NVIDIA Explores CoWoP Packaging for Next-Gen Rubin GR150 AI GPUs

NVIDIA's Exploration of CoWoP Packaging

NVIDIA, a leader in AI and graphics processing technology, is reportedly considering a significant shift in its packaging technology for its next-generation AI GPUs. The company is exploring the adoption of Chip-on-Wafer-on-Platform (CoWoP) PCB packaging, particularly for its upcoming Rubin GR150 AI GPUs 1

1

2

Understanding CoWoP Technology

CoWoP is an advanced packaging technology that removes the package substrate and connects the interposer directly to the motherboard. This approach offers several potential benefits over the current Chip-on-Wafer-on-Substrate (CoWoS) technology, which has been the industry standard for high-performance computing (HPC) and AI chips for nearly 14 years 2

2

Source: Wccftech

Key Benefits of CoWoP

The potential advantages of CoWoP include:

1. Improved signal and power integrity
2. Reduced substrate losses
3. Closer voltage regulation to the main GPU die
4. Enhanced NVLINK IC capabilities
5. Direct contact between thermal solutions and silicon, eliminating the need for a package lid
6. Potential cost reductions due to the elimination of the package lid 1

   1

   2

   2

NVIDIA's Testing Roadmap

NVIDIA has reportedly begun early testing of CoWoP technology:

1. July 2025: Initial tests with a dummy GB100 GPU/HBM solution
2. August 2025: Functional GB100 CoWoP testing, evaluating manufacturability, structural and electrical functionality, thermal design, and NVLINK interface throughput
3. 2026: Testing with Rubin GR100 chips using CoWoP packaging in an SXM8 form factor
4. Late 2026: Production-ready Rubin GR150 CoWoP solution expected
5. 2027: Anticipated availability of Rubin GR150 GPUs 1

   1

   2

   2

Related Stories

Nvidia Shifts to Advanced CoWoS-L Packaging for Next-Gen AI Chips, Impacting TSMC and Supply Chain

NVIDIA Accelerates Next-Gen 'Rubin' AI GPU Launch, Featuring Advanced 3nm Process and HBM4 Technology

Nvidia unveils Vera CPU with 88 Arm cores to power next-generation AI infrastructure

NVIDIA's Thermal Challenge: Innovative Cooling Solutions for Next-Gen AI GPUs

Challenges and Considerations

While CoWoP offers potential benefits, it also presents significant challenges:

1. High initial costs for adopting the new technology
2. Complexity in setting up a new supply chain
3. Design changes required for existing motherboards
4. Potential production bottlenecks and cost increases

It's worth noting that NVIDIA is not planning to abandon CoWoS entirely. The company is likely to utilize both CoWoS and CoWoP technologies moving forward, depending on specific product requirements and market conditions 1

1

2

Source: TweakTown

Industry Perspectives

Despite NVIDIA's exploration of CoWoP, some industry analysts remain skeptical about its immediate adoption. Morgan Stanley and several investor firms have expressed doubts about NVIDIA implementing CoWoP in its next-generation GPUs. The final decision will likely depend on various factors, including market trends and supply/demand dynamics 2

2

As the AI and HPC markets continue to evolve rapidly, NVIDIA's potential shift to CoWoP packaging represents a significant development in the pursuit of more efficient and powerful AI GPUs. The success of this transition could have far-reaching implications for the future of AI hardware and NVIDIA's position in the market.