NVIDIA is transitioning from its Blackwell AI GPUs to next-generation Rubin GPUs as it prepares for increased AI computing demand. In 2025, Blackwell shipments are expected at 5.2 million units, dropping to 1.8 million in 2026, while Rubin GPU deliveries are projected to surge to 5.7 million units. This shift reflects NVIDIA’s strategy to maintain AI market dominance.
How Many NVIDIA Blackwell GPUs Are Expected to Ship in 2025 and 2026?
JP Morgan projects NVIDIA to ship 5.2 million Blackwell AI GPUs in 2025. By 2026, this number is forecasted to decrease to 1.8 million units as the company ramps up production of its next-generation Rubin GPUs. The reduction in Blackwell shipments is primarily due to a strategic shift toward Rubin, which offers higher performance and improved AI capabilities.
Why Is NVIDIA Shifting from Blackwell to Rubin GPUs?
The shift from Blackwell to Rubin is driven by NVIDIA’s goal to stay competitive in the rapidly evolving AI GPU market. Rubin GPUs support next-generation HBM4 memory and are optimized for AI workloads, providing faster processing, lower latency, and higher energy efficiency compared to Blackwell GPUs. This transition ensures NVIDIA maintains its leadership position while addressing increasing AI demand.
How Will AMD’s Instinct MI400 Affect NVIDIA’s AI GPU Market?
AMD’s Instinct MI400 accelerators, featuring HBM4 memory, will increase competition in the AI GPU segment. While NVIDIA currently dominates the AI GPU market, Rubin’s superior performance is critical for countering AMD’s growing presence. Industry analysts expect a more competitive landscape in 2026, driving innovation and faster adoption of Rubin GPUs.
What Are the Forecasted Rubin GPU Shipments for 2026?
Rubin GPU shipments are projected to reach 5.7 million units in 2026, reflecting NVIDIA’s strategic focus on next-generation AI hardware. Alongside Rubin, NVIDIA plans to ship approximately 1.5 million Vera CPUs, enhancing the overall AI server ecosystem and providing enterprises with scalable computing solutions.
| GPU Series | 2025 Shipments | 2026 Shipments |
|---|---|---|
| Blackwell AI | 5.2M | 1.8M |
| Rubin AI | N/A | 5.7M |
When Could Rubin GPUs Impact the AI Hardware Market?
Rubin GPUs are expected to be available in the second half of 2025, with full market penetration in 2026. Their introduction will accelerate AI workload adoption, particularly in data centers, cloud computing, and enterprise AI applications. Companies like WECENT are preparing to provide tailored Rubin solutions for high-performance AI deployments worldwide.
WECENT Expert Views
“NVIDIA’s transition from Blackwell to Rubin represents a major milestone for AI computing. Rubin GPUs combine cutting-edge HBM4 memory with optimized architecture, providing enterprises with powerful AI solutions. At WECENT, we see strong demand from clients seeking scalable, high-performance servers. By integrating Rubin GPUs into their infrastructure, organizations can achieve superior AI processing speeds and energy efficiency, future-proofing their IT ecosystems.”
How Can Businesses Benefit from Rubin GPU Adoption?
Businesses can leverage Rubin GPUs for accelerated AI workloads, including machine learning, data analytics, and high-performance computing. Rubin’s efficiency reduces operational costs while enhancing performance. WECENT provides consultation, deployment, and maintenance services, ensuring enterprises maximize ROI and maintain secure, reliable AI infrastructure.
| Deployment Area | Benefits |
|---|---|
| Data Centers | Faster AI processing, reduced energy costs |
| Cloud AI | Scalable GPU resources, improved reliability |
| Enterprise Servers | Optimized workload performance, long-term support |
Conclusion
NVIDIA’s AI GPU roadmap highlights a strategic shift from Blackwell to Rubin to sustain market leadership. Rubin GPUs offer superior performance, energy efficiency, and scalability, making them ideal for enterprises and data centers. WECENT supports businesses in implementing these next-generation solutions, ensuring reliable AI infrastructure and a competitive edge in AI-driven markets.
Frequently Asked Questions
What is the main difference between Blackwell and Rubin GPUs?
Rubin GPUs feature next-gen HBM4 memory and improved AI processing, while Blackwell GPUs are previous-generation hardware with lower performance efficiency.
When will Rubin GPUs be widely available?
Rubin GPUs are expected to roll out in the second half of 2025, with widespread adoption throughout 2026.
Can existing servers be upgraded to Rubin GPUs?
Yes, many enterprise servers can be retrofitted or upgraded with Rubin GPUs, and WECENT provides expert support for seamless integration.
Who benefits most from Rubin GPU adoption?
Data centers, cloud providers, and enterprises running AI and high-performance computing workloads gain the most from Rubin GPU deployment.
Are Rubin GPUs compatible with Vera CPUs?
Yes, Rubin GPUs are optimized to work alongside Vera CPUs, offering a complete high-performance AI server solution.
What is the NVIDIA AI GPU shipment forecast for 2025?
In 2025, NVIDIA’s Blackwell architecture is expected to lead, with approximately 5.2 million AI GPUs (B100/B200) shipped. This high-volume deployment supports large-scale AI workloads in cloud and enterprise environments, while the next-generation Rubin platform remains in limited early availability for late 2025 testing.
How will NVIDIA GPU shipments change in 2026?
By 2026, Blackwell shipments are projected to drop to around 1.8 million units as NVIDIA transitions to the Rubin (R100) architecture. Rubin GPU shipments are expected to reach 5.7 million units, driven by higher performance, energy efficiency, and growing demand for advanced AI computing in commercial and cloud applications.
What are the performance differences between Blackwell and Rubin GPUs?
Rubin GPUs use next-generation HBM4 memory and a 3nm process, delivering roughly 5× faster inference and 3.5× faster training performance than Blackwell. Rubin also features 60% more transistors and up to 10× more throughput per watt, targeting agentic AI workloads and enabling higher efficiency for AI servers and advanced compute systems.
When will Rubin GPUs become widely available?
Rubin GPUs are expected to be available in limited quantities during the second half of 2025, with widespread deployment starting in 2026. This phased rollout ensures Blackwell remains dominant in 2025 while Rubin scales up to meet demand for next-generation AI workloads in enterprise and cloud environments.
What additional NVIDIA hardware complements Rubin GPUs?
Alongside Rubin GPUs, NVIDIA plans to ship about 1.5 million Vera CPUs in 2026. These CPUs are designed to integrate with the Rubin ecosystem, enhancing AI server performance, energy efficiency, and supporting agentic AI workloads in combination with Rubin GPUs for next-generation computing infrastructure.
How is NVIDIA preparing for increased Rubin GPU demand?
To meet projected Rubin demand, TSMC is expanding CoWoS (Chip-on-Wafer-on-Substrate) capacity to 120,000–130,000 wafers per month by late 2026. This increase supports mass production of Rubin GPUs and enables NVIDIA to deliver high-volume AI servers to cloud providers, enterprises, and data centers worldwide.
What is the focus of Blackwell GPUs?
Blackwell, dominant in 2025, is optimized for large-scale AI computing. Systems like the GB200 NVL72 are standard for major cloud providers, providing reliable performance for inference and training workloads. Blackwell remains crucial in 2025 while Rubin scales up, bridging the transition to next-generation AI platforms.
Why is Rubin considered NVIDIA’s next-generation AI platform?
Rubin is designed for “agentic AI” with superior energy efficiency, higher transistor count, and advanced throughput per watt. Its combination of HBM4 memory, 3nm process, and integrated architecture allows up to 10× better performance per watt than Blackwell, making it the foundation for NVIDIA’s 2026 AI server ecosystem.