Nvidia Broadens CUDA Reach with RISC-V Embrace

Nvidia Expands CUDA Support to RISC-V, Signaling Strategic Push into China’s Growing Chip Market

By Archyde Staff

July 26, 2025

Nvidia, the dominant force in AI hardware, is making a strategic move to broaden the reach of its powerful CUDA software stack by extending support to the RISC-V instruction set. This progress, announced at the RISC-V Summit in Shanghai, is seen as a significant step in capitalizing on China’s burgeoning interest in open-source processor architectures.

While RISC-V cores have long been integrated into Nvidia’s GPUs – with estimates suggesting over a billion such cores are present across the company’s product line, typically between 10 and 40 per GPU – this latest announcement signifies a deeper software-level integration.CUDA is the linchpin for developers to harness the computational power of Nvidia’s Graphics Processing Units (GPUs), and its availability on a new instruction set opens up new avenues for hardware designers and software engineers.

On the surface, extending CUDA support to RISC-V might appear as a natural progression, mirroring its existing compatibility with x86 and Arm-based cpus, both of which employ RISC principles. From Nvidia’s technical standpoint, the integration is not expected to necessitate major architectural overhauls.

China: A Lucrative Frontier for RISC-V

The critical question surrounding this announcement revolves around its timing and context. Nvidia’s decision to align CUDA with RISC-V, particularly at an event hosted in Shanghai, underscores its strategic focus on the Chinese market. In recent years, China has intensified its efforts to reduce its reliance on Western CPU technologies, with RISC-V emerging as a central pillar in this ambitious endeavor. While some Western tech companies have scaled back their RISC-V initiatives, Chinese firms like Alibaba remain deeply invested in the open-source architecture.

Furthermore, this move aligns with recent permissions granted to Nvidia to sell its H20 AI chips in China. By facilitating the use of its flagship AI processors with the increasingly popular RISC-V instruction set, Nvidia could present a compelling value proposition to Chinese customers. Despite its current multi-trillion-dollar valuation, Nvidia is clearly seeking to unlock further growth opportunities, and its strategy around RISC-V in China is a key element of this expansion.

The broader trajectory of RISC-V adoption outside of China, especially within data centers, remains a subject of keen observation. While RISC-V’s open-source and royalty-free nature are significant advantages, its maturation for demanding workloads is still an ongoing process. Reports suggest that new RISC-V projects are in development that could perhaps rival Arm’s established presence. Nvidia’s CUDA support could thus act as a significant catalyst, accelerating RISC-V’s progress and its ability to compete with incumbent instruction sets, particularly in the high-growth AI sector.

How might Nvidia’s RISC-V adoption impact the long-term cost of developing and deploying AI/ML applications?

The Shift Towards Open Architectures

Nvidia, traditionally known for its proprietary CUDA platform and dominance in GPU computing, is making significant strides in embracing the open-source RISC-V instruction set architecture (ISA). This move signals a potential reshaping of the high-performance computing (HPC) landscape and offers developers increased flexibility and control.For years, CUDA has been the de facto standard for GPU-accelerated computing, notably in fields like artificial intelligence (AI), machine learning (ML), and scientific simulations.However, the closed nature of CUDA has prompted a growing demand for open alternatives, and RISC-V is emerging as a leading contender.

What is RISC-V and Why Does it Matter?

RISC-V (pronounced “risk-five”) is a free and open-source hardware instruction set architecture.Unlike proprietary ISAs, RISC-V allows anyone to design, manufacture, and sell chips based on the architecture without licensing fees. This openness fosters innovation and competition.

Here’s why Nvidia’s embrace of RISC-V is noteworthy:

Reduced Vendor Lock-in: Developers are no longer solely reliant on Nvidia’s ecosystem.

Customization: RISC-V’s modular design allows for tailored hardware solutions optimized for specific workloads.

Innovation: The open-source nature encourages community contributions and faster advancement cycles.

Cost Reduction: Eliminating licensing fees can lower the overall cost of hardware and software development.

Nvidia’s RISC-V Initiatives: A deep Dive

Nvidia’s commitment to RISC-V isn’t a sudden pivot; it’s a phased integration. Several key initiatives demonstrate this:

Cu-Core: Nvidia has announced Cu-Core, a fully programmable RISC-V core designed for data processing units (DPUs). These DPUs are increasingly used for offloading networking, storage, and security tasks from the CPU, freeing up resources for core applications.

Networking and Data Center Focus: Initial RISC-V efforts are heavily focused on networking and data center infrastructure. This is a strategic move,as DPUs are becoming critical components in modern data centers.

Software Support: Nvidia is actively working on porting CUDA libraries and tools to RISC-V, enabling developers to leverage their existing CUDA code on RISC-V hardware. This includes efforts to ensure compatibility with popular frameworks like TensorFlow and PyTorch.

Collaboration: Nvidia is collaborating with other industry leaders and open-source communities to accelerate the development of the RISC-V ecosystem.

Benefits for Developers and Businesses

The integration of RISC-V with Nvidia’s technologies presents several advantages:

Enhanced Performance: Optimized RISC-V cores, coupled with Nvidia’s GPU acceleration, can deliver significant performance gains for specific workloads.

Greater Flexibility: Developers can choose the hardware and software components that best suit their needs,without being constrained by proprietary ecosystems.

Reduced costs: Open-source licensing and increased competition can led to lower hardware and software costs.

Accelerated Innovation: The open-source nature of RISC-V fosters collaboration and faster development cycles.

Improved Security: The transparency of the RISC-V architecture allows for more thorough security audits and vulnerability detection.

CUDA Compatibility and the Future of GPU Computing

A key question is how Nvidia will maintain CUDA compatibility while embracing RISC-V. The company’s strategy appears to be focused on providing tools and libraries that allow developers to port their CUDA code to RISC-V with minimal effort.

Here’s what we can expect:

1. CUDA-to-RISC-V Compilers: Tools that automatically translate CUDA code into RISC-V instructions.
2. Hybrid Architectures: Systems that combine Nvidia GPUs with RISC-V CPUs for optimal performance.
3. Open-Source Libraries: Continued development of open-source libraries that support both CUDA and RISC-V.
4. Ecosystem Growth: A thriving RISC-V ecosystem with a wide range of hardware and software options.

Real-World Applications and Use Cases

While still in its early stages,the combination of Nvidia and RISC-V is already finding applications in several areas:

Data Center Infrastructure: DPUs powered by RISC-V cores are being used to accelerate networking,storage,and security tasks in data centers.

Edge computing: RISC-V’s low power consumption and small footprint make it ideal for edge computing applications.

AI and Machine Learning: RISC-V-based accelerators are being developed to accelerate AI and ML workloads.

Automotive: RISC-V is gaining traction in the automotive industry for applications such as autonomous driving and in-vehicle infotainment.

Practical Tips for Developers

For developers looking to explore the Nvidia-RISC-V ecosystem:

Familiarize yourself with RISC-V: understand the architecture and its benefits. Resources like the RISC-V Foundation website (https://riscv.org/) are excellent starting points.

Experiment with Cu-Core: Explore Nvidia’s Cu-Core and its capabilities.

**Utilize CUDA porting tools