NVIDIA Denies Rubin and Kyber Delay Rumors, Says AI Chip Roadmap Remains on Schedule
NVIDIA has pushed back against recent claims that its next-generation AI chip plans are facing delays, stating that its roadmap remains unchanged. The company says its upcoming Rubin, Rubin Ultra, and Kyber rack products are still progressing as planned, despite speculation suggesting otherwise.
The rumors centered on NVIDIA’s future AI infrastructure, particularly Kyber racks powered by Rubin Ultra GPUs. Some reports claimed that these systems had been delayed until 2028, while others suggested NVIDIA had changed its Rubin Ultra design from a quad-die configuration to a dual-die solution. NVIDIA has now denied that there has been any change to its official roadmap.
Rubin is one of NVIDIA’s most important upcoming AI architectures and is expected to play a major role in the company’s data center strategy. The standard Rubin chips are planned for deployment in Oberon rack systems, while the more powerful Rubin Ultra GPUs are intended for Kyber racks. These high-end systems are expected to scale up significantly, with configurations such as the NVL576 rack designed to support as many as 576 GPUs.
That level of GPU density is aimed at next-generation artificial intelligence workloads, including large language model training, inference, scientific computing, robotics, and advanced data center acceleration. As demand for AI compute continues to rise, NVIDIA’s future rack-scale platforms are expected to be a key part of its strategy to maintain leadership in the AI hardware market.
NVIDIA has previously shown a roadmap that includes Blackwell, Rubin, and Feynman architectures across the 2024, 2026, and 2028 timeframes. These platforms are tied to evolving NVLink technologies, new CPUs, and increasingly powerful GPU designs. According to NVIDIA’s latest response, that schedule remains intact.
The company’s quick denial is notable. Rumors about delays can affect investor confidence, customer planning, and partner relationships, especially in the AI server market where major cloud providers and enterprise customers make long-term purchasing decisions. By responding quickly, NVIDIA appears to be trying to reassure the market that its AI product pipeline remains stable.
This is not the first time NVIDIA’s AI chips have been surrounded by speculation. Similar claims previously appeared around Blackwell and Blackwell Ultra, with rumors suggesting design problems, thermal challenges, or compatibility issues with high-bandwidth memory. Those claims were later dismissed as NVIDIA moved ahead with production and delivered early samples to partners.
Rubin also faced earlier speculation, but NVIDIA later confirmed progress on volume production and delivery timelines. The company’s messaging has remained consistent: its next-generation AI hardware is moving forward, and customer shipments are aligned with its public roadmap.
NVIDIA’s advantage is not only in GPU hardware. The company continues to benefit from its CUDA and CUDA-X software ecosystem, which remains one of the strongest barriers to competition in AI computing. These software tools help developers optimize machine learning, high-performance computing, simulation, and data processing workloads across NVIDIA hardware.
While rivals continue to promote competing AI accelerators with claims of lower total cost of ownership, better efficiency, or improved performance in specific workloads, NVIDIA remains the dominant force in many widely used AI benchmarks and real-world deployments. Its combination of GPUs, networking, rack-scale systems, and mature software gives it a broad platform advantage.
For now, NVIDIA says there is no delay to Rubin or Kyber. The company’s AI chip roadmap remains on track, and its future platforms are still expected to play a central role in the next wave of AI data center expansion.
