-standard alternatives." This statement, reported by Data Center Dynamics, underscores the urgency perceived by major cloud players.

Software Maturity: The Critical Challenge

Hardware power alone is not enough. To compete with x86 and Arm, RISC-V must prove its software robustness. SiFive is investing heavily in this ecosystem, with support for professional Linux distributions and, crucially, a port of CUDA, Nvidia's parallel computing platform widely adopted in machine learning.

This CUDA compatibility is a decisive issue. It would allow developers to use their existing libraries and tools on RISC-V architectures, significantly reducing adoption barriers. Combined with optimization efforts on Red Hat and Ubuntu, this software development transforms RISC-V from an academic curiosity into a viable industrial solution.

The expansion of engineering teams announced by SiFive precisely targets this maturation. Beyond silicon, the battle is fought in compilers, mathematical libraries, and AI frameworks. This holistic approach is reminiscent of the one adopted by AWS with its Arm-based Graviton processors, which have gradually gained credibility through a rich software ecosystem.

Data Centers and AI: Fertile Ground for RISC-V

The emergence of agentic AI is reshaping infrastructure needs. These autonomous systems require architectures capable of simultaneously orchestrating intensive computation, high memory bandwidth, and low latency. CPUs play a central role in this orchestration, coordinating specialized accelerators and managing complex data flows.

SiFive positions its RISC-V solutions as particularly well-suited to these constraints. The customization allowed by the open architecture enables optimizations impossible with standard chips. A hyperscaler can thus design a processor tailored for its specific workloads, integrating only the necessary functionalities and maximizing energy efficiency.

This energy efficiency becomes critical as data center power consumption explodes. Custom RISC-V architectures, stripped of superfluous elements, promise substantial gains in watts per operation – a decisive metric for data center operators.

Connections with high-performance storage and ultra-fast connectivity technologies like USB4 v2 and Thunderbolt 5 illustrate the evolving hardware ecosystem. Open architectures like RISC-V integrate naturally into this heterogeneous technological mosaic.

The RISC-V Ecosystem Beyond SiFive

SiFive is not alone in this market. Several players are developing RISC-V solutions for high-performance computing. However, the unique combination of massive funding, partnership with Nvidia, and already available products positions SiFive at the forefront.

Recent launches demonstrate this lead:

Performance P870-D: high-performance data center processor
XM Series: dedicated AI accelerators with NVLink Fusion support
Software Ecosystem: RHEL, Ubuntu support, CUDA port in progress

This product strategy contrasts with other RISC-V initiatives that have remained at the academic prototype stage. By directly targeting hyperscalers with production-ready solutions, SiFive accelerates the transition to open infrastructures.

The investment from Apollo Global Management, a major asset manager, suggests that the financial sector anticipates significant adoption. The presence of T. Rowe Price Investment Management reinforces this interpretation: beyond technological enthusiasm, the economic fundamentals appear solid.

Challenges and Prospects

Despite these advances, RISC-V faces considerable obstacles. The inertia of the x86 and Arm ecosystems, built on decades of optimization, does not dissipate overnight. Critical applications require years of validation before migration.

Fragmentation represents an inherent risk to open standards. Without unified governance, different RISC-V implementations could diverge, compromising software compatibility – precisely the competitive advantage sought. RISC-V International, the organization managing the specification, will need to maintain consistency while encouraging innovation.

The question of intellectual property also raises concerns. While the architecture is open, specific implementations from SiFive and competitors remain proprietary. Patents accumulated around RISC-V could create legal gray areas, particularly in a tense geopolitical context.

Nevertheless, the momentum seems favorable. Nvidia's commitment, a key player in AI, confers crucial legitimacy. The $400 million funding provides SiFive with the visibility needed for long-term developments and customer investments. The $3.65 billion valuation reflects market confidence in this trajectory.

Comparison of Processor Architectures for Data Centers

Characteristic	RISC-V	x86 / Arm
License	Open Source	Proprietary (Intel, AMD, Arm)
Customization	High (no royalties)	Limited (vendor-dependent)
Royalty Cost	None	Yes
Vendors	Multiple companies (e.g., SiFive)	Generally single (Intel, AMD, Qualcomm etc.)