NVIDIA Vera Rubin AI Platform Targets Q3 2026 Shipments as Blackwell Demand Pushes GPU Prices Higher
NVIDIA has confirmed the timeline for its next-generation Vera Rubin AI platform, setting the stage for another major leap in AI computing infrastructure. While the company’s current Blackwell systems continue to dominate demand across hyperscalers, cloud providers, and frontier AI model developers, NVIDIA is already preparing for its next wave of data center growth.
During its latest earnings update, NVIDIA reported record quarterly revenue of $81.6 billion, an 85% increase from the same period last year. The company’s Data Center business remained the clear growth engine, generating $75.24 billion in revenue. Edge Computing contributed another $6.3 billion, showing that demand for NVIDIA’s AI hardware is expanding across multiple markets.
The numbers highlight just how central NVIDIA has become to the global artificial intelligence buildout. Hyperscale customers accounted for $37.87 billion in revenue, while AI cloud, industrial, and enterprise customers generated $37.38 billion. Data Center revenue rose 92% year over year, driven by enormous demand for AI training and inference systems.
Blackwell remains NVIDIA’s biggest growth driver
NVIDIA’s Blackwell platform, including the GB300 and NVL72 systems, continues to see exceptional demand. The company said hundreds of thousands of Blackwell GPUs have already been deployed worldwide, with frontier AI labs and hyperscale cloud operators leading adoption.
According to NVIDIA CFO Colette Kress, Blackwell has become the fastest product ramp in the company’s history. A major reason is its performance in AI inference, where NVIDIA says Blackwell offers the lowest token generation cost. That matters because inference, the process of running AI models after they have been trained, is becoming one of the largest and fastest-growing areas of AI infrastructure spending.
NVIDIA also noted that the number of partner data centers exceeding 10 megawatts has nearly doubled in just one year, now surpassing 80 sites. This reflects the scale of the AI data center race, as companies rush to deploy massive GPU clusters capable of supporting advanced generative AI, reasoning models, and agentic AI applications.
Older NVIDIA GPUs are becoming more expensive
Even as Blackwell becomes the flagship platform, demand for previous-generation NVIDIA GPUs remains extremely strong. The ongoing shortage of AI compute capacity has pushed prices higher for older chips such as Hopper H100 and Ampere A100.
NVIDIA said the cost of renting H100 GPUs has risen around 20% year to date, while A100 cloud pricing is up nearly 15%. This trend shows that AI companies are not only chasing the newest hardware but are also relying heavily on existing GPU fleets to meet immediate compute needs.
NVIDIA CEO Jensen Huang has described this effect as a kind of “fine wine” trend, where older GPUs gain value because AI demand continues to outpace supply. Cloud providers are making more of their existing hardware available to AI startups and enterprises, but demand remains intense across training and inference workloads.
Vera Rubin is NVIDIA’s next major AI platform
While Blackwell is still ramping, NVIDIA is already turning attention to Vera Rubin, its next-generation AI platform. Vera Rubin is designed as part of NVIDIA’s broader Extreme Co-Design strategy, combining GPUs, CPUs, networking, memory, and software into a tightly integrated AI infrastructure platform.
The platform is expected to play a major role in powering what NVIDIA calls agentic AI factories. These are large-scale AI systems built to support autonomous AI agents, advanced reasoning models, automated workflows, and next-generation enterprise AI services.
NVIDIA confirmed that Vera Rubin is on track for initial shipments in Q3 2026. A larger volume ramp is expected in Q4 2026, followed by stronger shipment growth through the first half of 2027.
The company is also preparing its Vera CPUs for deployment in rack-scale systems. Early shipments of Vera CPUs have already gone to major AI and cloud customers, including OpenAI, Oracle, Anthropic, and other large-scale infrastructure partners.
Jensen Huang says Vera Rubin is off to a strong start
Jensen Huang said NVIDIA’s annual product cadence remains one of the company’s biggest competitive advantages. After Vera Rubin, NVIDIA plans to introduce Rubin Ultra in 2027 and Feynman in 2028, keeping the company on a rapid release schedule for AI infrastructure platforms.
Huang also emphasized that inference is becoming a major growth opportunity for NVIDIA. As more companies deploy AI applications at scale, the need for efficient inference hardware is rising quickly. He said NVIDIA is gaining inference market share rapidly and expects Vera Rubin to be even more successful than Grace Blackwell.
According to Huang, nearly every frontier AI model company is expected to adopt Vera Rubin from the beginning. That would mark a stronger launch position than Blackwell had at its debut, suggesting that demand for Vera Rubin could be extremely high once shipments begin.
AI infrastructure demand shows no sign of slowing
NVIDIA’s latest results show that the AI hardware boom remains in full force. Blackwell is delivering record revenue, older Hopper and Ampere GPUs are seeing price increases, and the company’s next-generation Vera Rubin platform is already attracting significant interest from major AI players.
The combination of training, inference, cloud AI services, and agentic AI workloads is creating a massive and growing market for NVIDIA’s data center products. With Vera Rubin scheduled for Q3 2026 shipments and a broader ramp expected soon after, NVIDIA is positioning itself to remain at the center of the AI infrastructure race.
As demand for artificial intelligence compute continues to climb, NVIDIA appears ready for the next phase. Blackwell is driving today’s growth, but Vera Rubin is shaping up to be the platform that powers the next generation of AI factories.
